War Stats Do Not Measure Up ‘
Exploring the limits of knowledge with military casualty statistics in Ukraine and other wars, and what we can do to manage uncertainty
DRONES cause 70 percent of the casualties in Ukraine.’ For a few critical months from early- to mid-2025 this was an accepted and commonly repeated statistic, and it lives on in some quarters. This number influenced key Western understanding of modern warfare and, in turn, at least indirectly influenced the debate over new military force design.1 And
like all war statistics, this one about drones in the Ukraine War is ultimately unprovable.2 In an ideal world it would be dropped from public discourse. But in the real world it has traction. For some, it reinforces existing beliefs about drone warfare. For others, it satisfies a broader and unrelenting need to find clarity where clarity cannot reasonably exist.
1See additional sources below and note that 80 percent and even 90 percent figures have also been floated in various venues. Suggested percentages have risen significantly in the first half of 2025, from 70 to more frequent mentions of at least 80 percent. While I focus on this one claim the analysis herein applies to any similar claim. This 70 percent figure has been repeated in, at least: Sean Rayment and Thomas Godfrey, “Killer Drones: Combat drones responsible for almost 80% of all deaths in blood-soaked Ukraine conflict, Sun on Sunday probe reveals,” The Sun, April 12, 2025; David Kirichenko, “Ukraine’s drone wall is Europe’s first line of defense against Russia,” Atlantic Council, July 2, 2025; Mila Tanghe, “Russia Assails Ukraine’s Drone Wall,” Center for European Policy Analysis, June 4, 2025; Greg Myre, “The many ways Ukraine carries out unprecedented drone attacks,” National Public Radio, June 7, 2025; Paul Adams, “Kill Russian soldiers, win points: Is Ukraine’s new drone scheme gamifying war?,” BBC, mid-July, 2025; etc. The Sun lists both a high-end 80 percent number and the 70 percent figure. Most reports appear to circle back to Marc Santora, Lara Jakes, Andrew E. Kramer, Marco Hernandez, and Liubo Sholudko, “A Thousand Snipers in the Sky: The New War in Ukraine,” The New York Times, March 3, 2025; see below. Note that a Royal United Services Institute report states “...tactical UAVs account for 60-70% of damaged and destroyed Russian systems.” This is a ranged estimate specific to systems like vehicles and artillery pieces and not directly to human casualties. Jack Watling and Nick Reynolds, Tactical Developments During the Third Year of the Russo-Ukrainian War, London, UK: Royal United Services Institute, 2025, p. 10. Oddly, another article lists the 70 percent human casualty figure but appears to cite the Watling and Reynolds, 2025 monograph: Jorge Rivero, “Innovating under fire: Lessons from Ukraine’s frontline drone workshops,” Modern War Institute, March 25, 2025.
2As the reader moves through this monograph it is important to keep in mind that classified information may exist that shows far greater detail than is available in public sources. Some of that data inevitably will be far superior to public data. However, relatively better accuracy for a specific datum and greater quantity of data does not necessarily convey overall accuracy or representative value.
In all likelihood, common Western use of the 70 percent figure started with the publication of a New York Times article quoting Ukrainian colonel and parliamentarian Roman Kostenko.3 While the original quote was not directly cited –meeting a loose standard for journalism but not a higher standard for policy knowledge – we can assume it derives from something Kostenko said in early 2025.4 However, in a recorded interview with Ukrinform in April one month after publication of the Times article, Kostenko said: “According to last year’s [2024] statistics, 65 percent of Russian soldiers on the battlefield were killed by Ukrainian drones...”5 He then added that about 75 percent of Russian vehicles were also destroyed by drones.6
Kostenko may or may not have accurately reflected more detailed Ukrainian military statistics in either quote.7 This does not reflect poorly on either Kostenko or the Ukrainian Ministry of Defense: Even the most earnest estimates of this kind by the most powerful militaries in the world rarely stand up to even surface-level scrutiny. In that light the Ukrainians might even be commended for some of their intentional vagueness. American statistics on war are,
as we shall see, particularly misleading in great part because they are so convincingly precise and publicly detailed.8
Assuming my argument is correct –that many war statistics are inherently inaccurate or at least indefensibly precise –why do people continue to seek them out and lend them so much credence? I focus on several key reasons here, including cultural influences and political pressure. I
describe ways in which the psychological need for closure can draw any of us away from the darkness of uncertainty towards a deceptively clarifying light.
Dodging the uncertainty maxim in US military analysis
This antithetical pursuit of clarity is actively discouraged at the surface level of Western military theory and doctrine. States of the North Atlantic Treaty Organization
3Santora, et al., 2025. The video of Kostenko’s interview was posted on YouTube on April 7th, 2025, which also appears to have been the date of airing. It is also possible the NYT reporters averaged the 65 percent and 75 percent figures given here, and that may have been given in a previous interview by Kostenko. If so, merging casualty and vehicle kill statistics would generate further inaccuracies. Importantly, Santora, et al., is cited and linked in other articles as the original source of the 70 percent figure. For example, the NYT references its own article in: Andrew E. Kramer, “Amid Uncertainty About U.S. Support, Ukraine Pins Its Hopes on Innovation,” The New York Times, April 28, 2025. This secondary NYT article is then cited and linked in Tanghe, CEPA, 2025. We can see the vaguely substantiated original figure cemented through repetition as Tanghe links to Kramer, Kramer links to Santora, et al., with Tanghe attributing the third- or fourth-hand 70 percent figure to the Ukrainian military: “... according to the Ukrainian military.” Tanghe, CEPA, 2025.
4There was no link to the original source in Santora, et al., 2025. I searched in both English and Ukrainian for possible sources and did not identify any publicly available, still-online source quoting Kostenko directly on this subject. He is quoted frequently in various social media and news sources on related or other subjects in both English and Ukrainian. It is of course possible that I failed to discover the quote; all online searches are necessarily imperfect. Logical reasoning requires some evidence that the individual statements within the argument are true. So in this case, one could not reasonably arrive at the conclusion that drones cause 70 percent of casualties in Ukraine if one could not show the source of that statistic to be reliably accurate. Journalists often protect their informants, rendering the ultimate source of their claims opaque and therefore of questionable reliability for serious policymaking. In this case they appear to have simply left out a hyperlink.
5Roman Kostenko, interview with Ihor Dolgov, Ukrinform, undated, posted on April 7, 2025. Also see: “Roman Kostenko: Ukrainian drones responsible for 65% of Russian casualties and 75% of destroyed equipment,” Odessa Journal, April 8, 2025.
6Kostenko did not appear to be selling these numbers as something they were not. He mentioned the 65 percent figure in passing within a long and wide-ranging discussion. Overvaluing of the reported statistics appears to have happened on the Western side of the equation.
7I could not find reference to publicly available Ukrainian government statistics that might reveal the details behind this 70 percent figure. On 03 January, 2025, the Ministry of Defense published an uncited press report detailing Russian military losses in 2024. This article does not reference visible or verifiable data: “Понад 430 000 солдатів,
9 000
Ukrainian Ministry of Defense website, January 3, 2025. I also checked: the Brookings Institute Ukraine Tracker; the Stockholm International Peace Research Institute (SIPRI); the Armed Conflict Location & Event Data research from 01 January, 2025; other trackers including Oryx and Warspotting; and medical reporting from both Ukraine and Russia (see below). None of these sources revealed official data on dronecaused casualty percentages that matched the 70 percent figure, though see below for the examination of a widely-cited Russian medical journal report.
8I have written extensively about this subject and offer detailed analysis and references in: Ben Connable, Embracing the Fog of War: Assessments and Metrics in Counterinsurgency, Santa Monica, Calif.: RAND Corporation, 2012.
commonly describe war as complex, chaotic, unpredictable, disorderly, hard to observe and even harder to understand.9 All these common factors can be distilled to a single Western military maxim: War is inherently uncertain
That’s all well and good in theory. But in previous studies on campaign assessment and on will to fight in warfare I argued that in practice the American military barely pays lip service to this maxim of uncertainty.10 When Americans have been thrown into chaotic environments like Vietnam (1960-’70s), Iraq (20032017) and Afghanistan (2001-2021), they have generally abandoned the uncertainty maxim and worked desperately to measure war to Swiss-watch tolerances.
In Embracing the Fog of War: Assessments and Metrics in Counterinsurgency, I describe some of the better- and lesswell-known examples of wartime overquantification from the Vietnam, Iraq and Afghanistan wars.11 In all three of these wars, desperate policymakers put enormous pressure on military staffs to generate quantifiable evidence of progress. They were driven towards unreliable statistics in part by a range of perverse
“When Americans have been thrown into chaotic environments, they have generally abandoned the uncertainty maxim and worked desperately to measure war to Swiss-watch tolerances.”
incentives including political pressures to show headway and short-term public receptiveness to comforting quantification. Three egregious examples of overquantification stand out: body counting; the Hamlet Evaluation System; and SIGACTs [significant activity].
(1) Body counting. American military forces reported enemy casualties in earnest during the stalemate phase of the Korean War as political pressure to
show success mounted.12 Counting of dead enemy bodies evolved into one of two primary measures of success during the Vietnam War.13 In practice, accurate body counting proved to be impossible and the data proved to be valueless. Body counting created perverse incentives that corrupted military integrity and probably led to unnecessary civilian deaths.14 American defeat ensued despite years of ostensibly favorable casualty ratios.15
(2) Hamlet Evaluation System. Measuring population control balanced out body counting: Show we are killing the enemy and also show we are securing support of the people.16 To measure population control, each South Vietnamese hamlet was rated on
9For example, the 2017 capstone doctrine of the American Joint Force describes modern war in unequivocal terms: Despite the evolution of advanced technologies, war has remained chaotic and uncertain. U.S. Joint Chiefs of Staff, Doctrine for the Armed Forces of the United States, Joint Publication 1, Washington, D.C.: Joint Staff, as of July 12, 2017, p. I-3. This version was temporarily released to the public and then removed from the Joint Staff website. I purposefully use the word “frequently” here to avoid furthering unsustainable claims of universality. War as a broad phenomenon may be inherently uncertain, but some instances of war are not necessarily chaotic or uncertain. For example, it is often possible in the Ukraine War to observe Russian armored units attacking over open ground, driving in linear columns, in broad daylight, and with no obscuring smoke. This is a clear and certain situation, even though the tactical solution to that assault might be neither clear nor certain.
10This may apply to other NATO militaries as well. My research focused on the United States so I restrict my observations here. On will to fight see: Ben Connable, Michael J. McNerney, William Marcellino, Aaron Frank, Henry Hargrove, Marek N. Posard, S. Rebecca Zimmerman, Natasha Lander, Jasen J. Castillo, and James Sladden, Will to Fight: Analyzing, Modeling, and Simulating the Will to Fight of Military Units, Santa Monica, Calif.: RAND Corporation, 2018.
11Connable, Embracing the Fog, 2012. This monograph presents a much wider array of quantification issues across these three wars and others, focusing primarily on irregular warfare assessment. It also directly addresses theoretical issues important to understanding the phenomena of over-quantification and need for certainty.
12See: Scott Sigmund Gartner and Marissa Edson Myers, “Body Counts and ‘Success’ in the Vietnam and Korean Wars,” Journal of Interdisciplinary History, Vol. 25, No. 3, Winter 1995, pp. 337-395; and Connable, Embracing the Fog, 2012, pp. 140-146.
13The other was population support or control, commonly referred to by a legacy French colonial term, pacification. Many other measures were applied in concert with body counting, including hamlet evaluations; see below. U.S.G. Sharp and William C. Westmoreland, Report on the War in Vietnam as of 30 June 1968, Washington, D.C.: U.S. Government Printing Office, 1968.
14For example, see conflicting accounts of Operation Speedy Express: Connable, Embracing the Fog, 2012; “Pacification’s Deadly Price,” Newsweek, June 19, 1972, pp. 4243; and Gunter Lewy, America in Vietnam, New York: Oxford University Press, 1978, p. 81.
15Connable, Embracing the Fog, 2012, pp. 140-146; Sharp and Westmoreland, Report on the War, 1968.
16Together, these two objectives form a somewhat sound counterinsurgency policy. However, in practice neither is truly measurable. For more on these approaches see, for example: Ben Connable and Martin Libicki, How Insurgencies End, Santa Monica, Calif.: RAND Corporation, 2010; and Christopher Paul, Colin P. Clarke, and Beth Grill, Victory Has a Thousand Fathers: Detailed Counterinsurgency Case Studies, Santa Monica, Calif.: RAND Corporation, 2010.
a scale ranging from enemy controlled to friendly controlled. There may have been anywhere from 8,000 to 14,000 hamlets; the total number was never verified. A hamlet could have 50 or up to 20,000 people. Most hamlets had never been visited let alone carefully observed. Nonetheless, every known hamlet was rated on a precise scale each month, generating a misleadingly precise overall pacification statistic. These reports epitomized the concept of precision without accuracy.17
(3) SIGACTs. Significant activity data recorded daily incident reports from the Iraq and Afghanistan wars.18 A report could describe anything from a vehicle accident to an ambush to a full-scale battle. Unit leaders were responsible for submitting reports on all significant activities. In practice, reporting was uneven, sometimes skipped, sometimes
overstated and often inaccurate. Data quality worsened as reports were repeated and re-typed up the line to headquarters. Ultimately, SIGACTs did not tell an accurate story of either war. Nonetheless, they were used to inform strategy.19
Modern Western conventional war analysis also centers on sports-like statistic tracking. Success in the 2011 Libya intervention under Operation Odyssey Dawn was measured in the number of targets struck by bombs and missiles.20 Success in the 2020 Nagorno-Karabakh
war was measured primarily by reported drone kills, which could be seen and counted in readily available online videos.21 And Americans commonly measure relative progress in the Ukraine War by using comparative vehicle kill data or square kilometers of ground controlled. Relentless pursuit of refined measurement is the center-rest American mindset.
Drivers of quantification and the uncertainty of war
At least three broad factors influence this metrics-fixated mentality. First, the general psychological need for certainty – or at
“Americans commonly measure relative progress in the Ukraine War by using comparative vehicle kill data or square kilometers of ground controlled.”
17For a detailed accounting of HES and a long list of citations see: Connable, Embracing the Fog, 2012, chapter 7. Accuracy is the closeness of a measurement to a known value, while precision is the closeness of two not-necessarily-grounded measurements.
18For more on SIGACTs see Connable, Embracing the Fog, pp. 157-166, and sources cited in that section.
19I am focused here on the broader strategic understanding of war, and on the approach researchers and analysts take to support strategic decision-making. It is important to note here that detailed tactical information and data can be and have been used to effectively improve tactical action and reduce combat risk. For example, the Joint Improvised Explosive Device Defeat Organization (JIEDDO, later with new names and acronyms) and Task Force Paladin both used SIGACT data to help forecast attacks, optimize movements and reduce risks. See any number of online references for both organizations, including the hyperlinks here. Tactical value with limited data applied with an understanding of relative certainty is not the same as complex policy understanding of war-wide trends or statistics.
20For example, see the flurry of numbers and measures in: Hailey Staker, “28th BW reflects on Odyssey Dawn launch,” press report, 28th Bomb Wing Public Affairs, U.S. Air Force, March 28, 2016.
21For example, see: John F. Antal, 7 Seconds to Die: A Military Analysis of the Second Nagorno-Karabakh War and the Future of Warfighting, Havertown, Penn.: Casemate Publishers, 2022.
least for comforting approximation – often motivates sorting, tracking and counting behaviors. Second, predilection to quantify and measure is a deeply embedded Western cultural norm. And third, military professionals and politicians are most likely to generate and fix on performative statistics when military progress is in question.
Need for closure and comforting clarity
Psychological understanding of the mind changes over time, but there is a general acceptance of a human need for closure. Probably the best article on this subject, “Motivated Closing of the Mind: ‘Seizing’ and ‘Freezing,’” by Arie Kruglanski and Donna Webster, describes the need for closure as the pursuit of “a firm answer to a question and an aversion toward ambiguity”.22 Closing in this context means closing the uncertainty gap.
Kruglanski and Webster are careful to point out that need for closure is not universal.23 People can also be motivated to avoid certain kinds of knowledge. Sometimes knowing is uncomfortable, like when new knowledge challenges a closely held belief. For example, the great Italian astronomer Galileo Galilei was tried and punished for publicly dismantling the contemporaneous view that the sun revolved around the Earth. His new scientific knowledge challenged frozen and deeply personalized beliefs, and his claims in turn generated fear, anger and ill-advised rejection of the most logical scientific conclusion.
“Sometimes knowing is uncomfortable, like when the new knowledge challenges a closely held belief. For example, Galileo was tried and punished for publicly dismantling the contemporaneous view that the sun revolved around the Earth.”
People under pressure to know are far more likely to seize on any reasonable answer, even if that answer is a radically biased and misleadingly precise guesstimation. In doing so they collectively harden the perceived value of what should be understood as soft guesses. Often, they trade accuracy for speed. As I describe below, wartime pressures are particularly
acute and often drive even the most thoughtful leaders and analysts towards hasty acceptance of half-baked and often misleadingly precise statistics.
Inclination to quantify and measure
General human compunction towards measurement is compounded by cultural proclivities in the Western world.24 Two books effectively capture the Western obsession with quantification. James Vincent’s more accessible Beyond Measure traces the global human history of measurement.25 Vincent’s book sets the context for Alfred W. Crosby’s more academic treatment of Western measurement culture, The Measure of Reality 26
Crosby argues that quantification in Europe started as a mindset and evolved through forced practice into a complex net of interrelated cultural norms, values, symbols and other influences. War and capitalism were probably the two primary drivers of quantification and measurement in Europe. Proximity, diversity and population density on the continent intensified human interaction and gave competitive advantage to societies that could out-think and outpace adversaries.
This measurement-centric view of Western history is fairly straightforward, and it translates directly to both military and national policy-making. In short and using basic examples: If you build a
22Arie W. Kruglanski and Donna M. Webster, “Motivated Closing of the Mind: ‘Seizing’ and ‘Freezing,’” Psychological Review, Vol. 103, No. 2, 1996, pp. 263-283, p. 264. While I recommend reading Kruglanski and Webster, there is a broad and shifting literature on the need for closure. Nothing they have written is inherently right, and probably several of their insights and arguments have since been improved upon or refuted. Consider this a good starting point for deeper review. Also worth reading: Christian D. Schunn and J. Gregory Trafton, “The Psychology of Uncertainty in Scientific Data Analysis,” book chapter, Applied Psychologies of Science; Rasmus Bruckner, Hauke R. Heekeren, Matthew R. Nassar, “Understanding learning through uncertainty and bias,” Communications Psychology, Vol. 3, No. 24, 2025; Paul D. Windschitl and Gary L. Wells, “Measuring Psychological Uncertainty: Verbal Versus Numeric Methods,” Journal of Experimental Psychology, Vol. 2, No. 4, 1996, pp. 343-364; Nathan J. Evans, Babette Rae, Maxim Bushmakin, Mark Rubin, and Scott D. Brown, “Need for closure is associated with urgency in perceptual decision-making,” Memory and Cognition, No. 45, 2017, pp. 1193-1205.
23In fact there are many nuances and inherently contradictory components within the need-for-closure concept. See Kruglanski and Webster for more on, for example: urgency and permanence tendencies, consensus and consistency strivings, the point of belief crystallization, personal versus situational influences, etc. Read Evans, et al., “Need for closure,” 2017, to understand the speed-accuracy trade-off.
24In this section I simplify my explanation of regional and national cultures to help describe broad influences on behavior in war. In practice, culture is a complex and generally unbounded phenomenon. National cultures are useful descriptive constructs and nothing more. A person who happens to live in the United States does not have their behavior determined by an American national culture. In practice, individuals and groups interpret identity in many different and dynamic ways. My intent here is to introduce the idea that cultural norms, values, icons and related factors have some influence on the behavior of particularly American military personnel and politicians.
25James Vincent, Beyond Measure: The Hidden History of Measurement from Cubits to Quantum Constants, New York, NY: W.W. Norton & Company, Inc., 2022.
26Alfred W. Crosby, The Measure of Reality: Quantification and Western Society, 1250-1600, Cambridge, UK: Cambridge University Press, 1996. I would read Vincent before reading Crosby.
better artillery cannon than your enemy and reproduce it in a factory to specific tolerances through precise and accurate measurement, you can win more wars. If you design a faster sailing ship guided by relatively more accurate navigational aids you can trade more safely and efficiently than your competitors and make more money. More money allows for stronger military forces – more and better cannons, etc. – which in turn allows your society and culture to dominate and grow. Over time, societies that successfully quantified and measured spread the dominant continental norms, values and symbols that in turn formed the globally dominant metricsfixated European, and then over time, American culture.
Building from European culture, Americans did what they do best: They amplified everything they believed to be useful with a collective intensity probably unmatched in human history. From a broadly American perspective, if quantification and measurement were generally good then extreme calculating must be even better.
Moneyballing best epitomizes this way of thinking in American culture.27 Exhaustive measurement of every observable characteristic and phenomenon presumably offers a crucial edge in the highly competitive worlds of American sports, business, politics, social media influencing, etc. Whether or not this approach to life is genuinely or broadly effective, it is now central to American culture. So of course, Americans tend to apply this same culturally informed approach to war.
Military speed-accuracy trade-offs, need for closure, and public analysis When it comes to understanding war, focused minds are far more likely to seek than to avoid closure. Military and civilian leaders need to understand what is happening to make sometimes physically existential – or at least politically important – wartime decisions.
Under tremendous pressure they often succumb to the speed-accuracy tradeoff. 28 Quick decision-making is sometimes
necessary, but it does not leave time for careful evidence gathering or reasoned exploration of alternative evidence or theories. Even the best wartime leaders are often inclined to quickly settle on any seemingly possible answer to a problem even if a more patient, detailed examination might produce far more accurate knowledge.29 As I note, sometimes this trade-off is necessary and useful.30 Sometimes, though, the speed-accuracy trade-off is harmful and impractical.
Pressure induced trade-offs and the need for closure combine to reinforce the existing human and Western – and then also American – inclination to quantify and measure. Sometimes this is a well-meaning and honest process.31 But in the most confounding situations a bit of misdirection and some ill intent can creep in.
If we can’t figure out if we’re winning or losing in the Vietnam War, we apply junk methods like body counting, hamlet evaluations and other measurement systems to help figure things out.32 If
“Pressure induced trade-offs and the need for closure combine to reinforce the existing human and Western inclination to quantify and measure.”
27This term is informally verbed from the title of the 2003 book (and 2011 movie adaptation) Moneyball. Michael Lewis, Moneyball: The Art of Winning an Unfair Game, New York, NY: W.W. Norton & Company, 2003.
28See below for more citations. This hyperlink connects to a neuroscientists explanation of the phenomenon. It offers a narrow but useful explanation and lens into this complex concept. Richard P. Heitz, “The Speed-Accuracy Tradeoff: History, Physiology, Methodology, and Behavior,” Frontiers in Neuroscience, Vol. 8, June 2014, pp. 1-19.
29The late Daniel Kahneman introduced many people to this concept in his popular book Thinking, Fast and Slow. See: Daniel Kahneman, Thinking, Fast and Slow, New York, NY: Farrar, Straus, and Giroux, 2011. Also see Kruglanski and Webster, 1996, and other cited sources on the subject of the need for closure. It is important to note that Kahneman, Kruglanski and Webster, et al., provide narrow scientific insights into human behavioral selection. Many psychologists would disagree with their perspectives and adhere to or offer alternative theories. Moreover, experts across other fields and sub-fields including but not limited to anthropology, social science, neuroscience, philosophy and psychiatry offer a range of other theories and models that reinforce, supplement, contradict or even confound the perspectives I offer here. There is not as yet a single, centralized, generally agreed-upon theory or model of human behavior.
30Speed of decision-making is essential in many military contexts. Historically, great military leaders have been recognized for possessing coup d’oeil, which can be loosely defined as a learned capability to see battlefield challenges, rapidly synthesize them in immediate context, and make equally rapid and effective decisions based on that intuiting process. Or, more briefly, it is the ingrained capability to make rapid and effective decisions. For more on this concept see: Trent J. Lythgoe, “Coup d’oeil and Cognition: How to Build Adaptive Tactical Experts,” Military Review, March-April 2003, pp. 95-107; Ami-Jacques Rapin, Shrewdness, coup d’oeil, and genius: the cognitive attributes of the consummate general (Greek antiquity, Byzantine era, modern times), preprint edition, Lausanne, Switzerland: University of Lausanne, 2023.
31Having worked with and on staffs tasked with campaign assessment and also examined those processes in detail as a researcher, I can attest to the fact that the majority of U.S. and (in Afghanistan) NATO staff members worked hard to provide the best quality, most accurate, most honest and most transparent data and assessments possible.
32See Connable, Embracing the Fog, 2012, for many examples of these other systems including PACES, TFES, TIRS, RDCS, AKS, RS, VHRS, PSDFS, and VCINS. To decipher these acronyms and find original source references see: Connable, Embracing the Fog, 2012, pp. 99-100, 113-115, 126, etc.
we can’t convince political leaders that progress is being made, we deliver sometimes misleadingly precise data that feed a logically unsound sense of impending victory.33 And if we can’t convince our population the war is being won, we cherry pick any of those measurements to make our public case. Misleading precision, cherry picked statistics and needy and speedy guesstimations are the enemies of analytic accuracy.
Public pressure to measure war adds even more speed, misleading precision and another bit of chaos to the mix. As of mid2025, members of the public are far more involved in the day-to-day measurement of war than at any other time in human history. Online data, satellite imagery, videos and instant social media news from the frontlines have allowed for the rapid growth of both a cottage industry of opensource intelligence outlets and a broad community of sometimes-quite-talented individual analysts.
In this “do your own research” era of military assessment, both demand for and supply of war stats have skyrocketed. Full-time analysts, journalists and freelancers competing for our attention try to out-data each other, accelerating the human, Western, American and military predilections to measure. In this frantic and uncontrolled information environment, existing precision-accuracy and speedaccuracy gaps are widening.
This is my subjective take on war knowledge in the 2020s: As we push harder to measure what cannot be reasonably measured with speedy precision but no better proven accuracy than
May Contain Lies, 2024, p. 192.
during the Vietnam War, the gap between actual knowledge and perceived or false knowledge is almost certainly growing.34
Accepting the practicably unbreachable restraints on human knowledge in war Western military theory has it right: War is frequently complex, chaotic, unpredictable, disorderly, hard to observe and even harder to understand. This inherent uncertainty generates an abiding dissonance between the need to know and the practical limitations of knowing. It is perfectly reasonable to continuously try to close the knowledge gap to improve decisionmaking. It is not reasonable to believe, expect or claim to eliminate that gap with
impossibly precise quantification and measurements.
Understanding our behavioral proclivities to avoid uncertainty is an essential first step to building a more effective approach to war analysis. With inherent biases on the table, we can apply scientific logic to help reduce uncertainty to the greatest extent possible and then shape our analyses for better policy-making.
Alex Edmans provides a thorough but digestible explanation of this scientific approach to applied uncertainty in May Contain Lies: How Stories, Statistics, and Studies Exploit Our Biases – And What We Can Do About It 35 He walks the reader through a laddered series of understanding replicated here in Figure 1.36 I recommend reading Edmans’ book. Here is a brief summary of his research-driven explanation:37
n Statements are unsubstantiated and not necessarily meaningful claims. Example: “Aliens exist.” On their own, statements neither represent nor convey knowledge.
n Facts are true but discrete items of information. Example: “Humans exist.” Facts are necessary to build knowledge but may not be independently useful.
n Data is (or are) aggregated information that may or may not be made of facts. Example: A tabulation of total vehicles destroyed in Ukraine. Data are manipulable.
n Evidence is information showing something to be generally true or valid.38
33This technique was used prolifically during the Vietnam, Iraq and Afghanistan wars. See Connable, Embracing the Fog, 2012, Chapters 6-8. Also see Appendix D of Embracing the Fog for examples of contemporaneous official reports. One official report lists the number of bars of soap distributed to the Vietnamese population as a measure of progress (Connable, 2012, p. 99). SIGACT trends were often used as measurements of progress in both Iraq and Afghanistan.
34This problem is compounded by the decreasing emphasis on primary-source field research in war and, arguably, by the increasing possibility that artificial intelligence will muddy online information and data accuracy. On field research see:
35Alex Edmans, May Contain Lies: How Stories, Statistics, and Studies Exploit Our Biases—And What We Can Do About It, Oakland, Calif.: University of California Press, 2024.
36See Edmans, May Contain Lies, 2024, p. 192, e.g.
37These are my aggregated and curtailed definitions not drawn directly from May Contain Lies.
38This is a complex concept that requires further explanation. Again, I recommend reading May Contain Lies for that accessible and useful insight.
Proof Figure 1: Edmans – Ladder of Understanding. Adapted from: Edmans,
Example: We had good evidence that a glass of red wine a day was beneficial.39 But...
n ...proof, which might be defined as a universal truth, does not always emerge from evidence.40 Example: Sadly, it turns out the red wine evidence proved nothing.41
Edmans uses the word facts in this construct, but I prefer to use information 42 In common parlance a fact is perceived to be inherently true and accurate. But in many aspects of life – and certainly of war – discrete items of information should be considered relatively more or less true and accurate rather than inherently factual.
Information has more prospective value than a statement because it ostensibly has some grounding in the real world. But as we all know, information can be easily fabricated or manipulated. In some cases, black and white truth is simply impossible to ascertain in part because the standards for accuracy are not absolute. And in many cases, statements are confused with facts. Consider a notional case:
Statement: A person says, “I am a genius”.
“What we’re giving you is facts and conclusions based on solid intelligence.”
– Colin Powell, 2003
Information: A test shows them to have an intelligence quotient (IQ) of 140, which ostensibly qualifies them as a genius.
Is this information also a fact? No, it is not. IQ tests probably show some relative advantages in knowledge and intellect but generally are not considered to be reliable measurements of practical intelligence.43 And because the word “genius” is subjective, this test result cannot be considered a fact. Nor could it be considered evidence or proof of the person’s inherently unprovable genius.
Now consider a real-world case:
On February 5th, 2003, then-US Secretary of State Colin Powell presented a brief to the United Nations (UN) Security Council on the status of Iraq’s weapons of mass destruction programs. His brief constituted a prelude to war. He stated plainly: “These are not assertions. What we’re giving you are facts and conclusions based on solid intelligence.”44 He cited human intelligence sources to back that claim. But it later turned out that at least one key human source was found to be
39For example: Paul E. Szmitko and Subodh Verma, “Red Wine and Your Heart,” Circulation, Vol. 111, No. 2, pp. 10-11, 2005; J.B. German and R.L. Walzem, “The Health Benefits of Wine,” Annual Review of Nutrition, No. 20, 2000, pp. 561-593; and Ramesh Vidavular, Hajime Otani, Pawan K. Singal, and Nilanjana Maulik, “Significance of wine and reservatrol in cardiovascular disease: French paradox revisited,” Experimental & Clinical Cardiology, Vol. 11, No. 3, Fall 2006, pp. 217-225.
40I write “might be defined” because proof can be understood in both narrow and confined contexts (idiosyncratic proof) and in broader, universal contexts (general truth). In terms of scientific understanding, idiosyncratic proof of causation does not conflate with universal proof of causality. In other words: I can prove that in one specific case, (a) action caused (b) reaction to occur, but I cannot universally prove that the same (a) action always, in every case, causes the (b) reaction to occur. So, in a loose example, I can prove that my isolated action of dropping a metal ball causes the ball to fall. However, I cannot universally prove that dropping a ball always causes a metal ball to fall. Other factors might alter the outcome. The ball might be drawn upwards by a strong magnet that was not present in my other experiment, or perhaps if I drop the ball from an upsidedown position it sits in my hand rather than falls, etc.
41For example: Jakob Manthey, Kevin Shield, and Jurgen Rehm, “Alcohol and Health,” The Lancet, Vol. 400, No. 10365, November 2022, pp. 1764-1765; Benjamin O. Anderson, Nino Berdzuli, Andrew Ilbawi, Devora Kestel, Hans P. Kluge, and Rudiger Krech, “Health and cancer risks associated with low levels of alcohol consumption,” The Lancet: Public Health, Vol. 8, No. 1, January 2023, pp. 6-7; Alice Callahan, “How Red Wine Lost Its Health Halo,” The New York Times, February 17, 2024.
42Fact is more technically accurate in some contexts, and probably within many of the explanations of logical argument construction. I believe the way Edmans intends its use is as a clear, logical differentiator from statement. In this view, any non-factual claim would be relegated to the lowly status of statement. Facts would be few and dependable. That is a decent scientific approach but it falters in the real world. In practice, use of fact in this context does not necessarily denote something that is accurate and true. Popular use of fact does connote both characteristics, neither of which may in truth be present. My objective here is to help the reader think about information in war like an intelligence analyst: Take nothing at face value. Intelligence professionals purposefully use the term “information” rather than evidence or any similar term because it denotes inherent uncertainty. Therefore, I am bending the rules a bit to suit my purpose. I do also tend to believe that the use of “fact” is unhelpfully and unrealistically dichotomous.
43For more on IQ testing, see the Mensa.org website as well as: Anouk van Hoogdalem and Anna MT Bosman, “Intelligence tests and the individual: Unsolvable problems with validity and reliability,” Methodological Innovations, Vol. 17, No. 1, pp. 6-18; Jeff Renaud, “Western University-led research debunks the IQ myth,” press release, Western University, December 19, 2012; A.I. Gates, “The Unreliability of M.A. and I.Q. Based on Group Tests of General Ability,” Journal of Applied Psychology, Vol. 7, No. 1, 1923, pp. 93-100; Stephen Murdoch, IQ: A Smart History of a Failed Idea, Hoboken, N.J.: John Wiley & Sons, Inc., 2007.
44Colin S. Powell, U.S. Secretary of State Colin Powell Addresses the U.N. Security Council, February 5, 2003, slide 18.
Picture: UN Photo/Mark Garten
What is a scientificallyrepresentative random sample?
I refer throughout this monograph to the idea of scientific representativeness and random sampling. Briefly: In order to know if the data you have represent a broader phenomenon within a given group, you have to be able to show that the data represent a group phenomenon within a reasonable margin of error. This is how political polling and other types of representative analysis work. It may seem counterintuitive, but randomness helps generate reasonable accuracy. If something random happens frequently in a sample, it is more likely to be common to the group. Sampling is an imperfect way to understand a large-group phenomenon, but it is necessary when we cannot observe the whole group in detail. Perhaps the least reliable kind of sampling is called convenience sampling: Survey only the most available subjects at the relatively high risk of misrepresenting the whole group. Most military casualty data appear to result from convenience sampling.47
“In war statements dominate, information has relative value, datasets are questionable and reliable evidence that can be used to support decision-making is scarce.”
lying.45 Some of the other visual and technical information he presented had been misinterpreted.46 So, he was presenting a combination of statements and at least partly inaccurate information, not facts.
In cases other than war we can often achieve some level of knowledge below the high threshold for proof. Understanding a complex issue and closing the knowledge gap for improved decision-making requires at least some useful information, hopefully some wellorganized and factual data, and ideally some dependable evidence. Uncertainty is relative. In Powell’s case, these clearly were assertions and conclusions based on necessarily uncertain intelligence. Perhaps given those qualifications the invasion of Iraq might have been reconsidered. In war statements dominate, information has relative value, datasets are questionable and reliable evidence that can be used to support decision-making is scarce. As I recommended above and reiterate in the conclusion, it is better to approach knowledge in war like an intelligence
professional: Treat every bit of information as relatively more or less accurate and never as absolutely factual or false.
Back to the Ukraine statistics
Now return to the wartime claim that 70 percent of all casualties in Ukraine are caused by drones. At this stage of analysis we can surmise that this is simply a statement provided by one person that is (absent a clear source) not backed by observable information or data. It has been misleadingly extrapolated in the press and on social media from just Russian casualties to all casualties. Therefore, this statement constitutes neither evidence nor proof that drones cause 70 percent of all casualties in Ukraine. Even more importantly, it could not reasonably be extrapolated to support a claim that drones might cause any percentage of all casualties in any other war, and particularly in a prospective future war.
Still, it might still be true that drones cause 70 percent of all casualties in the Ukraine War.48 One unsubstantiated statement does not rule out the existence of perhaps
45This source is commonly referred to as “Curveball.” See, for example: Martin Chulov and Helen Pidd, “Defector admits to WMD lies that triggered Iraq war,” The Guardian, February 15, 2011; and Robert E. Kelley, “Twenty years ago in Iraq, ignoring the expert weapons inspectors proved to be a fatal mistake,” Stockholm International Peace Research Institute, March 9, 2023.
46For ex post analysis of Iraq’s WMD programs see: Pat Roberts and John D. Rockefeller IV, Report on the U.S. Intelligence Community’s Prewar Intelligence Assessments on Iraq, report to the U.S. Senate Select Committee on Intelligence, U.S. Senate, undated; Charles Duelfer, Comprehensive Report of the Special Advisor to the DCI on Iraq’s WMD, Langley, Vir.: Central Intelligence Agency, 2004; and Laurence H. Silberman and Charles S. Ross, The Commission on the Intelligence Capabilities of the United States Regarding Weapons of Mass Destruction, report to the President of the United States, Washington, D.C.: Government Printing Office, March 31, 2005.
47See the analysis below for more on how these data are collected and used. In general, casualty data appears to be drawn from hospitalization records, which necessarily show only those casualties treated at a medical facility and not casualties left untreated in more dangerous and inconvenient places like frontline combat positions, or dead bodies buried under rubble, burned beyond recovery or disintegrated. I express my concerns with the reliability of random sampling in war in Embracing the Fog of War and in other publications. For more on the use of data analysis in modern wars, see: Eli Berman, Joseph H. Felter, and Jacob N. Shapiro, Small Wars, Big Data: The Information Revolution in Modern Conflict, Princeton, N.J.: Princeton University Press, 2018; Leo J. Blanken, Hy Rothstein, and Jason J. Lepore, Assessing War: The Challenge of Measuring Success and Failure, Washington, D.C.: Georgetown University Press, 2015; Jonathan Schroden, “Why Operations Assessments Fail – It’s Not Just the Metrics,” Naval War College Review, Vol. 64, No. 4, 2011.
48I return here to the possibility that classified data might exist that provide better insight but also reiterate the observations and arguments I am making throughout this monograph: Even more precise information does not necessarily connote representative accuracy. The same could be true of the alternative 65 percent figure provided by Kostenko in his April interview, or of the 80 percent figure sometimes used by Ukrainian military leaders and analysts in other contexts, or of any other claimed figure. Kostenko, interview with Ihor Dolgov, April 7, 2025. For an example of a notional 80 percent casualty rate see, or listen to: Michael Kofman, podcast with Amos Fox, “Michael Kofman on the Russo-Ukraine War and the State of Military Thought,” Revolution in Military Affairs podcast, August 4, 2025.
hidden but well-substantiated evidence. What, then, would be required to build solid evidence to support this claim or a similar claim? I offer these practical and at least logically reasonable requirements while acknowledging they could only be met in rare circumstances:49
1. Clear definition of a casualty: Do we mean wounded or dead, or both? Would a wounded soldier who continued to fight and was not evacuated still count?50
2. Representative casualty data: Ideally casualty data would be 100 percent complete and accurate, but a scientifically representative sample would be sufficient for analysis.51
3. Reliable evidence of causation: What caused each casualty and how do we know? How would we address multiple causes (e.g., a bullet and a drone wound)?
4. Time and space bounding: Which time period are we analyzing and why? How do we geographically and spatially bound our dataset, and why?
What I ask for here is, effectively, proof in at least the Ukraine War context. I set this exceptionally high bar because the consequences of lazy analysis and
Source: U.S. Department of Defense, Defense POW/MIA Accounting Agency, 2025
U.S. MISSING IN ACTION FROM ALL WARS AS OF MID-2025
WORLD WAR II: 71,862
KOREAN WAR: 7,401
VIETNAM WAR: 1,566
COLD WAR: 126
GULF WARS AND LIBYA: 7
Total U.S. servicemembers still missing in action in 2025: 80,963
inaccuracy in this case are significant. If we accept this 70 percent figure, or any similar figure without well-substantiated evidence, then we run the risk of extrapolating false knowledge from the Ukraine War context to a universal context.52 And from there we may change our own military forces in ways that unintentionally make them less ready for future fights.
In the following section I provide a survey of historical Western casualty analysis to help highlight all these measurement challenges. Understanding the vagaries inherent in, for example, even the most carefully recorded World War II statistics will help put the Ukraine War analyses in context.
Historical casualty statistics
Own-side casualties are generally far easier
49For example, see the cited works below on U.S. casualties in the 1991 Persian Gulf War.
to quantify and measure than most other factors in war. Well-organized military forces like those of Great Britain and the U.S. tend to keep good records on individual soldiers and casualties. Particularly starting in the late-20th century they have had a reasonably good handle on how many people are serving, who they are and whether they are alive and well, wounded or dead.
However, even the best record keeping in history cannot overcome the inherent uncertainty of war. Soldiers frequently go missing and are never found, often because they are dead and unrecoverable. Many wounds and deaths are not directly observed and therefore are not carefully recorded. Units that have been routed often leave both wounded and dead
50I also assume here that “casualty” would not include destroyed vehicles. In general, foundational definitions are essential to good research and evidence building. Everyone involved must agree upon a central understanding of the problem and define terms before facts can be established, data can be amassed and analyzed, and evidence can be built. What if, in the case of casualties, one person defines a casualty as only a combat wounded person while tracking deaths separately; another person uses the term casualty broadly to include all combat wounded and killed; another person uses the term only for combat casualties and not disease or mishap losses that often constitute a fairly high percentage of overall casualties in war; while another person uses casualties to include all losses including from disease and mishaps? With that kind of disparity in place at the outset of our analysis we would have no way of building a single set of reliable evidence that had both internal and external validity. As we will see in the following section, problems with definitions have had a real-world impact on our understanding of historical casualty figures in war.
51As I will show in the next section, a scientifically random sample would be extraordinarily difficult to acquire in the Ukraine War or in any war. In order to know a sample is representative it would be necessary to know the bounded limits of the data; to have some scaled confidence in the quality of the data both sampled and not sampled; and to be able to time-bound and location-bound the data to some extent. So in this case, generally, and with a clear definition of casualty in place: How many total casualties have been inflicted on both sides (or perhaps just the Russian side)? How many of those casualties can be confirmed as combat casualties? What are the categories of wounding agent (i.e., bullet, fragment, burn, etc.)? How have the data aggregators differentiated between the types of wounding agents, and are those differentiations reliable? At least in my review of available public data, none of these questions can be answered with confidence.
52For example, the U.S. Department of Defense recently released a memorandum describing drones as “the biggest battlefield innovation in a generation”. This statement suggests a limited understanding of the evolutionary history of drone development and perhaps an exaggerated perception of drone utility in modern warfare. While this memorandum does not directly cite the Ukraine War as the driver behind the inclusive directive, the Ukraine War is widely cited by defense officials as one of, if not the single most important, narratives shaping their perceptions of modern war. See: Peter B. Hegseth, Unleashing U.S. Military Drone Dominance, memorandum, Washington, D.C., Office of the Secretary of Defense, undated. For more on the evolution of drones and the impact of drone use in the Ukraine War on modern policy see: Ben Connable, Ground Combat: Puncturing the Myths of Modern War, Washington, D.C.: Georgetown University Press, 2025, chapters 1 and 9.
53As of August 15, 2025: https://www.dpaa.mil
behind; those soldiers, too, may never be accounted for. Sometimes, unit records are poorly recorded or are even destroyed in or out of combat.
And while general casualty data (e.g., overall dead and wounded) in wellorganized militaries are probably fairly reliable in most wars, data on the causes of wounding and death are probably far less reliable. In fact, even the data on own-side causes of wounding and death in some of the best recorded wars in history appear to be generally unreliable. Applying my criteria above – clearly defined, representative, reliable and timeand-space-bound – to historical casualty data reveals many gaps, flaws and inconsistencies.
I focus here on ground combat casualties since these data are far more plentiful and easier to compare than air and sea battle casualties. For reasons that will become apparent, I bookend this examination of casualty statistics with the American Civil War and the Ukraine War.
Cause-of-casualty data in the American Civil War
As wars go, the American Civil War (18611865) was a comparatively straightforward affair. Armies often maneuvered en masse and frequently engaged each other in open spaces under direct observation of commanders and even some curious civilian onlookers.54 However, neither the Union federal nor Confederate armies had truly reliable bookkeeping practices in place at any time during the course of the war. Confederate books were comparatively atrocious and many military records were destroyed or dispersed during and after the war.55 Table 1 (below) shows aggregated common estimates of total soldiers serving and killed on both sides in the Civil War.56
Note that the number of serving Confederate soldiers throughout the war is no better than a wild estimate that might be off by 100 percent. Total Confederate wounded is effectively unknown.58 Possibly over 100,000 Confederate soldiers deserted, adding to the confusion.59 All other statistics for both sides are just estimates based on census data, snapshots of unit rolls and uneven enlistment records. These data are so unreliable that one scholar showed evidence that overall combined deaths were off by at least 130,000 or by as much as 230,000.60 On its website in 2025
54Of course this was not always the case. Many battles and skirmishes were fought in wooded areas, between dispersed forces well out of sight of any senior commander, and perhaps also on the move (e.g., cavalry raiding operations). In many cases wounded and dead soldiers were left on the battlefield as one side or the other was routed. Nonlinearity also was common in World War I. Operations in Europe from 1914-1918 were far more complex than most people imagine. For good examples of these commonly nonlinear operations see: Infantry in Battle, second edition, Richmond, Vir.: Garrett & Massie, 1939.
55Some records were ordered destroyed as the Union advanced and many records were purposefully and in some cases unintentionally destroyed when Richmond, Virginia, was burned. See: “Richmond in Flames and Rubble,” American Battlefield Trust, October 9, 2023; Trevor K. Plante, “Researching Confederate Marines in the Civil War,” Prologue Magazine, Vol. 33, No. 4, Winter 2001, no pages given; DeAnne Blanton, “Confederate Medical Personnel,” Prologue Magazine, Vol. 26, No.1, Spring 1994, no pages given. Articles from Prologue Magazine and other helpful resources are available on the U.S. National Archives website.
56These numbers are drawn from the U.S. Defense Casualty Analysis System, Principal Wars In Which the United States Participated – U.S. Military Personnel Serving and Casualties (1775-1991), undated, DCAS 3.02.01 as of August 14, 2025. Confederate data are listed as estimates produced by the Provost Marshal General. I rounded these figures off and used the higher-end figure for Confederate deaths in Union prisons. For more detailed and contemporaneous data on the Confederate forces, see: Thomas L. Livermore, Numbers and Losses in the Civil War in America, 1861-65, Cambridge: Houghton, Mifflin and Company, 1900.
57Note that this figure is far lower than the 2,898,304 enlistments reportedly recorded in the U.S. War Department in the late 1800s. See Livermore, Numbers and Losses, 1900, p. 63.
58See Livermore, Numbers and Losses, 1900, for more detail on these estimates and for the original sources of field data.
59Livermore, Numbers and Losses, 1900, p. 5. The original source for this estimate comes from Confederate muster rolls retained at the time by the U.S. War Department.
60These revised data are disputed. See: J. David Hacker, “A Census-Based Count of the Civil War Dead,” Civil War History, Vol. 57, No. 4, December 2011, pp. 307348; M. Flotow, “J. David Hacker’s ‘A Census-Based Count of the Civil War Dead,’” The Journal of the Abraham Lincoln Association, Vol. 42, No. 2, 2022; Joan Marcelo, Jeffrey L. Jensen, Leonid Peisakhin, and Haoyu Zhai, “New estimates of US Civil War mortality from full-census records,” PNAS, Vol. 121, No. 48, 2024, pp. 1-3. I found Flotow, 2022, to be one of the more informative articles on the subject. Also see: Nese F. DeBruyne and Anne Leland, American War and Military Operations Casualties: Lists and Statistics, Washington, D.C.: Congressional Research Service, January 2, 2025.
Table 1: Aggregated U.S. Civil War Serving and Casualties, U.S. Government Statistics
the American Battlefield Trust highlights key challenges:61
Compiling casualty figures for Civil War soldiers is a complex process... even 150 years later no one has, and perhaps no one will, assemble a specific, accurate set of numbers, especially on the Confederate side... combat threw armies into administrative chaos and the accounting done in the hours or days immediately following a battle often raises as many questions as it answers. For example: Who are the missing? Weren’t many of these soldiers killed and not found? What, exactly, qualifies a wound and did armies account for this the same way? What became of wounded soldiers? Did they rejoin their unit; did they return home; did they die? A wholly accurate count will almost certainly never be made.
With these underlying unknowns in place we can focus on the cause-of-casualty statistics. What types of weapons caused battle casualties in relative proportion to each other? For the Civil War these data are rendered with precision. For example, the U.S. Army’s medical research center listed the sources of Union Civil War casualties in a table, replicated above.62 At least rows three and four of this remarkably precise
Weapon or Missile
Rifle or smooth-bore musket
Fragments from shells
Cannonall or grapeshot
Cutting
table show results down to the individual soldier. However, the table only lists 124,000 casualties and not the estimated 360,000 total Union combat dead. Is it perhaps scientifically representative? Maybe, but it is not immediately clear that this sample is randomly reflective of all Union combat deaths.63
When we pull the thread on the sourcing for this table we find it was extracted from Trevor N. Dupuy’s The Evolution of Weapons and Warfare. 64 In turn, Dupuy’s
table is not directly cited. His numbers may have been drawn from a 1952 book, Battle Casualties: Incidence, Mortality, and Logistic Consideration, or perhaps from Albert G. Love’s War Casualties 65 However, neither of these two sources appear to contain this Civil War data.66 Therefore, no conclusions about the Civil War or about warfare in general should be drawn from the numbers in Table 2, left: They are not verifiable facts, they do not constitute valid evidence and they are far from proof of any generalizable trend.
Cause-of-death casualty data from World War I
Citation checks reveal a continuing trend from the Civil War, through the world wars, and into the latter half of the 20th century. In all cases, professional journals published and-or referred primarily to knowingly incomplete hospitalization records to identify causes of injury and death. In other words, in many cases these data represent only recorded information that may or may not have been factual. And in no case could any of the cited information or data be considered representative evidence
Warfare evolved from the 1860s into the 1910s, becoming more industrialized and in comparative scale more lethal. Massed artillery fires probably caused far more
61American Battlefield Trust, “Civil War Casualties, The Cost of War: Killed, Wounded, Captured, and Missing,” updated as of September 15, 2023. Anyone interested in even more dated casualty estimates would enjoy the detailed work by Catherine Rubincam in: Catherine Rubincam, “Casualty Figures in the Battle Descriptions of Thucydides,” Transactions of the American Philological Association (1974-2014), 1991, Vol. 121, pp. 181-198. Also see: Gaston Bodart, Losses of Life in Modern Wars: Austria-Hungary; France, Oxford, UK: Humphrey Milford, 1916.
62Ronald F. Bellamy and Russ Zajtchuk, “Assessing the Effectiveness of Conventional Weapons,” book chapter in: Ronald F. Bellamy and Russ Zajtchuk, eds., Conventional Warfare: Ballistic, Blast, and Burn Injuries, Washington, D.C.: Center of Excellence in Military Medical Research and Education, 1989, p. 56.
63It would also be reasonable to at least gently question the ability to clearly differentiate wounds from a shoulder-fired weapon from those of grapeshot and shrapnel. Shoulder weapons fired much smaller projectiles, typically just over half an inch to about 7/10ths of an inch in caliber compared to two inches to four inches for a grapeshot ball, while shrapnel projectiles could have been any size probably up to a few inches in jagged shape. However, wounds from all three of these causes were often horrendous, leaving gaping holes and rips, and bodies could be hit by all three types of projectile standing and prone.
64Trevor N. Dupuy, The Evolution of Weapons and Warfare, Fairfax, Vir.: Hero Books, 1984, p. 171. Note that this page number is taken from the Da Capo Press reprint of Dupuy’s original manuscript.
65Dupuy lists two books, Numbers and Losses and Battle Casualties in the same section but attaches no citation directly to the table. A review of Livermore, Numbers and Losses, 1900, shows no correlating data. No such table or data seem to appear in Battle Casualties or in Love’s War Casualties, though I may have missed the reference in my review. All three of these books present statistical data from official medical records, reinforcing the idea that detailed information on casualties – including on cause of wound – is derived primarily from medical records that are sometimes compared with unit rolls and reports. In turn, this suggests that only those soldiers directly examined by medical professionals were counted for cause-of-wound estimates. Gilbert W. Beebe and Michael E. DeBakey, Battle casualties: Incidence, mortality, and logistic considerations, Springfield, Ill.: Thomas, 1952; Albert G. Love, War Casualties, Army Medical Bulletin Number 24, Carlisle Barracks, Penn.: U.S. Army Medical Field Service School, 1931. See below for more on the source of Love’s data.
66Dupuy’s The Dupuy Institute references Love elsewhere. See, for example: Shawn Woodford, “‘Love’s Tables’: U.S. War Department Casualty Estimation in World War II,” The Dupuy Institute, March 19, 2018.
Table 2: Purported Sources of Union Casualties in the Civil War
Source: Bellamy and Zajtchuk, 1989, p. 56
casualties in World War I than artillery did during the Civil War.67 Artillery often shattered, fully disintegrated or completely buried bodies in rubble, making even rudimentary casualty counting difficult.68 And as in all wars, some armies tracked their casualties more carefully than others.69
German and French casualty tracking was haphazard, particularly near the end of the war as both bureaucracies started to shake apart under the strain of prolonged conflict.70 German casualty estimates may have been no better than those of the Confederates in the U.S. Civil War. Depending on the source, Germany fielded anywhere from four million to 13 million troops during the war and suffered between about 760,000 and 1.8 million war deaths.71 Some estimates derive from a mix of census data and enlistment records while others rely on hospital records.
1914-1917
*Shells, grenades and mortar bombs
Table 3: Bellamy and Zajtchuk World War I Cause-of-Death Data
It looks like the ultimate source of the most detailed German casualty figures is the Sanitätsbericht über das Deutsche Heer (Deutsches Feld-und-Besatsungheer) im Weltkriege 1914/1918, which aggregates the official statistics of the German medical service.72 Two analysts, and James Quirk, examined what they refer to as the Sanitats and wrote:73 “The source of the Sanitats data
“German and French casualty tracking was haphazard, particularly near the end of the war as both bureaucracies started to shake apart under the strain of prolonged conflict.”
In other words, even the most detailed evidence available has an unclear provenance; we do not really know how information was collected and how the dataset was built. Yet these numbers were put to use and repeatedly cited military and scientific publications, sometimes without caveat. Causeof-death analysis for German troops appears to have been limited to some unknown percentage of only those bodies examined by military doctors the field.74 For example, see Table 3 above. Missile Bullet Fragments*
is not known with certainty, but strong circumstantial evidence indicates that these data come from the records of the medical units in the field and hospitals...”
67I write “probably” here out of caution and to drive home my point that the data supporting even broad statistical analyses of World War I casualty causation are generally inadequate except in some specific battle cases. Masses of anecdotal evidence suggests artillery caused most casualties, but massed anecdotes alone do not constitute scientific evidence.
68For an anecdotal account of the impact of artillery see, for example: Lord Moran (Charles McMoran Wilson), Anatomy of Courage, reprint, New York, N.Y.: Carroll & Graf, 2007 (1945).
69I focus on the German and American armies here. For another perspective see: David Noonan, Those We Forget: Recounting Australian Casualties of the First World War, Victoria, Australia: Melbourne University Publishing, 2014; Pablo Duarte, Marcel Freidinger, and Andreas Hoffmann, “Military Casualties and Exchange Rates During the First World War: Did the Eastern Front Matter?,” The Economic History Review, Vol. 72, No. 4, November 2019, pp. 1312-1334; Victor Gay and Pauline Grosjean, “Morts Pour la France: A Database of French Fatalities of the Great War,” Explorations in Economic History, Vol. 90, October 2023.
70On this point for the German army see: James H. McRandle and James Quirk, “The Blood Test Revisited: A New Look at German Casualty Counts in World War I,” The Journal of Military History, Vol. 70, No. 3, July 2006, pp. 667-701; and Robert Weldon Whalen, “War Losses (Germany),” in: International Encyclopedia of the First World War, Version 1.0, October 2014.
71For these various estimates see: Whalen, “War Losses,” 2014; McRandle and Quirk, “The Blood Test Revisited,” 2006, p. 697; and Gerhard Hirschfeld, Gerd Krumeich, and Irina Renz, eds., Enzyklopädie Erster Weltkrieg: Aktualisierte und erweiterte Studienausgabe, Paderborn, Germany: Ferdinand Schöningh, 2014, pp. 664665; and Leonard P. Ayres, The War With Germany: A Statistical Summary, Washington, D.C.: Government Printing Office, 1919, p. 119 and others. McRandle and Quirk refer to the German Sanitats data held on microfilm at the University of Tennessee Library; see below.
72Sanitätsbericht über das Deutsche Heer (Deutsches Feld-und-Besatsungheer) im Weltkriege 1914/1918, archived data publication, Berlin: Reichskriegsministerium, 1934, accessed on microfilm through the University of Tennessee Library as cited in McRandle and Quirk, “The Blood Test Revisited,” 2006. McRandle and Quirk also referred to contemporaneous German data from the “Zentral Nachweiseamt (Central Enquiry Office for War Casualties and War Graves),” which they used only to try to align changes in overall casualty data. McRandle and Quirk, p. 674. Graves data is frequently used to help check and round out overall casualty data. However, I see no evidence that grave registration data contributes in any meaningful way to our broader understanding of combat causes of death.
73McRandle and Quirk, “The Blood Test Revisited,” 2006, p. 674.
74“Sanitätsbericht über das Deutsche Heer (Deutsches Feld- und Besatzungsheer) im Weltkriege 1914/1918: III. Band, Die Krankenbewegung bei dem Deutschen Feld- und Besatzungsheer, Berlin 1934, German Ministry of War, archive, created in 1934, Federal Archives, BArch PH 2/1998.
The reference cited on the original of this U.S. Army table appears to be the same Sanitats cited by McRandle and Quirk. And here the authors are cautious about the original source data.75 Cause-ofcasualty percentages are drawn from an examination of only 14,486 total casualties out of possibly 5.5 million total German casualties in the war.76 And even these few casualties represent only an unknown percentage of combat wounded and dead brought to a medical facility and cannot under any circumstance be considered scientifically random and representative of all German battle deaths, or of war deaths more broadly.
So, no matter how reliable, representative and compelling the British, American and French data might be, the title of this table is misleading: It does not show what it purports to show. It should read something like “Non-Representative Sample: Some Sources of Casualties from Various Armies in World War I”. It does not constitute evidence.
And American and British statistics are
Table 4: Uncited Casualty Data Example in Ayres, 1919 Source: Ayres, The War With Germany, 1919, p. 123
not necessarily any more accurate or representative than the German stats.77 Studies on American casualties generally refer to three sources, the first of which is Leonard Ayres’ The War With Germany: A Statistical Summary 78 Ayres was meticulous and precise but did not directly cite his original sources.79 Therefore, his data are loosely informative but have
no defensible accuracy; they are not practicably useful.80 See Table 4, above, for a sample diagram based on Ayres’ original.81
Next we have Battle Casualties, which goes into great detail on the challenges and limitations of casualty data in war, and refers back to Albert Love’s War Casualties
75They state, “Published data on the sources of casualties by weapon frequently fail to state whether the data pertain to the total population... or only to those casualties who were hospitalized. Uncertainty... also arises from the data collectors’ inability (a) to count all the dead; and (b) to accurately identify the weapons that were the cause of death. Assessments of the causes of death are sometimes based upon surprisingly small samples.” Bellamy and Zajtchuk, “Assessing the Effectiveness of Conventional Weapons,” 1989, p. 57.
76Bellamy and Zajtchuk, “Assessing the Effectiveness of Conventional Weapons,” 1989, p. 57.
77Some detailed sources on both American and British casualties are available in archives, some of which are cited here. British archives are extensive and include a register with the Commonwealth War Graves Commission, records in the National Archives, and an 80-volume list of the dead published by His Majesty’s Stationary Office in 1921. For U.S. casualties: Some official histories of the war suggest that at least the initial casualty reports were estimated due to battlefield chaos and the inability to recover dead and wounded in an expeditious manner. For example, see references to the rather generalized casualty reporting from the front in: U.S. Army Center of Military History, United States Army in the World War, 1917-1919: Reports of the Commander-in-Chief, Staff Sections and Services, Washington, D.C., 1991, pp. 28 and 39. On page 39 the report lists an initial estimate of approximately 11,000 casualties for a specific battle but notes that a final report issued later “places the casualties at less than 7,000”. This wide disparity suggests many unknowns, many of which may or may not have been ultimately resolved.
78Ayres, The War With Germany, 1919. For an example of a report citing Ayres see: Evan Roberts and Alexandra Burda, Correlates and Consequences of American War Casualties in World War I, working paper, University of Minnesota, July 2018.
79See, for example: Ayres, The War With Germany, 1919, pp. 121-123. Ayres presents data in tables and charts without citation. He refers obliquely to the Central Records Office of the American Expeditionary Forces but does not cite this office directly. These records are located at the U.S. National Archives as Records of the American Expeditionary Forces (World War I), Record Group 120. There are 26,016 cubic feet of records here, including many references to casualties. It appears casualties were recorded by the adjutants and surgeons, suggesting they were accounted for by enlistment-officer record, unit roll, unit reporting and hospital reports. There are no explicit indications of wound causation records but they may be included.
80Other sources cite the standing official data provided by the U.S. Department of Defense which, in turn, appears to have been drawn from a report by the U.S. War Department in 1924 and an amendment to that report published by the Statistical Services Center, Office of the Secretary of Defense (OSD), in 1957. It is not clear but highly likely that these data were drawn in turn from Battle Casualties and-or War Casualties, or from their cited primary source reports. See, for example: John T. Correll, “Casualties,” Air Force Magazine, June 2023, pp. 48-53, p. 51. Correll refers to the Encyclopedia Britannica, which in turn refers to the loosely cited War Department and OSD reports. As of August 18, 2025.
81While these data must be considered non-representative, they do highlight another important caveat in wartime casualty analysis. Disease and other non-combat causes have historically removed a significant number of soldiers from the battlefield. Relative losses to disease appear to have fallen significantly over time as medicine has advanced worldwide, but some percentage of soldiers is always incapacitated by non-combat factors.
for its World War I data.82 Love’s book refers to the original hospital-record details and comparative statistical analyses he compiled for the exhaustive official medical records of the Medical Department of the U.S. Army.83 These show much more specific causes of wounding and death. See Table 5, right, which shows reported causes of wounds.84 Even these exceptionally detailed data from World War I, all of which were recorded on an index card system designed in 1917, still only show medical professionals’ assessments of casualties treated and not of all casualties lost on the battlefield. Therefore, they might be both precise and accurate for all Army medical patients but still have no representative value for U.S. casualties in World War I as a whole.85
Possibly the best insight into American cause-of-wound and cause-of-death data from World War I is offered by Gilbert W. Beebe and Michael E. De Bakey, the authors of Battle Casualties:86 “Of equal importance [to the medical method of wound examination] is the questionable validity of the World War I figures on fatality by type of missile. This judgment stems not only from the essential nature of the characteristic observations which lie behind all such figures, but also from the fact that half of the wounded and 60 percent of the deaths in World War I are unknown as to type of missile, so that the unknown category has the highest case-fatality rate.”
Bottom line on World War I cause-of-death
82Beebe and Bakey, Battle Casualties, 1952.
records: As with the American Civil War, no reliable record showing causes of death appears to exist. Therefore, we can draw no scientifically reliable conclusions – or even generally useful estimates – about the specific and relative causes of death in war thus far.87
Cause-of-casualty data from World War II War information density increased significantly through World War II and then into the second half of the 20th century. More data generated greater precision and a higher volume of analytic reporting. In turn, increased precision and higher report volume fed a greater collective sense of
Table 5: Official U.S. Army World War I Causeof-Wound Statistics
Source: Love, Medical and Casualty Statistics, 1925, p. 1020.
accuracy compared to data from previous wars. Shouldn’t the sheer weight of evidence be adequate to prove causation? No, it would not, because without absolute value – i.e., we know everything down the finest detail – or randomized representative data it remains uncompelling. In fact this volume and precision of information led to a considerable precision-without-accuracy, precision-without-representative-value problem. Thankfully we have ample expert insights on data from World War II describing these data problems. At least 50 countries directly contributed to the fighting in World War II, probably deploying more than 100 million soldiers collectively.88 Six powers probably contributed the
83Albert G. Love, The Medical Department of the United States Army in the World War, Statistics, Part Two: Medical and Casualty Statistics, Washington, D.C.: Government Printing Office, 2025. This is both an analysis and data book providing individual soldier-level data from the index card system. Reader note: This original reporting goes to great pains to separate casualties by skin color, using outdated terms that some might find offensive.
84This chart is derived from: Love, Medical and Casualty Statistics, 1925, p. 1020. The original shows both purported cause of wound and percentage of those wounded who later died by cause (e.g., projectile, gas).
85It is not clear if these data also include U.S. Marine casualties. The Marine Corps brigade served under the U.S. Army Second Division, so it would be logical that all casualties were included. Official Marine Corps accounting shows 1,459 Marines were killed in action, 991 died of wounds after treatment, 27 died of accident, 269 died of disease, and 12 died from other causes. These numbers are from: Edwin N. McClellan, The United States Marine Corps in the World War, Washington, D.C.: Headquarters, U.S. Marine Corps, 1920, p. 64.
86Under a section entitled, “Sources of Casualty Data,” Love states that his information was derived from Beebe and Bakey, Battle Casualties, 1952, p. 128. They cite Love, War Casualties, on this point and reference his data on gas and unknown casualties.
87I must leave open here the possibility that other wars through World War I have revealed greater and more reliable detail. However, a review of the cited sources also includes other wars, none of which appear prima facie to provide that kind of useful accuracy. Perhaps some small engagements did provide more useful data, but those would not be usefully generalizable. See more on smaller wars data below.
88This is a loose estimate based on a wide array of conflicting sources. Definitions matter a great deal here: What does “fighting” mean?
majority of the fighting forces in the war: Germany, Italy, Japan, the United States, the United Kingdom Commonwealth and the Soviet Union.89 So, if someone were to make the claim that ‘combat deaths for the major powers fighting in World War II were caused by (e.g., bullets, artillery shells, tank shells, bayonets, etc.) in relative proportion,’ they would at least have to provide a reliable random sample study that would include causes of wounding and death for the entire period of fighting and for each of the six powers:90
n Total soldiers serving over time;
n Total ground combat casualties;
n Clear and reliable sources for casualty information;
n Combat wound examination or reporting data for each combat death applied;
“Probably the vast majority of [Soviet Union] soldiers were left where they died or were buried hastily in field cemeteries and could not have been carefully examined by medical staff.”
n [at least] Reliable random sample of combat deaths with reliable causation.
Before we even dig into the data it looks clear that this may be an impossible task. No two powers – or states within powers, or military forces within states, etc. – consistently recorded even their serving soldiers over time, let alone their casualties. American data probably are the most relatively useful, but they are offset by larger disparities. For example, the Soviet Union may have committed as many as 35 million servicepeople
to the war. It lost perhaps as few as eight million and as many as 14.6 million soldiers dead.91 These are rough estimates based on multiple, conflicting and in many cases biased sources. Probably the vast majority of these soldiers were left where they died or were buried hastily in field cemeteries and could not have been carefully examined by medical staff.
Неизвестного Солдата [Tomb of the Unknown Soldier] in Moscow’s Alexander Garden is dedicated to the Soviet soldiers killed during World War II.
Historian Christopher Lawrence estimates that about 20 percent of Red Army soldiers lost in each battle were missing in action, so they were never physically examined.92 In general there is no reasonable possibility of accurately counting Soviet combat casualties, let alone causes of death, or the cause of the wounds that led to death. Russian anthropologist Boris Sokolov had this to say about counting even those who served in uniform:93 “...there is no way whatsoever to calculate precisely how many persons actually served in the Red Army during the war inasmuch as a considerable number of these soldiers were mobilized (conscripted or even dragooned) directly from local regions into
89Some might add France, Poland and China here as well, and with defensible reasons. I am providing a gross overview in this section and not attempting to count all contributions and deaths.
90In order to claim generalizability across all participating nations in World War II, or to claim, “combat deaths in World War II were caused by these weapons in relative proportion,” every nation would have to be represented with equivalent consistency and reliability in the dataset.
91These numbers do not include Soviet citizens but do include members of the Soviet sea and air services, so they are not strictly limited to ground combat forces. The Price of Victory (below) does break out losses by service. See, for example: Boris Sokolov, “How to Calculate Human Losses During the Second World War,” The Journal of Slavic Military Studies, Vol. 22, 2009, pp. 437-458; Alfred Vagts, “Battle and Other Combatant Casualties in the Second World War,” parts I and II, part 1 in The Journal of Politics, Vol. 7, No. 3, August, 1945, pp. 256-294; Edwin Bacon, “Soviet Military Losses in World War II,” The Journal of Slavic Military Studies, Vol. 6, No. 4, 1993, pp. 616-633; and Lev Lopukhovsky and Boris Kavalerchik, The Price of Victory: The Red Army’s Casualties in the Great Patriotic War, South Yorkshire, UK: Pen & Sword Books Ltd., 2017. Note that Vagts relies heavily on non-academic articles published in The New York Times and probably has the least credible estimates of these four. See the cited works in Sokolov, in particular, for the best guide to original source material on this subject. Lopukhovsky and Kavalerchik, and to some extent the other authors cited all point out the politicization of the official statistics and manipulation of data over time.
92See the cited explanation and table for methodology. Christopher Lawrence, War By Numbers: Understanding Conventional Combat, Lincoln, Neb.: Potomac Books, 2017, p. 191. Lawrence refers in part to Trever N. Dupuy’s work on attrition. Dupuy is careful to list a range of caveats for cause-of-death statistics and wrote plainly, “The raw statistical data [for all wars] is not entirely reliable.” (p. 82). Dupuy does combine a range of statistics in pursuit of a more accurate dataset, but it is not clear that these combinatory efforts outweigh the challenges with the underlying data for cause of death. Trevor N. Dupuy, John R. Brinkerhoff, C. Curtiss Johnson, and Peter J. Clark, Handbook on Ground Forces Attrition in Modern Warfare, Historical Evaluation & Research Organization, Fairfax, Vir., September 1986.
93Sokolov, “How to Calculate,” 2009, pp. 453-454.
advancing Red Army formations and units from 1941 through 1944. Furthermore, the Red Army maintained no centralized records of these levies.”
Data were no better for the German, Japanese or Italian forces.94 Readers should try to imagine the collapse of Axis forces just at one place and time: Stalingrad, 1942-43. Casualty numbers for the German, Romanian, Italian and other forces may have exceeded half a million for this one battle.95 As the Axis forces collapsed through early 1943, perhaps hundreds of thousands of corpses lay where they fell, to be buried in snow, rubble and eventually the earth, only then to be continually unearthed as partial skeletons in the decades that followed. Yet still, impossibly precise cause-of-death ratios for Axis soldiers on
the Eastern Front have been published and are still referenced.96
How accurate and representative are the American and Commonwealth data? To a great extent they are better than those of the other powers. Canadian, British and several other Commonwealth nations kept fairly meticulous enlistment and reporting records, as did the Americans.97 So while we cannot even venture a reasonable guess at the total number of Soviet troops who served in World War II, we do have probably reliable data on total rolls for Allied powers. But “better” in this case remains insufficient for evidence or proof.
Allied casualty data are less reliable than service enlistment data, and cause-ofcasualty data are even less reliable. As in World War I, casualties in World War
II generally were tracked by triangulating hospital records, enlistment records and unit reporting. Probably the most frequently referenced table is this one from the official Medical Statistics in World War II, Table 6 (right).98
In blurry letters at the bottom of Table 6 are these words: “Excludes cases where the specific causative agent was not recorded or was unknown.” According to the same source, “about 52 percent of the medical reports of death for the killed in action either were listed as unknown or were not recorded for causative agent”.99 Therefore, medical officials claimed to have recorded a causative agent for just half of known total deaths, with the actual total number of deaths being unknown.100 For those cases recorded, we cannot rely on the accuracy of the diagnoses and must assume wide
94Some Japanese casualty figures are particularly poor in part because of Imperial Japan’s policy of defensive collective sacrifice during the Allied island hopping campaign from 1942 through 1945. On islands across the Pacific Theater including Guadalcanal, Tarawa, Peleliu, Iwo Jima, Okinawa, the Philippines, and Saipan, and in China –particularly in Manchuria in 1945 – hundreds of thousands of Japanese soldiers died and were unaccounted for by their own armed forces. No reasonable claims of reliable cause-of-death analysis could be made regarding Japanese land force casualties in World War II. To get a broad sense of the scale of unrecorded loss see: Mari Yamaguchi, “75 years later, 1 million Japanese war dead still missing,” Associated Press, August 12, 2020; Tomoko Otake, “The long search for Japan’s lost soldiers,” The Japan Times, August 15, 2025; Eri Ohtani, Haruyki Makishima, and Kazuhiro Sakaue, “The recovery and repatriation of the remains of Japanese war dead and the roles of physical anthropologists,” Forensic Science International, Vol. 324, July 2021; and The HALO Trust. The hyperlinked article on Italian casualties is: Dmitri A. Jdanov, Dana A. Glei, and Domantas Jasilionis, “A modeling approach for estimating total mortality for Italy during the First and Second World Wars,” Genus, Vol. 66, No. 1, 2010, pp. 17-36.
95Casualty figures for Stalingrad are somewhat disparate and hard to interpret. Eminent Russia analyst David Glantz presents a low end of 217,000 Axis casualties total and then what he calls a “reasonable” high end of 580,000 killed, wounded and missing, while Robert Forczyk suggests 235,000 dead in the German 6th Army alone. Footnotes and citation on the Stalingrad casualty figures are hard to come by and not necessarily useful when they are examined. Glantz is known for his archival work in Russia, but he does not directly cite the many quantitative claims in his book. He does state that as of January 10th, 1943, the German 6th Army stopped counting its casualties. Forczyk offers no citation. Many other sources and examples exist but (arguably) non provide more useful clarity than Glantz. See: David M. Glantz and Jonathan M. House, Endgame at Stalingrad, Book Two: December 1942-February 1943, the Stalingrad Trilogy, Volume 3, Lawrence, Kan.: University Press of Kansas, 2014, pp. 600-601; Robert Forczyk, Stalingrad 1942-43 (3): Catastrophe: the Death of 6th Army, Oxford, UK: Osprey, 2022, p. 81. Also see: Frank Ellis, The Stalingrad Cauldron: Inside the Encirclement and Destruction of the 6th Army, Lawrence, Kan.: University Press of Kansas, 2013.
96Given the available data, any claim on this point would be highly suspicious. Yet Bellamy and Zajtchuk published a table purporting to show German cause of wound and cause of death data from the entire Eastern Front, which included the Battle of Stalingrad. They report the source as the German Central Archives for Military Medicine in Berlin. Cited material included clinical records and field reports. However, they also write: “Little is known about the methods used in the field to obtain the original data.” Spot checks for wound cause may have been carried out, but the number of checks is no longer available. In other words, the German medical services could have carried out one or perhaps a million spot checks. There would be no way to know whether these checks were done in a way that would generate scientific randomness, but it seems almost impossible that could be the case given varying and difficult battlefield conditions on the Eastern Front. Even though the authors provide caveats that effectively nullify the prospective representative value of the data, they still republish the statistics as if they were proven. Bellamy and Zajtchuk, “Assessing the Effectiveness,” 1989, p. 58. An example of a newer reference to this work: John B. Holcomb, Lynn G. Stansbury, Howard R. Champion, Charles Wade, and Ronald F. Bellamy, “Understanding Combat Casualty Care Statistics,” The Journal of Trauma, Injury, Infection, and Critical Care, Vol. 60, No. 2, 2005, pp. 397-401.
97British sources on individual dead are quite detailed, though they do not necessarily provide any reliable insight into cause of wound or cause of death. See the Imperial War Museum’s website for a range of resources available for further study, including the National Archives and the Commonwealth War Graves Commission.
98John Lada and Frank A. Reister, eds., Medical Statistics in World War II, Office of the Surgeon General, Washington, D.C.: Department of the Army, 1975. Reister quotes his own data here in a later work on the Korean War, at least passively reinforcing the belief that these are valid and representative data: Frank A. Reister, Battle Casualties and Medical Statistics: U.S. Army Experience in the Korea War, Washington, D.C.: Department of the Army, 1973.
99Lada and Reister, eds., Medical Statistics, 1975, p. 20.
100Note that these figures probably would not include those soldiers listed as missing in action at the time the database was created. This accounts for tens of thousands of additional soldiers, which would significantly alter the overall data. And it is not clear if or how these data accounted for non-battle casualties.
Table 6: [Estimated] Percentage Distribution, Causative Casualty Agents, U.S. Army 1942-45
Source: Lada and Reister, Medical Statistics in World War II, p. 20
disparity from surgeon to surgeon and even recording clerk to recording clerk.101
But even if all these data were recorded with perfect accuracy – extraordinarily unlikely – they still would not constitute a random sample and could not be considered representative. This is at best a non-representative convenience sample.102 In other words, these data do not represent evidence. Nonetheless, with all caveats in place the authors draw firm conclusions about causative agents from this table, misinforming readers in perpetuity.103
Trends to this point are now clearer. Armies fielded by countries with inadequate bureaucracies, fighting in a state of disarray, retreating or advancing quickly, and perhaps on the verge of defeat generally record and provide terribly inaccurate and incomplete casualty
information. And even the best organized armies fighting under better conditions, perhaps on a drive to victory, still struggled to track their own casualties. No army described here, from the U.S. Civil War to World War I to World War II, provides us with any reliable insight into causes of wounding or death in war.
Cause-of-casualty data: End of World War II through the early 21st century
More recent data are no better and in most cases they are worse. There may have been anywhere between ten and 50 conflicts ongoing in any given year during the ensuing 80 years from the end of World War II (1946-2025).104 This includes but is far from limited to the following, listed here to drive home an important point: The Chinese Civil War, Greek Civil War, Korean War, Mau Mau Rebellion, Cyprus Insurgency, Suez Crisis, Dhofar War, Arab-
Israeli Wars, Algerian War of Independence, Vietnam War, Soviet-Afghan War, Iran-Iraq War, Lebanese Civil War, First Kashmir War, Bangladesh Independence War, Falkland Islands War, Guatemalan Civil War, Nigerian Civil War, East Timor War, Rhodesian Bush War, Eritrean Independence War, Angolan Civil War, Second Kashmir War, Mozambican War of Independence, Persian Gulf War, Former Yugoslavia Wars, NATO Afghanistan War, Iraq War, Democratic Republic of Congo Wars, Rwandan Civil War, Syrian Civil War, Libyan War of Independence, Libyan Civil War, ChadLibya War, Chadian Civil Wars, Yemeni Civil Wars, Philippines Insurgencies, Indian Insurgencies, Somalian Wars, NagornoKarabakh Wars, Malian Wars, EthiopiaTigray War, Gaza War, Ukraine Wars, and wars in Sudan and South Sudan
Given the failure of the most advanced
101Imagine one surgeon in a clean, safe, well-lit rear area hospital carefully examining a wounded or dead soldier and relaying his report to an unharried clerk who takes meticulous notes and then carefully files those notes with the appropriate adjutant section, and then another surgeon in a forward combat trauma center working perhaps under artillery fire with flashlights and processing tens of wounded and dead per hour, reporting his information to a clerk who has had little sleep and takes terrible notes. Now imagine the disparity of conditions everywhere between these two extremes and ask whether reliable consistency can be reasonably expected.
102Lada and Reister are clear that these data cannot be compared with any statistical significance to data from other conflicts due to wide variations in collection methods, completeness, representativeness, etc. Lada and Reister, eds., Medical Statistics, 1975, p. 20.
103Other available reporting on U.S. casualties from this period is equally inconsistent and offers no better possibility of representative value. Probably the best available source to understand the challenges with World War II causative agent data is Beebe and Bakey, Battle Casualties, 1952, pp. 128-136. They go into great detail and provide insights into the small number of more detailed medical studies conducted during the war that might provide better – though still non-representative – data on causation. This is compelling, old-school scientific writing. Unfortunately, the book is not available in digital format. My physical copy is falling apart. It is still worth reviewing the other cited sources including Bellamy and Zajtchuk.
104Definitions make even this broad tabulation difficult. What is a war? What is a conflict? This 10-50 figure is drawn from the 2024 PRIO, version 25.1 dataset. It is, by necessity, subjectively categorized. See Nils Petter Gleditsch, Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg, and Havard Strand, “Armed Conflict 1946-2001: A New Dataset,” Journal of Peace Research, Vol. 39, No. 5., 2002.
armies in the world to detail wound causation in the largest and probably best organized war in history (World War II), what kind of expectation can we have that armies from any of these wars might provide better data? There should be no such expectation. A few examples will be sufficient to support this point.105
Korean War, 1950-1953: U.S. Army statistics from the Korean War mirror the reliability and representativeness of the WWII data.106 They are no more accurate or representative than any previous data. On the other side, even overall North Korean and Chinese troop data are effectively a mystery; no reasonable cause-of-wound or
Snapshot of a dead North Korean soldier (circa 1950).
cause-of-death counts are readily available or could be reasonably amassed.107
Vietnam War, 1960-1975: Casualty
tracking for the American-led coalition in the Vietnam War was comparatively far better than that of the People’s Army of Viet Nam (NVA) or the People’s Liberation Armed Forces (Viet Cong).108 Overall casualty counts for both of these forces are rough estimates based in large part on badly flawed body count data.109 Despite exhaustive recordkeeping by U.S. military and defense officers, no war-wide data could be reasonably representative for any one military force or all forces.
Iran-Iraq War, 1980-1988: By the end of the Iran-Iraq War in 1988 approximately six million soldiers and paramilitary combat forces were supporting the war or arrayed
105Here are some additional resources I found while examining wars in this interim period: Robert L. Mabry, John B. Holcomb, Andrew M. Baker, Clifford C. Cloonan, John M. Uhorchak, Denver E. Perkins, Anthony J. Canfield, and John H. Hagmann, “United States Army Rangers in Somalia: An Analysis of Combat Casualties on an Urban Battlefield,” The Journal of Trauma Injury, Infection, and Critical Care, Vol. 49, No. 3, 2000, pp. 515-529; Susan L. Eskridge, Caroline A. Macera, Michael G. Galarneau, Troy L. Holbrook, Susan I. Woodruff, Andrew J. MacGregor, Deborah J. Morton, and Richard A. Shaffer, “Injuries from combat explosions in Iraq: Injury type, location, and severity,” Injury, Vol. 43, 2012, pp. 1678-1682; Jane Shen-Gunther, Richard Ellison, Charles Kuhens, Christopher J. Roach, and Steve Jarrard, “Operation Enduring Freedom: Trends in Combat Casualty Care by Forward Surgical Teams Deployed to Afghanistan,” Military Medicine, Vol. 176, No. 1 (or 67), 2011, pp. 67-78; William Eckhardt, “War-related Deaths Since 3000 BC,” Bulletin of Peace Proposals, Vol. 22, No. 4, December 1991, pp. 437-443; Bethany Lacina and Nils Petter Gleditsch, “Monitoring Trends in Global Combat: A New Dataset of Battle Deaths,” European Journal of Population, Vol. 21, No. 2/3, June 2005, pp. 145-166; and Dean Falk and Charles Hildebolt, “Annual War Deaths in Small-Scale versus State Societies Scale with Population Size Rather than Violence,” Current Anthropology, Vol. 58, No. 6, December 2017, pp. 805-813; Joshua McIntyre, “Syrian Civil War: A Systematic Review of Trauma Casualty Epidemiology,” BMJ Military Health, Vol. 166, No. 4, 2019, pp. 261-265.
106Frank A. Reister, Battle Casualties and Medical Statistics: U.S. Army Experience in the Korea War, Washington, D.C.: Department of the Army, 1973. In this volume Reister strips away some of the important caveats from his book with Lada, compounding the WWII data and Korean War data into a seemingly reliable source.
107See, for example: Xu Yan and Li Xixo-Bing, “The Chinese Forces and Their Casualties in the Korean War: Facts and Statistics,” Chinese Historians, Vol. 6, No. 2, 1993, pp. 45-48. Note that I did not fully review this article given its insultingly high $56.00 price tag at Taylor & Francis Online, but the abstract states, “Since the Korean War ended in 1953, however, the numbers of the Chinese troops involved in the Korean War and their casualties have remained secret.” Yan and Xixo-Bing, “The Chinese Forces,” 1993, p. 45. Also see: “New Evidence on North Korean War Losses,” Wilson Center, August 1, 2001. Also see this interesting aggregation of purported sources on casualties in the Korean War. Even if the data are not accurately represented the disparities between sources are interesting: Matthew White, Korean War Casualty Information: North/South Korean and Chinese Casualties, website, Korean War Educator.
108For U.S. figures see the National Archives, Vietnam War U.S. Military Fatal Casualty Statistics. In general, the North Vietnamese Army (NVA) is more accurately referred to as the People’s Army of Viet Nam (PAVN), while the forces of the southern communist National Liberation Front (NLF) are – arguably – better referred to as the People’s Liberation Armed Forces rather than the intentionally pejorative Viet Cong. For more on the wounding of U.S. personnel in the Vietnam War: James G. Garrick, Naval Battle Casualty Study, Springfield, Vir.: National Technical Information Service, December 1973; L.A. Palinkas and P. Coben, Combat Casualties Among U.S. Marine Corps Personnel in Vietnam: 1964-1972, San Diego, Calif.: Naval Health Research Center, 1985, and others.
109To get a sense of the complexity of the data challenge in this case see Connable, Embracing the Fog, 2012, and Thomas C. Thayer, ed., A Systems Analysis View of the Vietnam War, 1956-1972, Volume 6: Casualties and Losses, Washington, D.C., Office of the Secretary of Defense, February 17, 1975. Thayer describes in great detail how both U.S. military and Defense Department offices gathered and estimated information, assembled data and then attempted to build evidence on adversary casualties. Despite all the caveats that Thayer provides he still generates precise figures that were repeated and used to support contemporaneous decision-making and that are still in use in the mid-2020s. Probably the best analysis of U.S. casualties is a report describing the detailed medical examination of 500 U.S. Army fatalities between 1967 and 1968. This report details wounds and ballistics and identifies cause of wound for each case. However, a careful review of the cause-of-wound data shows that 96 of the 500 bodies had received wounds that could not be clearly identified, generating a nearly 20 percent gap before accounting for probable errors in just this non-random sample. One broader medical statistical study exists (cited below) but even if a precise, accurate and representative causative count existed it would not balance out the near total lack of information from the other forces. See: Ian Sunshine, Analysis of 500 US Army Combat Fatalities in Vietnam, July 1967 to November 1968, Edgewood Arsenal, Md.: Department of the Army, September 1970, p. 20 and others. Unfortunately, the original U.S. Army statistical study is not available online. It is referenced in: Spurgeon Neel, Medical Support of the U.S. Army in Vietnam, 1965-1970, Washington, D.C.: Department of the Army, 1991, p. 54. Original source is cited as: U.S. Army, Statistical Data on Army Troops Wounded in Vietnam, January 1965-June 1970, Medical Statistics Agency, Office of the Surgeon General (undated, not found online). Note that Neel integrates the Vietnam data from this report with the previously cited WWII and Korean War data, compounding the inaccurate perception that this is longitudinally-relevant evidence. Also see the Wound Data and Munitions Effectiveness Team (WDMET) database. WDMET includes examination records of 7,989 wounded and-or killed soldiers which, according to the coders, represented less than four percent of all injured or killed in the conflict. I presume that figure to account only for American casualties and not North or South Vietnamese, allied, militia or other forces. Howard R. Champion and Mary M. Lawnick, Coding the WDMET Database, Fort Detrick, Md.: U.S. Army Medical Research and Materiel Command, 2006. Also see this video regarding the original data collection effort: Ray Gauvin and Darrell Lane, “Seeking to Protect The Force: Experiences With The Wound Data and Munitions Effectiveness Team,” The West Point Center for Oral History, April 24, 2023.
against each other along the hundredsof-kilometers long front.110 Anywhere from 600,000 to one million soldiers on both sides were killed in the war, many being left where they were killed due to ongoing fighting or the presence of poisonous gas.111 There are no reliable estimates of wound causation for either or both sides.112
Iraq War, 2003: As with the 1991 Gulf War, U.S.-led coalition allies kept excellent casualty records for the 2003 invasion of Iraq. Probably every coalition casualty can be accounted for along with a reasonably accurate cause-of-wound, cause-of-death. However, in both the 1991 and 2003 wars, medical staffs capable of handling an order of magnitude of additional casualties were immediately on hand to treat, track and record coalition injuries.113 Records of Iraqi casualties from the 2003 war are effectively non-existent given the collapse of both the military and regime at the end of the invasion.114 Therefore, it would not be possible to describe or estimate causation for the vast majority of casualties (Iraqi) in this war.
Ethiopia-Tigray War, 2021: Little is known
“U.S.-led coalition allies kept excellent casualty records for the 2003 invasion of Iraq. Probably every coalition casualty can be accounted for along with a reasonably accurate cause-of-wound, cause-of-death.”
110These numbers are only rough estimates. Neither side retained or publicly provided precise or provably accurate enlistment, retention or casualty data throughout this eight-year war. For both a breakdown of these numbers and citation see: Ben Connable, Iraqi Army Will to Fight: A Will-to-Fight Case Study with Lessons for Western Security Force Assistance, Santa Monica, Calif.: RAND Corporation, 2022, p. 40 and others.
111Estimates vary considerably and no works I reviewed claimed accuracy. See, for example: Connable, Iraqi Army, 2022, pp. 37-38; Pierre Razoux with Nicholas Elliott, tr., The Iran-Iraq War, Cambridge, Mass.: The Belknap Press of Harvard University Press, 2015; Dilip Hiro, The Longest War: The Iran-Iraq Military Conflict, London, UK: Grafton Books, 1989; Rob Johnson, The Iran-Iraq War, London, UK: Palgrave-Macmillan, 2011; Anthony H. Cordesman and Abraham R. Wagner, The Lessons of Modern War, Volume II: The Iran-Iraq War, Boulder, Colo.: Westview Press, 1990.
112One study purported to describe detailed examination of Iranian casualties. It reports, or at least clearly suggests, that Iranian medical records were available for fewer than 50,000 total casualties. That would constitute approximately ten percent of all Iranian casualties on a low-end estimate without random sampling. There is no indication that cause-of-wound or cause-of-death analyses were conducted on any of these 50,000 corpses or on the possibly non-random sample of 1,176 corpses examined for this specific study. Masoud Saghafi-Nia, Seyed Masoud Khatami, Nahid Nafissi, and Yadollah Rezaei, “Triage of war-injured troops in the Iran-Iraq War,” Journal of Research in Medical Sciences, Vol. 13, No. 1, 2008, pp. 8-11. Another study describes causation of wounds from a non-random examination of 7,352 Iranian war casualties. However, the data focused only on wounded and not on fatalities, and then only on wounds to extremities: Mohammad Zarei, Mahmoud Farzan, Mohammad Javad Deghani Firoozbadi, Furqan Mohammed Yaseen Khan, and Mehdi Tavokoli, “Extremity war injuries: A retrospective study of the Iran-Iraq War,” Archives of Trauma Research, Vo, 8, No. 1, January 2019, pp. 28-32.
113Expectations for coalition casualties in both the 1991 Gulf War and the 2003 invasion of Iraq far exceeded actual casualties, though in both cases tens of thousands of medical personnel were deployed and ready to support combat forces. For example, in preparation for the 1991 Gulf War, 44 Army-related hospitals were prepared to receive at least 13,580 patients with a surge of up to 25,000 additional patients. A total of 23,493 Army medical personnel alone (not including Air Force, Navy and civilian personnel, or any coalition medical personnel from the United Kingdom, France, etc.) were directly deployed to the combat area to treat fewer than 500 wounded-in-action U.S. soldiers. See, for example: Frank F. Ledford, Jr., “Medical Support for Operation Desert Storm,” The Journal of the U.S. Army Medical Department, January-February 1992, pp. 3-6; and U.S. Government Accountability Office, Operation Desert Storm: Full Army Medical Capability Not Achieved, Washington, D.C., August, 1992. For more on Operation Iraqi Freedom see: James M. Zouris, G. Jay Walker, Judy Dye, and Michael Galarneau, “Wounding Patterns for U.S. Marines and Sailors during Operation Iraqi Freedom, Major Combat Phase,” Military Medicine, Vol. 171, No. 3, 2006, pp. 246-252. Also see: Rigo Hoencamp, Eric Vermetten, Edward C.T.H. Tan, Hein Putter, Luke P.H. Leenen, and Jaap F. Hamming, “Systematic review of the prevalence and characteristics of battle casualties from NATO coalition forces in Iraq and Afghanistan,” Injury, Vol. 45, 2014, pp. 1028-1034; and Noelle N. Gronroos, James M. Zouris, and Amber L. Wade, “Odds of Hospitalization Among Marine Corps Personnel by Military Occupational Specialty and Causative Agents During OEF and OIF,” Military Medicine, Vol. 174, No. 7, 2009, pp. 715-720.
114This gap in Iraq data also exists for the 1991 Gulf War casualties. See Connable, Iraqi Army, 2022, and cited sources in chapter four.
about this massive war, fought between perhaps several hundred thousand soldiers in total. All sides in this war kept journalists and external observers from the battlefield and none published detailed casualty reports, if they existed. Even rough toplevel estimates are not to be trusted. Perhaps as few as 100,000 or well over 800,000 soldiers and civilians were killed in the war.115 No cause-of-wound, cause-ofdeath estimates are possible.
If we take into account all the wars fought between the end of World War II and 2025 – not including the 2022-2025 phase of the Ukraine War – we might be able to identify the weapons and munitions that caused wounds or death in as many as 100,000 cases, and only then from convenience sampling and with varying and uncertain accuracy and reliability.116 We would know very little about wound causation in many major wars like the Ethiopia-Tigray War. And even if some studies were indeed representative of all casualties within given wars (I found none), they could not in any way be globally representative.
Now refer back to the standard for knowing. In order to identify a scientifically representative sample of casualty causation in a war we need to have at least a reasonable idea of the total number of casualties. At the global level we would need to have a reasonable knowledge of the total number of combat casualties
in all wars. That overall casualty number, however, is practicably unknowable.
Why is that the case? War databases are useful for some things, but casualty estimates have always been problematic.117 As I wrote above, even the best efforts cannot overcome real-world gaps in knowledge. Top-level estimates generally blend combat and non-combat casualties without distinction. When they are applied, definitions of “combat casualty” vary greatly from source to source and database to database. Given these gaps, uncertainties and disparities in combat casualty data from the U.S. Civil War through early 2022, it is safe to conclude the following: Preceding the expanded Russian invasion of Ukraine in 2022, there were no accurate, reliable or representative data on cause of wound or death in global warfare.118
In other words, despite all the historical efforts to track and measure casualty data in war, we have never had a reliable, extrapolatable understanding – evidence or proof – of combat casualty causation. Therefore, any successful measurement of this factor in the Ukraine War would be novel and groundbreaking. As the reader may safely assume at this point, no such novelty or broken ground exists. Still, it is worth examining the Ukraine War resources in greater detail to get all the way to the bottom of the 70 percent drone-
casualty figure and to set the stage for a more reasonable approach.
Ukraine War casualty causation
As I wrote up front, the 70 percent figure –‘drones cause 70 percent of all casualties in Ukraine’ – appears to be uncited, contradicted by the same source and of uncertain provenance. I also wrote that one unsupported statistic does not obviate the possibility that better measurement might be available. We know the eventual outcome here. But it is important to peel back the layers of data and consider many confounding factors in the Ukraine War in order to improve understanding of other measurements in this war and in others. And no matter how remote I believe the chances might be, I must leave open the possibility that new scholarship will prove me wrong.
Is there evidence to support the 70 percent figure or another measurement?
How might we ascertain cause-of-wound, cause-of-death for a reliable statistic for one or both sides in the Ukraine War? As a reminder, here are the four rudimentary standards I set to build that kind of data and, ideally, evidence:
1. Clear definition of a casualty;
2. Representative and reliable casualty data from one or both sides;
3. Reliable evidence of wound causation
115It is barely worth citing these data since all reasonable sources effectively acknowledge that they are guesses. I led a research team that engaged with some of the leading experts on Ethiopia and conflict violence within Ethiopia; none had any significant insight into the fighting let alone casualty data. For example, see: Teklehaymanot G. Weldemichel, “Tigray war: Modern geographies of mass violence and the invisibilization of populations,” Political Geography, Vol. 118, April 2025, no page numbers; Helena Laurent, “Ethiopia: The daunting task of reporting the Tigray conflict,” DW, January 27, 2022. Also see the work of the Tigray War Project at Ghent University and Every Casualty Counts, with some data primarily on civilian casualties. For more sources on this war and more information on the gaps in data see: Ben Connable, Ground Combat: Puncturing the Myths of Modern War, Washington, D.C.: Georgetown University, 2025; and the Ground Combat Database-Version 1, Excel database of modern battles.
116This is a rough estimate based on my review of publicly available causation studies, and then only those published or republished in English. I was able to identify fewer than 50,000 casualties studied and must assume other studies exist. Applying this assumption I doubled the highest end of my review to 100,000. This is what might be best referred to as a SWAG, or Scientific Wild Ass Guess, and should be used for nothing.
117I reviewed the Uppsala Conflict Data Program, the Correlates of War project, Costs of War project, Global Peace Index, Project Mars, and others. Explaining the gaps within and comparisons between all these datasets and others would require a separate paper. Luckily, Milton Leitenberg at Cornell University has conducted a detailed, thoroughly cited and insightful analysis of war databasing challenges. His table, running from pages 73-77, shows how little understanding there is of specifically military casualties in war just through the year 2000. I attempted to code combat casualties in the 423 battles I examined from 2003-2022 in the Ground Combat Database. Despite citing well over 20,000 sources my results were paltry. In the vast majority of cases even at the battle level I was unable to determine gross casualty rates except through scant and often biased self-reporting from military units. See: Milton Leitenberg, Deaths in Wars and Conflicts in the 20th Century, occasional paper, 3rd edition, Ithica, N.Y.: Cornell University, 2006. Also see: Douglas M. Gibler and Steven V. Miller, “The Militarized Interstate Events (MIE) dataset, 1816-2014,” Conflict Management and Peace Science, Vol. 41, No. 4, pp. 463-481; Gleditsch, et al., “Armed Conflict 1946-2001,” 2002; Ben Knight, “Global conflicts: Death toll at highest in 21st Century,” DW, June 28, 2023; as well as the dataset websites, all readily available through basic search online.
118This statement applies to our understanding of military casualties in wars at the service and country level, and to war writ large. It would of course be possible to assess military casualties in contained cases, like individual engagements or even larger battles. This has been done many times but idiosyncratically and without generating broader evidence.
for that random sample; 4. Time and space bounding for the selected data.
I conducted a review of publicly-available Ukrainian and Russian military statistics and medical evidence. While the Ukrainian Ministry of Defense puts out detailed statistics on a daily basis, it does not provide the raw data behind those stats for examination. Russia rarely puts out casualty data and when it does the data are highly suspicious; see below. Therefore, while the official counts provided by both countries are interesting for various reasons, they cannot be used to build representative and reliable evidence.
Perhaps medical studies might provide some insight. With that thought in mind I reviewed publicly-available medical reporting from both sides in the hope that I might find a study similar to the ones conducted during World War II, etc. Searching in English, Ukrainian and Russian, I found a range of interesting journal articles and reports.119 Two Russian medical reports specifically described causation, one for wounds and one for deaths.
Russian medical reports on casualties
First I found an article on an ostensibly independent Russian media website,
A Ukrainian soldier burying the remains of a Russian serviceman following the liberation of the village of Lukianivka on March 28th, 2022.
Picture: Serhii Nuzhnenko (Radio Free Europe/ Radio Liberty)/the Collection of war.ukraine.ua
Verstka, reporting casualty data recorded by Russian military anesthesiologists.120 According to these data, which represent a non-scientific sample of Russian soldiers treated at a Role-2 medical facility, 86.5 percent of soldiers were wounded by some sort of shell fragment, six percent were shot, three percent were burned and 4.6 percent of wounds were unlabeled.121
This report does not tell us what kind of weapon caused these wounds. Reported data are opaque (not publicly available) and could have been fabricated or misreported. These data also are
non-random, meaning there is no way to extrapolate this study to all Russian casualties. Therefore, this information is vaguely interesting but no more generalizable than any medical wound data from previous wars.
Next I found what appeared at first glance to be a more detailed Russian medical report with some incomplete cause-ofwound data. This article appears to have been re-reported in Newsweek, which in turn cited but did not link to “Vyorstka,” another transliteration for Вëрстка (Verstka, loosely translated as “layout”).122 Newsweek wrote that 74.5 percent of deaths were due to explosions of unspecified origin, and 14.7 percent were due to gunshot wounds.
It appears this Newsweek article may have indirectly referred to an article published in the Russian Military Medical Journal, “Causes of death of servicemen in modern warfare.”123 In the abstract the authors report that the primary cause of death amongst the 608 casualties studied was from combined injuries (38.2 percent), which I took to mean a mix of blast, burn, fragments and maybe gunshots.124 Second place at 18.8 percent was unknown cause, while third place was gunshot wounds to the head at 14.1 percent.125
119One of these was written by my colleagues at the Center for Naval Analyses. See: Gabriela Iveliz Rosa-Hernandez and Patrick Enochs, We Need a Medic! The Russian Military Medicine Experience in Ukraine, Arlington, Vir.: Center for Naval Analyses, November 2024. This is a good source for original-language material. I used it to find the first of the two listed Russian reports.
120“Осколки в руках и ногах. Какие ранения
Verstka, June 25, 2024.
121A Role 2 facility can be roughly equated to a small field hospital. Role 1 is a front line trauma center, while Role 3 would be a full hospital equivalent. There is a table in this article that seems to show comparative wound statistics from Afghanistan, Chechnya, and Ukraine. This would be a potentially valuable chart but it is hard to read and I could not access the original source. “Осколки в руках и ногах,” Vertska, June 25, 2024.
122“Russian Troops’ Most Common Cause of Death Revealed in Report,” Newsweek, October 17, 2024. As of August 24, 2025: newsweek.com/russia-ukraine-warexplosions-1970605.
123This article was also cited in the CNA report cited here. R.R. Kasmirov, I.M. Samokhvalov, A.A. Zavrazhnov, V.V. Kudryashov, S.A. Kovalenko, and I.A. Tolmachev, “Causes of Death of Russian Servicemen in Modern Warfare,” Military Medical Journal (or Voenno-medicinskij žurnal), Vol. 345, No. 8, 2024, pp. 11-16. Readers are free to discover the link to that article on their own. As the Newsweek article notes, at least the abstract of this article does not refer to Ukraine. It almost certainly reflects battlefield casualties in Ukraine, but this is another open question that casts doubt upon the utility of the data.
124Since this is an official Russian military publication I declined to purchase access; I will not indirectly but knowingly support Russia’s illegal military operation in Ukraine by funding the Russian military. This abstract is available through Google Scholar and other journal aggregating websites. It is important to acknowledge that the Englishlanguage abstract may not accurately reflect the data provided in the body text.
125This “severe gunshot wounds to the head” statistic was interesting given the anecdotal evidence of Russian suicide by gunshot across the Ukraine front. Having observed between one and two hours of combat footage per day since perhaps mid-2022, I can attest to probably 200 cases of suicide by grenade and gunshot. These typically are viewed and recorded by Ukrainian drones. Grenade suicides frequently failed with expectedly gruesome consequences, all observable by fellow soldiers, so it would make sense that gunshot suicides would become more common over time. One unconfirmed Ukrainian report suggests soldiers are given suicide instructions. For more on this phenomenon see: Sergio Miller, “The Russian Army Death Cult,” Wavell Room, May 14, 2024; and “Russian soldiers fighting in Ukraine given ‘suicide instructions’,” UKRINFORM, April 12, 2024.
Clearly the numbers between these articles do not align, so it is hard to tell if there is yet another source, if this is a case of circular reporting or if it is just inaccurate regurgitation. Either way, none of these cited Russian data are representative, random or reliable, and we have no way of knowing how medical professionals in different studies labeled different types of wounds. And further, none of this refers to drones so it does nothing to clear up our central question: Which weapons cause most casualties?
Paramedics provide medical care to a wounded Ukrainian soldier at a first-aid post in Bakhmut, Donetsk region, on November 22nd, 2022.
Picture: Serhii Nuzhnenko (Radio Free Europe/ Radio Liberty)/the Collection of war.ukraine.ua
Ukrainian medical reports on casualties I examined a range of Ukrainian medical reporting, looking for cause-of-wound or cause-of-death data. Due to close collaboration and relationships with Western medical experts and institutions, more Ukrainian than Russian data and articles are available in the public domain.126 Still, I found no professional articles from the Ukrainian side claiming to offer representative data.127 So this is also a dead end in terms of attributing causation. It is, however, informative to see the continuity of research methods from war to war.
As in previous wars, some medical teams in Ukraine have conducted large-n or large batch casualty case studies. One of these
examined 415 patients to help describe maxillofacial injuries and treatment.128 Most of the injuries examined were caused by unspecified blasts from “shells, mines, bombs, missiles, and drones”.129 It then refers to many of the same data I cite above to put these numbers in historical context. In doing so, the authors represent them as generalizable evidence rather than what they are: small, non-representative datasets.130 This is a continuation of approach from at least World War I.
Ukrainian medical professionals are rightly focused on improving medical care for their patients and (as reflected in the literature) seem to be generally less focused on building datasets. And since we are attempting to understand a war that is still underway, reporting lag is generating perception lag. In other words: Studies conducted in 2023 might be published in 2024, or even in 2025, giving the impression that old data is new and that the present reality does not yet exist. In any event, circumstances on the battlefield, confounding variables and other factors further degrade chances of building a representative dataset.
Why are Russian casualty statistics considered to be so unreliable?
Now that we are fairly confident the professional research data cannot show reliable, representative weapons-to-wound causation, we can start to layer in a range of confounding conditions and factors on the Ukraine battlefield that further reduce prospective accuracy.
126I will only cite here articles relevant to cause of wound or cause of death. Several journals published Ukrainian medical articles or articles on Ukrainian casualties. See, for example, the Ukrainian Journal of Military Medicine, International Journal of Surgery Case Reports, Conflict and Health, Journal of Trauma and Acute Care Surgery, Military Medicine, Archives of Craniofacial Surgery, Legal Medicine, Trauma Surgery and Acute Care Open, Journal of Neurotrauma, EC Emergency Medicine and Critical Care, and others.
127Other articles examined include, but are not limited to: Lynn Lieberman Lawry, Amandari Kanagaratnam, Ashleigh Roberds, Jessica Korona-Bailey, Luke Juman, Miranda Janvrin, Zoe Amowitz, Tiffany E. Hamm, John Maddox, Oleh Berezyuk, and Tracey Perez Koehlmoos, “A qualitative assessment of disease and non-battle injuries in Ukraine since the Russian invasion,” Conflict and Health, Vol. 19, No. 19, 2025; Mikola Sinyuk, V. Polishchuk, P. Yuschak, and I. Burachok, “Management of war-related facial wounds in Ukraine: the Lviv military hospital experience,” BMJ Military Health, Vol. 171, 2025, pp. 12-15; Lynn Lieberman Lawry, V. Mani, T.E. Hamm, M. Janvrin, L. Juman, J. Korona-Bailey, J. Maddox, O. Berezyuk, A.J. Schoenfeld, and T.P. Koehlmoos, “Qualitative assessment of combat-related injury patterns and injury prevention in Ukraine since the Russian invasion,” BMJ Military Health, electronic publication, 2024.
128Oleksandr Prysiazhniuk, Roman Palyvoda, Yurii Chepurnyi, Tetiana Pavylychuk, Denis Chernogorskyi, Igor Fedirko, Yaroslav Sazanskyi, Danylo Kalashnikov, and Andrii Kopchak, “War-related maxillofacial injuries in Ukraine: a retrospective multicenter study,” Archives of Craniofacial Surgery, Vol. 26, No. 2, 2024, pp. 51-58.
129Prysiazhniuk, et al., “War-related,” 2024, p. 56.
130This is partly due to the prima facie representation of the same non-representative data by Joshua McIntyre in what he calls a meta-analysis of combat casualty data. Since none of the analyses McIntyre cited had equivalent datasets, methods or purposes, this is at best a comparative review, not a scientific meta-analysis. In any event, by improperly re-reporting these data at face value, McIntyre cements them and encourages others to repeat them as representative facts. McIntyre, “Syrian Civil War,” 2020.
Since we cannot know or even reliably estimate casualties for the whole Ukraine War without having useful data from both sides, we can focus on one side to narrow the research burden. Casualty statistics from both sides are unreliable, but Russian statistics are notoriously bad and almost certainly worse than those of the Ukrainian Ministry of Defense and of Western military establishments.
Why are Russian numbers considered to be so unreliable? There are many reasons for doubt:131
Propaganda. Most military forces try to hide their weaknesses in war, but the Russian Ministry of Defense has long considered statistics as manipulable propaganda tools.132 Official Russian casualty statistics since February 2022 are hard to find and, when they are available, almost impossibly low.133 Conversely, Russian statistics on
“Most military forces try to hide their weaknesses in war, but the Russian Ministry of Defense has long considered statistics as manipulable propaganda tools.”
Ukrainian losses are impossibly high.134 It would be reasonable to state that there are no readily available or reliable official Russian statistics on Russian casualties.
Mercenaries and foreign troops. Russia routinely employs mercenaries both in distinct units, as it did with the Wagner Group and other private military
companies in the Battle of Bakhmut, and also integrated within its own armed forces.135 Possibly well over 100,000 mercenaries and other foreign troops from across Russia, Central Asia, North Korea, Africa, China, South Asia and elsewhere have fought on the front lines.136 Mercenary casualties generally are not tracked by official government sources and are exceptionally hard to identify by other means.
Corruption and the missing. Extensive anecdotal evidence – which does not constitute reliable scientific evidence – suggests that some Russian commanders are putting little to no effort into recovering front-line casualties.137 In some cases the combat situation may be too difficult, but in others it appears that corrupt commanders prefer to list soldiers as missing so they can drain their bank accounts and cynically help the Russian government to keep official
131These are anecdotal or expert observations and do not constitute scientific evidence or proof. They are intended only to inform our collective understanding of the existing data gaps in Ukraine War casualty data.
132For more on the Soviet, then Russian approach to propaganda see: Ben Connable, Stephanie Young, Stephanie Pezard, Andrew Radin, Raphael S. Cohen, Katya Migacheva, and James Sladden, Russia’s Hostile Measures: Combating Russian Gray Zone Aggression Against NATO in the Contact, Blunt, and Surge Layers of Competition,” Santa Monica, Calif.: RAND Corporation, 2020. See the cited sources in chapters 1 and 2 for original material and more detailed analyses.
133Russia’s Ministry of Defense website has previously published some casualty figures, though official running totals do not appear to be provided. In general, the Russian government has dodged questions about overall casualty figures. I do not provide hyperlinks to Russian websites for security reasons. See, for example, these semi-official reports from government-friendly news agencies and websites like Lenta, Tass, and Pulse19: Nikita Abramov, “Песков
Lenta.ru, April 17, 2024;
СВО,” Lenta.ru, October 14, 2024; “Шойгу:
21 июня,” Pulse19, June 21, 2025.
Tass.ru, March 2,
Danilov,
134See, for example, this summary ostensibly derived from Ministry of Defense figures. It suggests that the Ukrainians had lost over 65,000 tanks and armored vehicles from the start of the war through August 26, 2025. That would far exceed the total number of vehicles in the Ukrainian inventory collectively (and perhaps also in the Russian inventory, combined) throughout the entire war and is effectively an impossible statistic. This is not an official website but it is indicative of the generally exaggerated approach to tabulation. “ПОТЕРИ
Mskvremya.ru, August 26, 2025. For other perspectives see: Dmitry Kobak, Alexey Bessudnov, Alexander Ershov, Tatiana Mikhailova, Alexey Rakhsha, “War Fatalities in Russia 2022-2023 Estimated Via Excess Male Mortality: A Research Note,” Demography, Vol. 62, No. 2, April 1, 2025, pp. 335-347; Valery Dzutsati, “Russian Casualties in Ukraine Continue to Rise,” Eurasia Daily Monitor, July 16, 2024; Benjamin J. Radford, Yaoyao Dai, Niklas Stoehr, Aaron Schein, Mya Fernandez, and Hanif Sajid, “Estimating troop losses on both sides in the Russia-Ukraine War,” analysis blog, The Loop, undated; “Russian losses in the war with Ukraine. Mediazona count, updated,” Mediazona, as of September 13, 2025.
135For more on Wagner in Bakhmut, in Ukraine, and related issues see, for example: Bryan Powers, “The Battle for Bakhmut: When Is a Battlefield Loss a Strategic Victory?,” U.S. Army, March 31, 2025; Jakub Ber, “From Popasna to Bakhmut. The Wagner Group in the Russia-Ukraine War,” OSW, April 28, 2023; “The price of Bakhmut. We reveal the staggering toll of Russia’s bloodiest battle since WW2 and Wagner’s inmates recruited to fight it,” Mediazona, June 10, 2024.
136This is a very rough estimate drawn from what may have been 50,000 or more mercenary troops engaged only at Bakhmut, at least 10,000 North Korean troops, as well as consistent employment of foreign fighters and mercenaries across the battlefield and over time. In my review of published and video evidence I have seen both mercenaries and enlisted soldiers who self-identified as Chinese, Cuban, Belarusian, Georgian, Central Asian, Indian, African (Nigerian, Cameroonian, and others), American, Western European and Eastern European. This rough estimate should not be taken at face value or repeated as information or evidence. On foreign troops see, for example: “Putin’s recruits: The young Africans fighting for Russia in Ukraine,” France 24, June 16, 2025; Miłosz Bartosiewicz and Piotr Żochowski, “Putin’s foreign legion: Foreigners fighting in the war with Ukraine,” OSW, February 20, 2024.
137This subjective insight is derived primarily from a wide array of Telegram messages and videos, YouTube videos, and other social media material that show what are purported to be messages from Russian soldiers at the front lines describing these activities. For example, see the YouTube channel Insights from Ukraine and Russia. This is a necessarily biased source but it presents what appear to be frontline videos from Russian soldiers explaining these activities.
casualties low and to avoid paying death benefits to soldiers’ families.
Caste system. A formal caste system in Russian front line units compounds the corruption factor.138 There are five general types of soldiers serving with the Russian land forces: (1) prisoners and other undesirables who serve in Storm-Z or Storm-V units; (2) foreign soldiers and mercenaries; (3) mobilized soldiers; (4) contract soldiers; and (5) independent units.139 Commanders tend to treat Storm-Z/V soldiers as cannon fodder while protecting their professional contract soldiers, and they may have no interest in or ability to report mercenary or independent unit casualties. According to unverified anecdotal reports cited here, caste position might determine whether or not a casualty is treated or officially reported.
Wounded but not officially wounded. Under enormous pressure from higher echelons of command, tactical commanders have routinely forced wounded Russian soldiers – typically
from the Storm units – back to the front lines and in some cases directly into infantry assaults while on crutches andor with large, festering wounds.140 So even without explicit corruption at play, they may be incentivized not to report the original wound to prevent the soldiers from being hospitalized and removed from combat.141
Taking all this into account it is clear that even Russia’s own secret, unreleased casualty data is probably unreliable. In other words: The Russians may have no idea what their own real casualty numbers might be. If that is the case then only Ukrainian or perhaps Western government estimates of Russian casualties might provide reliable wound causation data. This is where a range of additional biases and elided confounding factors come into play.
Battlefield biases and confounding data gaps
There are effectively two major biases and at least 20 of what might be called confounding data gaps, or other battlefield information that would directly reduce the accuracy and reliability of (in this case) the
70 percent drone casualty statistic.142
Bias 1 – technical: By design, drones have built-in cameras and the feeds from these cameras are frequently linked to communication centers that record and store all video feeds. Therefore, even while a successful drone strike may not necessarily cause a casualty, the availability of video evidence from nearly every drone strike creates a significant bias to weight the value of drones over weapons that do not have built-in cameras.
Bias 2 – dispositional: A considerable proportion of the Ukrainian military and also the civilian production sector is dedicated to drone operations.143 While there are no indications they are stacking the deck on casualty estimates, these people naturally tend to value drones over other weapons in part because of their roles and interests. Semiindependent drone operation units are also highly organized and are quick to deploy video montages. These expert units, like Birds of Magyar (Madyar), SIGNUM and others, play a critical role
138There has long been a caste system within the Soviet and then Russian armed forces. Prior to the Ukraine War there were generally two overall castes, conscript and contract soldiers, and then within those castes a sub-caste of non-Russian ethnicity versus white Russian ethnicity. During periods of war, and particularly during World War II, prisoners were forced into a lowest-caste system and sometimes used as cannon fodder. Efforts to expand available troops through mobilization, extensive use of prisoners and recategorization of soldiers after military punishment, integration of mercenaries, and use of independent units has contributed to what appears to be a new, more multi-layered and complex caste system. Prisoners and other soldiers who have received punishment still appear to take the brunt of the casualties in Storm units. Unfortunately, much of the information coming out of Russia is anecdotal, hard to verify, and extraordinarily difficult to translate into data or evidence. For background and more recent information on the caste system see, for example: Roger R. Reese, Why Stalin’s Soldiers Fought: The Red Army’s Military Effectiveness in World War II, Lawrence, Kan.: University of Kansas Press, 2011; Herbert Goldhamer, The Soviet Soldier: Soviet Military Management at the Troop Level, Santa Monica, Calif.: RAND Corporation, September 1975; Ytzhak Tarasulo, “A Profile Of the Soviet Soldier,” Armed Forces & Society, Vol. 11, No. 2, Winter 1985, pp. 221-234; Roman Kohanets, “Russia’s War Creates ‘Caste System’ Valuing Soldiers Unequally by Region,” United24 Media, August 15, 2025; and Kristaps Andrejsons, “Russia’s Convict-Soldiers Have Their Own Brutal Rules,” FP, May 16, 2023.
139These would include troops from non-defense ministries including the national guard, or Rosgvardia, regional units like the Chechen Akhmat, and intelligence special operations units.
140This insight is also derived from anecdotal evidence, though it does appear that many of the videos purporting to show wounded Russian soldiers returned to combat are accurate. Ukrainian drone footage routinely published on social media shows what are clearly Russian soldiers on crutches walking into assaults. For example, see: David Hambling, “Why Is Russia Sending Assault Troopers Into Battle On Crutches?,” Forbes, May 06, 2025; Daniel Kosoy, “Russia Is Sending Its Wounded Soldiers Into Meat Grinder Assaults, Even on Crutches,” United24 Media, February 17, 2025; and as of August 26, 2025: Nick Paton Walsh, “Injured Russian soldiers are being returned to the frontline,” CNN, February 22, 2025.
141There may be a dilemma here for some of the more corrupt commanders. If they report the wound and the soldier receives a wound-related payment, the commander can use the soldier’s bank card to extract some or all of his funds. If they do not report the wound they might be able to more easily force the soldier back into the assault. It is also possible that a commander might do both: report the wound, take the cash, and still force the wounded soldier forward.
142I hesitate to call these confounding variables because that term has a distinct scientific meaning. One might interpret the data I list here to be confounding variables, or they might prefer to call them something else entirely. For more on the basic concept of confounding variables see, for example: Chittaranjan Adrade, “Confounding by indication, Confounding Variables, Covariates, and Independent Variables: Knowing What These Terms Mean and When to Use Which Term,” Indian Journal of Psychological Medicine, Vol. 46, No. 1, February 2024, pp. 78-80; and this simple web explanation from Penn State University.
143It is not clear what percentage of the entire Ukrainian Armed Forces (UAF) is dedicated to drone operations, but it is a considerable proportion. For example, I was told by a well-known expert on the UAF that an infantry brigade might consist of over 20 percent drone operators and logisticians. Brigade command posts are dominated by drone feeds, and large portions of the Ukrainian front lines depend on large-scale drone support.
in the Ukrainian information warfare strategy and receive proportionally more attention than other units.144
These biases may or may not have influenced the emergence of the original 70-percent drone-casualty statistic.145 Real-world gaps in other data –prospectively confounding data – may be relatively easier to understand and perhaps to track over time.
If drones cause 70 percent of all casualties in Ukraine, that would necessarily mean that all other weapons and munitions together would cause only 30 percent of overall casualties. If we had a clear casualty count for both sides we could simply track all drone kills, compare those to the overall casualties, and determine a reliable percentage without examining any other data.146 Absent accurate casualty figures we need to clearly account for the remaining prospective casualty causing agents on
the Ukrainian battlefield. Those would include, but would not be limited to:
Infantry weapons: Rifles, hand grenades, submachine guns, machine guns, sniper rifles, heavy machine guns and grenade launchers, small mortars, recoilless rifles like the SPG-9, rocket launchers like the AT-4 and RPG-series, pistols, knives and other weapons are used to cause casualties in Ukraine.147
Armored vehicle weapons: Main cannons from tanks and lighter single-shot and automatic cannons from infantry fighting vehicles, vehicle mounted machine guns, vehicle mounted missiles like the TOW II, and crushing under wheels or treads cause casualties.148
Artillery and heavy mortars: Possibly between 300,000 and 2,000,000 artillery and heavy mortar shells of all type have been fired by
both sides in the Ukraine War each month.149 Volume of fires has reduced over time due to a range of factors.
A Ukrainian soldier walks past a captured Russian tank in the Kharkiv region, September 14, 2022.
Picture: Serhii Nuzhnenko (Radio Free Europe/Radio Liberty)/the Collection of war.ukraine.ua
144For more on this unit see: David Hambling, “Ukraine’s Elite Drone Unite Is A Lethal, Fast Growing Tech Start Up,” Forbes, April 14, 2025; David Kirichenko, “Madyar: Ukraine’s Birdman Hunting Worms From the Sky,” Kyiv Post, August 11, 2025; Yuri Zoria, “Elite Ukrainian drone unit becomes combat brigade, triples size in major expansion,” Euromaidan Press, November 12th, 2024. Many other units of this type exist, along with drone units embedded within infantry, artillery and other unit types. For an informal but nonetheless interesting comparative analysis see: Yurii Butusov, “Best UAV units – eliminators of occupiers, January ranking,” Censor.net, February 17, 2025. A number of other drone units, including Peaky Blinders, Predator Brigade and others, also routinely publish their video compilations online. Also see: Ivan Khomenko, “Ukraine Launches Point-Based Rewards for Drone Operators,” United24 Media, April 30, 2025.
145However, biases about drones have impacted statistics in other wars, most notably in the 2020 Nagorno-Karabakh War. For more on the biased influence of the Bayraktar corporation and Azerbaijani officials on the drone-casualty narrative in that war, see: Connable, Ground Combat, 2025, chapter 8; and Eado Hecht, “Drones in the NagornoKarabakh War: Analyzing the Data,” Military Strategy Magazine, Vol. 7, No. 4, 2022, pp. 31-37. Hecht shows the sharp differences between the official Azerbaijani statistics on drone strikes and data from the Oryx battle tracking website. For example, the government claimed 287 tanks were killed by Azeri drones, but Oryx was only able to confirm 101 drone kills of tanks. Overall, the government added about 40% more drone strikes than could be tracked, and 24% of the remaining possible strikes were not provable. Video evidence on Oryx showed with high confidence that a drone clearly had not caused the destruction in question in 39 cases, with 141 other cases being questionable out of a total of 1,267 items originally claimed. The Azeris may have had access to better data but if so I am not aware that they revealed it after Hecht’s article was published.
146Even this simple approach would be effectively impossible to apply. Some drone strikes clearly result in casualties and deaths: drones often record other drones at work and show exploding tanks, dismembered and charred corpses, and even soldiers finishing themselves off with their own rifles after being wounded by a drone. However, that compelling video data does not constitute evidence. We are shown videos of unsuccessful drone strikes far less frequently. It is not clear that all drone strikes cause casualties, and even some first-person-view drone strikes that directly hit individual soldiers may not lead to death. In one shocking pre-AI video case a Russian soldier head-butted an incoming attack drone, causing it to detonate on his helmet. He then ran away seemingly, but probably not unscathed. See Isabel van Brugen, “Russian Soldier Appears To Be Unscathed After Ukraine Drone Headbutt – Video,” Newsweek, August 15, 2024.
147For details on specific systems see the U.S. Army’s Worldwide Equipment Guide, or WEG.
148For details on specific systems see the WEG
149Estimates of total shells fired per month vary considerably by source. Aggregating from multiple sources, and without suggesting reliable accuracy, it might be safe to assume that in late 2025 Ukraine fires approximately 5,000 artillery shells per day while Russia fires between 10,000 and 20,000 per day. Heavier mortar shells from 81/82mm, 120mm, and (less so) up to 240mm are also consumed at a high rate but I found no reliable estimates. Note that some artillery shells are non-explosive (smoke, illumination, white phosphorous, mines, leaflet, etc.), and that some artillery shells are guided. Probably the most common round fired, however, is the standard explosive round carrying approximately (between, depending on type) eight to 24 pounds of high explosive between the common 122mm, 152mm and 155mm artillery systems. For more on artillery see, for example: David Axe, “New Guns, More Ammo: Ukraine’s Artillery Blasts Away At A Rate Of Millions of Shells A Year,” Forbes, February 28, 2025; Kateryna Denisova, “Ukrainian long-range strikes cut Russia’s shell fire rate by nearly half, Syrskyi says,” Kyiv Independent, April 9, 2025; Alex Vershinin, “The Return of Industrial Warfare,” Royal United Services Institute, June 17, 2022; Jack Watling and Nick Reynolds, Tactical Developments During the Third Year of the RussiaUkrainian War, Royal United Services Institute, February 2025; Open Source Centre and Royal United Services Institute, Ore to Ordnance: Disrupting Russia’s Artillery Supply Chains, undated. For details on specific systems see the WEG
Shell production on both sides has been challenging as has the production of artillery barrels, each of which can fire between 1,500 and 2,500 rounds before needing replacement.150
Anti-armor missiles: Guided anti-armor weapons like the FGM-148 Javelin antitank guided missile, ground-mounted TOW II, Stugna-P, Kornet and others are routinely used to damage or destroy adversary equipment and wound or kill personnel.151
Rocket and missile artillery: Both sides employ an array of ground-fired rocket and artillery systems including the BM-21 Grad, the M142 HIMARS, TOS-1-variant thermobaric warhead launcher and some longer-range systems like ATACMS, Neptune, and Flamingo. These weapons are capable of causing mass-casualty events.152
Mines: Probably more than two million anti-tank and anti-personnel mines have been laid in Ukraine, and all fixed defensive positions are buffered by extensive minefields.153 Furthermore, both sides drop mines in behind advancing forces with artillery-launched,
Emergency restoration works in Kyiv on July 31, 2025, at the end of a month in which Russia reputedly used over 5,100 glide bombs against Ukraine, more than 3,800 “shaheds” and nearly 260 missiles of various types – 128 of them ballistic.
drone-dropped and perhaps aerial dropped munitions.154
Aerial bombs, guns and rockets: Both sides routinely launch unguided rockets, bombs and shells, and also guided glide bombs against adversary front-line units. Some of these glide bombs exceed
3,000 pounds in overall weight and are capable of destroying industrial buildings. Statistics for Ukrainian aerial bomb use are unknown but the Russians produce over 40,000 per month and probably use approximately 5,000 per month against both military and civilian targets.155
Misadventure, murder, disease, etc.: Some percentage of casualties in every war derives from accidents, non-combat killing, underlying medical conditions or disease.156 Some unknown number of Russian soldiers dies each month by
150These estimates reflect howitzer/gun type and manufacturing standards but also the amount of propellant used per round. One way to count barrel use is to track aggregated propellant charges used rather than individual shells fired. Lower estimates would apply particularly those made to poorer standards in places like North Korea. I found no reliable field evidence on barrel wear data and replacement decision-making from either side in the Ukraine War. For more on artillery barrel production see, for example: David Axe, “Russia Needs Fresh Artillery Barrels, Bad. It’s Yanking Them Off Old Guns By The Thousand,” Forbes, January 12, 2024. This is a highly informative video about barrel production in the United States. “Why the US Has Only One Factory for Cannon Barrels,” Business Insider, YouTube.
151-152For details on specific systems see the WEG
153Estimates are difficult to verify and no accurate count would be possible. For more on this topic see, for example: Richard L. Garcia and Colin Colley, Russian Minefield Tactics Pose Challenge to Mobility, Red Diamond newsletter, U.S. Army TRADOC G-2, November 13, 2024; Mitzi Perdue, “An Explosive Choice: Landmines and Ukraine,” CEPA, November 22, 2024; “Ukraine Has Largest Minefield In The World, Prime Minister Says,” Radio Free Europe/Radio Liberty, January 8, 2023; “It is an elephant’: Ukraine’s unexploded mine problem,” UN News, United Nations, June 5, 2025; and see the Landmine and Cluster Munition Monitor
154Ukrainian forces can deploy Remote Anti-Armor Munitions, or RAAM, and the Area-Denial Artillery Munition, or ADAM, Sensor Fused Munition (SMArt), as well as a range of other deployable systems. Russia remotely deploys mines from a range of artillery and rocket systems. For more on remotely-deployed mines see: Jess Daly, “Ukraine’s New US Weapons That Strike Russian Troops from Below,” United24 Media, January 30, 2025; and Charles Bartles, “Russia Fields the ‘Zemlediliye’ Engineering System for Remote Mining,” U.S. Army TRADOC G-2, April 2, 2022. Probably one of the best resources on mines in Ukraine is: Bob Seddon and Nadine Lainer, Explosive Ordnance Guide for Ukraine, third edition, 2025.
155For details on specific systems see the WEG
156Data on disease rates in military units should be treated similarly to other casualty data: It is not practicably useful unless it is scientifically, reliably representative. Keep in mind that not all soldiers who suffer or even die from illness ever reach a medical facility. For some military forces, crashes and misadventures are recorded separately and may not show up in overall casualty data. Certainly murders and probably some suicides are recorded unevenly, though at this point we should recall the interesting 14.1 percent of examined Russian casualties that died from a single gunshot wound to the head (Kasmirov, et al., “Causes of Death,” 2024). Official U.S. casualty data show that 2,835 deaths in the Korean War (7.8% of total), 10,786 deaths in the Vietnam War (18.5%), 235 deaths in the Persian Gulf War (62%), 937 deaths in Operation Iraqi Freedom (21%) and 505 deaths in Operation Enduring Freedom (21%) resulted from non-battle causes including disease, heart failure, accidents, criminal activity and negligent behavior. Refer back to the Defense Casualty Analysis System (DCAS) for the official U.S. data. Given the relative competence of U.S. military forces compared to Russian military forces, it would at least be fair to assume that Russian non-battle deaths would be likely to exceed American non-battle deaths in war.
road accident, helicopter crash, slipping and falling or by negligently shooting themselves or others. Sometimes, drones force trucks off roads causing casualties without detonating. Some Russian soldiers are killed by their commanders or fellow soldiers. Some commit suicide. Others get sick and are treated as casualties whether or not they die from their illness.
Artillery is a prospective major confounding factor.157 Many estimates through mid-2024 assumed that artillery was causing between 70 and 80 percent of all casualties in the war.158 Those numbers amounted to loose statements, but for the sake of discussion only let us call them ‘broadly accepted,’ just as the 70-percent figure might be. Throughout that historical period (February 2022 to about mid-2024), Ukraine and Russia together may have fired anywhere from
“Some unknown number of Russian soldiers dies each month by road accident, helicopter crash, slipping and falling or by negligently shooting themselves.”
10,000 to more than 70,000 artillery shells per day.159 Let us assume that a rough average of 30,000 shells was fired per day, with wild variation by time period, from February 2022 through late 2024. Therefore, without taking any of this at face value: ~30,000 artillery shells per day
cause ~75 percent of all casualties, both sides.
Due to production and delivery shortages, Ukraine in 2025 now fires far fewer shells per day, so down to possibly 2,500 or less from as many as 10,000.160 According to the Ukrainian Armed Forces, as of April 2025 Russia was firing about 23,000 shells per day.161 So let’s go with a low combined estimate and assume that 22,000 artillery shells (2,000 Ukrainian + 20,000 Russian) are still fired per day as of mid- to late2025 with no appreciable decrease in pershot effectiveness.162 Therefore: ~22,000 artillery shells per day cause ~(<)30 percent of all casualties, both sides.
If drones now cause 70 percent or more of all casualties, then artillery must cause significantly less than 30 percent given all other casualty causing sources.163 If we take these generally accepted statistical
157Each of the other weapons listed here could receive a similar analysis. Probably the most eyebrow-raising gap from late-2024 through late-2025 is aerial munitions. Glide bombs alone almost certainly are causing significant casualties on the Ukrainian side and may be causing considerable Russian casualties as well. Glide bomb strikes are shown to wipe out entire trenches and command post buildings. However, individual casualties often are not visible and would therefore be harder to track. Infantry combat is probably more routine than most experts believe since infantry units probably do not report every engagement up the chain of command. And many missile strikes are unobserved so it would be difficult to track missile casualties. For more on glide bombs see, for example: Watling, et al., Tactical Developments, 2025, pp. 5, 7, 8, and others; “Ukraine says Russia dropped over 51,000 guided bombs since start of war,” Reuters, January 9, 2025; Michael Peck, “Breaking Russia’s Triple Chokehold,” CEPA, April 16, 2025; Michael Peck, “Glide Bombs: The Russian Wonder Weapon?,” CEPA, April 9, 2024.
158The 80-percent figure appears to have been popularized from one source but I did not find the original. RUSI’s 70-percent figure probably would be more widely accepted by experts based on the access and expertise of the authors. Note closely that the authors only refer to Ukrainian and not Russian casualties. “Ukraine’s frontline is ‘crumbling’ against Russian advances and Zelensky’s ‘misguided’ victory plan is partly to blame, says general,” RUSI, featured in The Daily Mail, October 31, 2024. Also see: Jack Watling, Oleksandr V. Danylyuk, and Nick Reynolds, Preliminary Lessons from Ukraine’s Offensive Operations, 2022-2023, Royal United Services Institute, July 2024; and Jack Watling and Nick Reynolds, Meatgrinder: Russian Tactics in the Second Year of Its Invasion of Ukraine, Royal United Services Institute, May 19, 2023, p. 11; Stephen Grey, John Shiffman, and Allison Martell, “Years of miscalculations by U.S., NATO led to dire shell shortage in Ukraine,” Reuters, July 19, 2024; Max Boot, “Weapons of War: The Race Between Russia and Ukraine,” Council on Foreign Relations, April 24, 2024; David Hambling, “How Is Ukraine Destroying So Much Russian Artillery,” Forbes, July 17, 2024; Mark F. Cancian and Chris H. Park, “Can South Korean 105-milimeter Ammunition Rescue Ukraine?,” Center for Strategic and International Studies, March 22, 2024; Jack Watling, “The Peril of Ukraine’s Ammo Shortage,” Time, February 19, 2024; Sam Cranny-Evans, “Russia’s Artillery War in Ukraine: Challenges and Innovations,” Royal United Services Institute, August 9, 2023; Bing West, “Has High Tech Made Artillery Obsolete?,” Strategika, August 30, 2024.
159Clearly this is a wide variation. Estimates are equally diverse. See, for example: Courtney Kube, “Russia and Ukraine are firing 24,000 or more artillery rounds a day,” NBC News, November 10, 2022; Michael K. Lima, “Munitions for Ukraine: Observations and Recommendations,” U.S. Army, April 23, 2024; Joe Saballa, “Ukraine Slowly Matching Rate of Russia’s Artillery Fires: Defense Official,” The Defence Post, October 2, 2024; Antti Ruokonen, “Ukraine’s Artillery Shell Shortfall,” Lawfare, April 3, 2024; and the various publications from RUSI cited here.
160Some estimates approach just 1,000 per day. For Ukrainian shell consumption see the cited reports by Jack Watling, et al., RUSI; the reports on artillery cited above; as well as, for example: Axe, “Artillery Blasts Away,” 2025; Patrick Tucker, “The West is underestimating Ukraine’s artillery needs,” Defense One, February 22, 2024; Katie Bo Lillis, Natasha Bertrand, Oren Liebermann, and Haley Britzky, “Exclusive: Russia producing three times more artillery shells than US and Europe for Ukraine,” CNN, March 11, 2024.
161Denisova, “Ukrainian long-range strikes,” 2025.
162Note that this does not include mortar shells. Mortar shell consumption has been tracked with much less vigor than artillery shell consumption and is therefore much more difficult to estimate.
163And of course those calculations are more extreme if we take at face value the newer 80-percent or 90-percent statements. At 90 percent drone casualty-causation rate, artillery would ostensibly cause fewer than 10 percent of all casualties, which would amount to more than 87-percent drop in the relative effectiveness of artillery, with all other factors remaining equal. If we take into account just prospective glide-bomb casualties, with over 5,000 glide bombs dropped on military and civilian targets combined per month, we go further into extreme territory. These statements may still of course be correct – we have not disproven them – but vague statements backed by no reliable data or evidence that also appear to be extremely unlikely deserve extra scrutiny.
statements at face value, that would be more than a 60-percent decrease from possibly 75 percent casualty causation in less than a year.164 At the same time average daily shellfire may have dropped by less than 27 percent, or about half as much.
So if casualty figures remained generally steady across that time, the supposed shift – 60-percent fewer artillery-caused casualties with less than 30-percent fewer shells fired – would be remarkable.165 Given the way that the drone statements from Ukraine are being absorbed in the West, some might be led by this apparently remarkable change to believe that the entire character or nature of war had revolutionized just between 2024 and 2025.
What does all this artillery estimation really tell us? Basically, it just reinforces the data problem and highlights more unknowns:
n Maybe all these rough estimates are junk, or maybe just one is way off from reality and skewing the whole calculation, or maybe data lag throws everything off.
n Maybe the roughly 75-percent artillery statistic statement was way off and drones or some other weapon were sneakily causing far more casualties than first believed.
n Maybe overall casualties have significantly increased, allowing vastly
Ukrainian border guards arm a UAV before a combat operation in August 2023.
Picture:
Petro
Zadorozhnyy
/ State Border Guard Service of Ukraine / the Collection of war.ukraine.ua
increased drone use to cause relatively more casualties without significantly decreasing artillery casualties.
n Maybe artillery has gotten worse as barrels wear down and accuracy decreases, or as drones force artillery units further from the front lines.
n Maybe drones do not cause 70 percent of all casualties.
n Or maybe any one of another number of factors is at play.
We don’t know the overall casualty figures and cannot track them over time. That means we have nothing stable to compare and contrast any of these numbers against. We have no reliable cause-of-wound, cause-of-death data. Absent comparative data, all these numbers – 75 percent
artillery, 30 percent artillery, 70 percent drones, 30,000 shells, 22,000 shells, etc. – all just become nearly valueless quantifications that confuse more than they help. At best, the percentages represent subjective expert judgments; I return to this point in the conclusion.
How do drones cause casualties?
Drones clearly do cause a significant number of casualties in Ukraine and – thus far without transparent, reliable evidence or proof – many experts agree they cause the most casualties in 2025. We needed to define casualty for our study, and we also need to know what it means for a drone to cause a casualty. Even this seemingly straightforward question is hard to answer in practice. Anyone who wants a better understanding of how drones cause casualties in the Ukraine War can review thousands of hours of video footage readily available online. Having already done that gruesome information collection, I offer up these observations.166
In a great many cases, drones clearly and almost unquestionably cause wounds or death in Ukraine.167 Probably the best example of this would be a first-personview (FPV) or grenade-drop drone strike against an individual soldier that is in turn recorded by a nearby observation drone. We can clearly see the attacking drone,
164Reported Ukrainian and Russian drone production over this period rose significantly. One estimate suggests Ukrainian production rose tenfold. However, not all drones reach the battlefield, not all drones are FPV or drop-drones, and not all drones that are launched reach their intended targets. A dated RUSI estimate suggests 60-80 percent of all attack drones launched fail to reach their targets due to environmental conditions, pilot error, electronic defenses or other factors. It is possible that as drone production has gone up, so has the rate at which drones are lost due to improving counter-UAS systems and tactics. Therefore, increased drone production should not necessarily be equated with increased drone munitions reaching targets. See: Mariam Halstian, “A First Point View: Examining Ukraine’s Drone Industry,” Georgetown Security Studies Review, May 15, 2025; Justin Bronk, “NATO Should Not Replace Traditional Firepower with ‘Drones,’” RUSI, August 4, 2025. The 60-80 percent loss figure is drawn from Watling and Reynolds, Tactical Developments, 2025.
165But even if accurate it still would not prove that drones were relatively more useful than artillery. Instead, it might show that drones have been used to fill in a major gap caused by shell shortages and artillery piece destruction on both sides. An approximately 50 percent drop in shells fired might lead to an equivalent drop in casualties caused by percentage (50 percent), which could have been backfilled by increasing drone use. So perhaps in at least relatively static modern wars fought on mostly open plains and with massed industrial support on both sides, artillery, drones and other types of fires might be interchangeable in some ways. Absent evidence, that all amounts to speculation.
166Both for my research in support of the development of the Ground Combat Database, and to keep abreast of the war in Ukraine, I have watched between one and two hours of Ukraine War combat footage, interview footage, documentary footage and related videos each day since late February 2022. Some days life intervenes and I watch less or none then catch up on following days. Other days I watch more in pursuit of a specific research agenda. Applying a low-end estimate that amounts to over 1,200 hours of Ukraine War video footage reviewed. My observations may therefore be interesting but remain anecdotal: Nothing I write based on my review of video evidence should be taken at face value or applied without question to any real-world problem. Video we are able to observe in the public domain is received passively from sources frequently intent on manipulating our perceptions through editorial selection, video editing and, probably with increasing frequency, AI video manipulation. I have viewed considerable primarysource video information, but it is also non-representative. Secondary- and tertiary-source information is drawn from Reddit Combat Forum and related sites, YouTube, Telegram and a range of other social media platforms.
167I write “almost unquestioningly” here because we have to stay committed to the idea that video can be manipulated and that in some cases it might be repeated. Some units put out drone-kill montages every day. While information is so prolific that cheating probably would be unnecessary, it would be easy to splice in old video with new on a slow day. Competition between units is very real and might (nothing to show this happens) motivate some disingenuous behavior. And more advanced video AI opens opportunities for outright manipulation or fabrication by any combatant in war.
the munition, the strike and the effects of the strike. In many cases the video evidence could with minimal doubt be taken as reliable proof of wounding or death. Many videos show Russian soldiers literally blown to pieces, ripped in half or decapitated, soldiers burning alive and transforming into charred husks or long, lingering shots of whole but immobile and clearly dead bodies.
In many other cases the mechanisms and results are less clear. Often, FPV drones hit targets but the strike is not recorded by another drone. Video feeds blinker out or freeze moments before prospective impact. In those cases we cannot claim to know that the munition on the drone detonated and that it also caused a casualty. Sometimes drones strike vehicles and detonate but may or may not cause casualties in the vehicle; casualties may too often be assumed. And in a great many cases, drones are used in concert with other weapons – use of combined arms –creating possibly the most interesting data complications.168
A notional example: Russian daylight armor-infantry assault
Let us take the notional example of a Russian frontal assault on a Ukrainian trench line. This is a daily occurrence across many parts of the 1,200-kilometer front. In this hypothetical case, Russian soldiers mount up in a mix of armored vehicles and drive in broad daylight down an open road directly towards the Ukrainian positions. They take this rather suicidal approach because the ground on both sides of the road is littered with mines and shell holes and is effectively non-trafficable. Their commanders have, in effect, chosen the slightly less suicidal option.
Two Ukrainian observation drones spot the attacking column. Video from the drones
is displayed in the brigade operations center. Brigade officers immediately order a response to stall the Russian attack. All at once they warn the infantry and nearby armor units, call up artillery and heavy mortar fire, and order front-line drone units to start launching defensive FPV and grenade-drop drones.
The first vehicle in the Russian column is a modified T-72 “turtle” tank encased in a massive add-on armored shell and mounting several electronic warfare devices built to stop incoming drones. Russian troops were told that their own drones had swept the road for mines and that the way would be clear. But the Ukrainians had slipped in a mine-laying unmanned ground vehicle that had planted TM-62 anti-tank mines at a critical bend in the road flanked by mud. Blinded by the add-on armor, the Russian tank driver puts his tread right over one of the mines, delinking a track and immobilizing his vehicle.
Now the rest of the column is in trouble: They have to either stop and then become
sitting ducks or try to drive around the lead tank into the mined fields, hoping for better fortune. The second vehicle in the column – another turtle tank – tries its luck and drives just to the side of the road in order to get by. It immediately hits another mine and becomes stuck in mud. Now the whole column is frozen. Soldiers can retreat, maybe survive, but face the wrath of their officers; push forward; or hunker down in no-man’s land.169
As the drivers and troops in the thinly armored infantry vehicles weigh their options, a mix of artillery and mortar fire starts to fall around the column. First, 82mm and 120mm mortars start to land in the fields nearby, creating chaos but no real damage to the Russian armor. Next, three 155mm artillery shells land closer in, guided by the observing Ukrainian drones. These disable two of the wheeled infantry-carrying armored vehicles, preventing them from moving and blocking the road to the rear.
A small Ukrainian team that had rushed forward at first warning of the Russian
168Combined-arms warfare probably is the most common type of military engagement on land. This is somewhat self-defining: Combined arms warfare is the use of more than one type of weapon and-or munition in order to fix and destroy enemy forces. Armies combine weapons like mortars and machine guns to create a dilemma for opposing troops. If they move in the open or attack, the machine guns can kill them with direct fire. But if they hide from the machine gun fire in defilade they are sitting ducks for the high-parabola mortar fire. Ideally, this dilemma creates uncertainty and freezing, which allows friendly troops to maneuver into position to seize their objectives. For more on combined arms see, for example, the Marine Corps Doctrinal Publication 1, Warfighting
169In some cases they might also face blocking units. Unproven anecdotal reports suggest some Russian units are applying a World War II-era technique: They put reliable contract soldiers in blocking units behind the advancing Storm units. These blocking soldiers have orders to force any retreating assaulters back to the front or to shoot them for retreating from combat. For more on this technique see Reese, Why Stalin’s Soldiers Fought, 2011, and, for example: David Axe, “‘They’re Just Choosing Where To Die.’ When Russia’s Poorly-Trained Stormtroopers Retreat, Russian Barrier Troops Gun Them Down,” Forbes, January 18, 2024.
Picture: Vadym Pliashechko / State Border Guard Service of Ukraine / the Collection of war.ukraine.ua
attack is able to put a Javelin anti-tank missile into the disabled lead tank. This rips off part of the turtle shell and shatters the electronic warfare devices on the tank, exposing it for follow-on attacks. Then the first FPV drone flies in and slams into the side of the second tank, ripping away part of its armor. More drones are on the way. A handful of soldiers from the two tanks abandon their vehicles and start to flee back down the road. Hapless Storm-Z troopers in the light armor decide to hunker down. Infantry jump out of their vehicles and take cover where they can: next to their disabled vehicles, in the road ditch or just in some tall grass nearby. Ukrainian mortar gunners zero in their fires, guided by the drone video relayed through the brigade operations center. Infantry casualties start to mount.
Now waves of FPV and grenade-drop drones swing into play. The two lead tanks are hammered by hit after hit. Each exploding shape-charge warhead rips away more of the turtle armor. The fifth drone through manages to hit the ammunition carousel on the first T-72, launching its turret sky-high in a spectacular explosion. However, the second tank is more stubborn, requiring 15 drones to set it on fire.
Drones hover and swarm over the Russian infantry who have been left on their own by their ruthless commanders and with no real options for survival. Mortar shells keep them pinned as drones cause wounds to legs and arms. A lone Ukrainian machine-gun position
velocity 7.62mm caliber bullets into their position, causing more casualties. Troops put tourniquets to gashed limbs and fire ineffectively at the drones. Then a Ukrainian M2 Bradley armored vehicle arrives. From more than a kilometer away the Bradley fires streams of 25mm explosive shells into the remaining troops, ending the assault. Drones chase the fleeing tankers back down the road and kill them.
I aggregated this notional scenario from hundreds of videos that show all or most of these actions from the Ukraine battlefield. Everything about this is quite real and commonplace. So, what happened here? How can we tell from this one engagement – or any real engagement like this one –which weapons caused which casualties?
In one extreme interpretation, drones indirectly or directly caused every single Russian casualty. Ukrainian defenders used drones to plant the mines that fixed the column in place, guided and adjusted their fires by drone, and ultimately wounded or killed some percentage of the Russians in the column with direct drone strikes. Other systems like artillery, mortars, mines, machine guns and cannons were just parts of the drone network.
In the other extreme, maybe drones only caused a handful of casualties. We don’t know if the mines or missiles wounded or killed anyone in the tanks, but they might
without the mines and artillery disabling the vehicles. Drones finished off some of the infantry and tankers but only after they were wounded by other weapons including a 1930s-era machine gun. And maybe the Bradley killed most of the Russians overall.
As with most complex things in life, some middle-ground answer with a bit of unsurety seems more reasonable. Mines created the condition that channelized the column that then enabled the drones to concentrate their fires. Drones worked together with the mines, artillery and missiles to destroy the vehicles. Drones worked together with the machine gun, artillery, mortars and Bradley to kill the infantry. Combined-arms warfare is commonplace, and this kind of fighting does not lend itself to causative analysis.
Piecing together quantifiable causation would require figuring out how each soldier was first wounded and then ultimately killed. Parsing those numbers would require information that could not be reasonably turned into data or evidence without examining every recorded engagement in the war or finding a reliable random sample. Since we do not know how many engagements have occurred, and since no two engagements are identical, even this extraordinary research effort probably would generate zero evidence.
Bottom line on Ukraine War
drone-casualty causation:
Subjective expert judgment
From a scientific perspective, our understanding of combat wound causation in the Ukraine War is no better than that of wound causation in the U.S. Civil War. Best guesses are offered. Some guesses are more respected than others. As I wrote above, these kinds of guesses are sometimes referred to as subjective expert judgments, or perhaps informed subjective judgments. In other words: A human being with extensive real-world expertise in a specific field or on a certain topic provides their best guess. There is real value to these kinds of subjective inputs. They can sometimes sway a court hearing or change the design of a skyscraper. Revered experts are particularly influential with senior military and political leaders and might be able to steer them towards an ideal outcome. In the best cases, subjective judgments are backed by transparent and reliable evidence: We want to both know what the expert thinks and be able to check their thinking against something tangible. In more difficult circumstances, experts provide their judgment absent reliable proof, evidence, data or even information.
Remember here that war is inherently uncertain. We have now gone through many of the challenges in just estimating one kind of wartime data. We know that cause-of-wound, cause-of-death estimates in Ukraine are not reliable. Still, experts are asked to provide their estimates. When they do, need for closure and cultural dispositions to quantify kick in. We wind up
“Approach information in war like a welltrained intelligence analyst. Intelligence analysis is tailored to support both military and policy decisionmaking under highpressure and often extraordinarily uncertain conditions.”
consuming statements about the Ukraine War and drones and then using and repeating them as evidence or even proof We can do better than this.
Conclusion: Act like an intelligence analyst
Can we extrapolate this examination of cause-of-wound, cause-of-death data out to other forms of wartime data, and then perhaps to our broader understanding of the practice of war? I argue here and have argued in previous works that we can do so with a high degree of confidence.170 In other words, we should have confidence in our lack of confidence: We should treat all knowledge in war as inherently uncertain. We should only place trust in information, data, and evidence when that trust is logically warranted.
This cautious approach to war knowledge is far more logical, intuitive and ultimately practical than the approaches we too often apply now: improper application of scientific methods; surface-level speed guessing; selecting the right answer by power of personality; lazy statistical interpolation; and then equally lazy and often manipulative extrapolation of unreliable junk measurements to broader phenomena.
How can we train and apply this mindset? Here I am clearly biased: Approach information in war like a well-trained intelligence analyst.171 Intelligence analysis is tailored to support both military and policy decision-making under highpressure and often extraordinarily uncertain conditions. Uncertainty is inherent in the intelligence discipline. In my interpretation and experience, good intelligence analysts consistently try to apply at least these basic principles:172
1. Start from an assumption of ignorance. Assume you know nothing at the outset.
2. Treat all information as uncertain. View information within a range of uncertainty.
3. Separate information from opinion. Treat information and opinions as distinct.
4. Explore all confounding factors. Examine and describe uncomfortable alternatives.
5. Be comfortable with not knowing. Acknowledge the limitations of knowledge.
6. Provide transparent sourcing. Always show your original sources in detail.
170For example: Connable, Embracing the Fog, 2012; Connable, Ground Combat, 2025; Connable, et al., Will to Fight, 2018; Ben Connable, “Learning from the VietnamEra Strategic Assessment Failure,” book chapter in: Andrew Williams, James Bexfield, Fabrizio Fitzgerald Farina, and Johannes de Nijs, eds., Innovation in Operations Assessment: Recent Developments in Measuring Results in Combat Effectiveness, Norfolk, Vir.: Allied Transformation Command, North Atlantic Treaty Organization, 2013; Ben Connable, Jason Campbell, Bryce Loidolt, and Gail Fisher, Assessing Freedom of Movement for Counterinsurgency Campaigns, Santa Monica, Calif.: RAND Corporation, 2012; Connable, How Insurgencies End, 2010; Ben Connable, “Russians Do Break: Historical and Cultural Context for a Prospective Ukrainian Victory,” War On The Rocks, September 25, 2024; and others.
171As a former intelligence analyst and researcher who studied the intelligence process I am biased towards the analytic method. Alternatives exist, including the scientific method. However, with the inherent uncertainties in war the scientific method is generally not applicable. For more on the misalignment between scientific methods and war analysis see Connable, Embracing the Fog, 2012.
172I reviewed a number of intelligence analysis guidelines and tradecraft documents as I prepared this list. I was surprised to find that most focused more on the relationship between the analyst and consumer and relatively less on the mindset necessary to conduct effective analysis. A few good guides exist and perhaps the most useful professional guidelines are the standing Intelligence Community Directives 203 and 206, both of which set high standards for analytic integrity, objectivity, and transparency. Also see: Robert Levine, “Principles of Intelligence Analysis,” Studies in Intelligence, Vol. 65, No. 4, December 2021, extracts; Bill Studeman, “10 Philosophies/Principles of Intelligence,” briefing, September 2013; New Zealand Institute of Intelligence Professionals, Intelligence Practitioner’s Handbook, August, 2023; and U.S. Marine Corps, Intelligence, MCDP 2, Washington, D.C.: Headquarters, United States Marine Corps, 1997.
7. Attach caveats to all findings. Keep the uncertainties closely tied to the claims. 8. Identify gaps and reattack. Continue to press for needed, useful knowledge. 9. Periodically revisit assumptions. Force hard checks on standing beliefs.
Most academic and military professionals are told at some point in their careers to check their biases and to be as objective as possible. But pure objectivity is unattainable.173 We all have some inescapable biases. In practice, very few people are capable of clearly separating fact from opinion, fewer still are comfortable with not knowing, and probably even fewer are comfortable expressing their lack of knowledge in a high-stakes setting. Therefore, it is best to treat these principles more as ideals. They will remain perfectly and permanently out of reach. It is the process of continually reaching for the ideals, falling short and reaching back up that builds the mindset necessary to improve. I have cited a range of useful readings here and I offer up a long-standing intelligence model for everyday war-study use.
Sherman Kent’s classic Words of Estimative Probability is probably the best resource to help understand the mindset I recommend.174 Kent is the godfather of the U.S. Central Intelligence Agency’s analytic division and probably the agency’s most important author on intelligence theory; the CIA’s analytic school was named after Kent.175 In this article he makes many of the same fundamental arguments I propose, focusing on uncertainty. In Table 7, above, Kent assigns a range of uncertainty percentages to words of estimative probability. None of this has any fixed meaning. It is instead a useful guide that anyone can rewrite for their own purposes using different percentage ranges and other words including, but not limited to: likely, unlikely, improbable, conceivable, may, etc.
Ideally, the words “certain” and “impossible” would never be used.
Table 7: Sherman Kent’s Words of Estimative Probability Table
Source: Kent, “Words of Estimative Probability,” 1964, p. 55.
Refer back to Edmans and the (rightly) extraordinarily high standard for proof, particularly for complex and largescale phenomena. Both certainty and impossibility require proof. In practice nearly all information, data and even evidence can be qualified without harming its value. In fact, reasonable qualifications like the ones Kent proposed might have helped prevent the United States and its allies from basing decision-making for the Iraq War at least in part on purported “facts” delivered by Colin Powell that turned out to be untrue statements.
Applying these principles, what can we say about the causes of casualties in the Ukraine War? I propose a more suitable statement than the early-2025 statement ostensibly linked to 2024 data, ‘drones cause 70 percent of all casualties in Ukraine’. All the original sources are listed above. With those source transparently in place, we can assess and opine:
Source assessment: One unsubstantiated subject-matter expert statement ostensibly based on calendar-year 2024 data from the Ukrainian Ministry of Defense suggests drones caused approximately 70 percent of all casualties in the Ukraine War during that calendar year. We do not have access to that data and therefore cannot verify its existence or veracity. In early 2025, several unnamed Ukrainian military officers reportedly independently stated or just
concurred with this 70-percent estimate. Because these human sources remain unidentified, neither their existence nor the reliability of their estimates can be accounted for. I found no contemporaneous independent sources that might reinforce the original estimate. It is highly likely that the Ukrainians possess more finely detailed multi-source combat reporting and field intelligence data; more below.
Analysis: Since the publication of the original estimate by The New York Times, circular reporting appears to have first solidified the original 70-percent estimate and then provided basis for new estimates that attribute 80-percent or 90-percent of all casualties in the Ukraine War to drones (i.e., it feels like more than before so it must be higher than 70, therefore 80 or 90). Note that the original estimate appears to have been drawn from Ukrainian military estimates of only Russian casualties; unsubstantiated extrapolation to all casualties may have occurred passively as circular reporting compounded online. Based on my cited assessment of previous efforts to track and attribute weapons and munitions to wounds and deaths in war, this kind of accounting appears to be infeasible. A range of information biases seem to be at work here, including expert bias, information volume imbalance generated by daily mass production of drone videos and political bias. Extensive information gaps do not support a clear judgment. Even with notably
173For more on this see the works of Sherman Kent, cited below, as well as: Richards J. Heuer, Jr., Psychology of Intelligence Analysis, Langley, Vir.: Center for the Study of Intelligence, U.S. Central Intelligence Agency, 1999.
174Sherman Kent, “Words of Estimative Probability,” second edition, Studies in Intelligence, Vol. 8, No. 4, 1964, pp. 49-65.
175For more on Kent see, for example: Jack Davis, “Sherman Kent and the Profession of Intelligence Analysis,” occasional paper, Vol. 1, No. 5, 2002.
more reliable and numerous classified data, estimates would still be non-random and therefore non-representative.
Bottom line: This is an unsubstantiated claim based on information that cannot be confirmed in the public domain, that logically cannot be representative and that should not be used for further analysis or decision support. However, we assess that drones probably cause relatively (but probably unevenly) more casualties on both sides in 2025 than they did in 2024 and preceding years in the current phase of the Ukraine War [sources cited]. This is due to a range of factors that may not be generalizable to other conflicts.
Gaps: Our primary information gap is lack of reliable and representative causeof-wound, cause-of-death data from the Ukraine battlefield. I do not believe this data can reasonably be obtained even with extraordinary resource expenditure.
“Do
we need to show a statistic in order to convince people that drones are important in modern warfare? Given the extraordinary global enthusiasm for drone technology as of late 2025, the answer to that question is no. We do, however, need to help military leaders, policymakers and other key decision-makers understand the practical limitations of knowledge in war.”
What can we do about this problem? First let us ask: Do we really need to know the answer to the question “which weapon causes the most casualties in the Ukraine War?” in order to best support the Ukrainian war against illegal Russian occupation? Having assessed a range of historical and ongoing wars, I do not believe this kind of statistical analysis is necessary or helpful. Ukrainian officers can clearly explain why
avoid the use of misleading measurements.
without having to resort to statistics. We can
Moving beyond Ukraine, do we need to know which weapon causes the most casualties in every observable war in order to understand changes in modern wars, and in doing so better forecast future wars? If it were possible to track this kind of data it might or might not be useful; that would be the subject of another paper. However, even if accurate it could prove to be misleading. As I show in the Ground Combat Database (GCD-V1), wars share some common characteristics but typically not all; Ukraine is not Gaza, is not Ethiopia, is not Iraq, is not Sudan, etc. Globalized generalizations are, unfortunately, unsubstantiated. Do we need to show a statistic in order to convince people that drones are important in modern warfare? Given the extraordinary global enthusiasm for drone technology as of late 2025, the answer to that question is no. We do, however, need to help military leaders, policymakers and other key decision-makers understand the practical limitations of knowledge in war. And we do need to help them identify the kind of unsubstantiated hyperbole surrounding advanced technologies like modern drones so they can make wise investments.
Anyone choosing to study modern war should ensure the questions they are asking or seeking to answer are necessary and helpful before they get to the point of asking, “is it answerable?”. If the answer to that question is no – it cannot be reasonably answered with information on hand or that might be collected – then the hard work of identifying gaps, building qualified and transparent estimates, and helping to condition leaders for inherent uncertainty follows.
REFERENCES
“«Понимал
Lenta.ru, October 14, 2024.
“It is an elephant’: Ukraine’s unexploded mine problem,” UN News, United Nations, June 5, 2025.
“New Evidence on North Korean War Losses,” Wilson Center, August 1, 2001. As of August 21, 2025: https://www. wilsoncenter.org/article/new-evidencenorth-korean-war-losses
“Pacification’s Deadly Price,” Newsweek, June 19, 1972, pp. 42-43; and Gunter Lewy, America in Vietnam, New York: Oxford University Press, 1978, p. 81.
“Putin’s recruits: The young Africans fighting for Russia in Ukraine,” France 24, June 16, 2025. Miłosz Bartosiewicz and Piotr Żochowski, “Putin’s foreign legion: Foreigners fighting in the war with Ukraine,” OSW, February 20, 2024.
“Richmond in Flames and Rubble,” American Battlefield Trust, October 9, 2023.
“Roman Kostenko: Ukrainian drones responsible for 65% of Russian casualties and 75% of destroyed equipment,” Odessa Journal, April 8, 2025.
“Russian losses in the war with Ukraine. Mediazona count, updated,” Mediazona. As of August 26, 2025: https://en.zona. media/article/2025/08/01/casualties_eng-trl
“Russian Troops’ Most Common Cause of Death Revealed in Report,” Newsweek, October 17, 2024. As of August 24, 2025: https://www.newsweek.com/russiaukraine-war-explosions-1970605
“Sanitätsbericht über das Deutsche Heer (Deutsches Feld-und Besatzungsheer) im Weltkriege 1914/1918: III. Band, Die Krankenbewegung bei dem Deutschen Feld-und Besatzungsheer, Berlin 1934, German Ministry of War, archive, created in
1934, Federal Archives, BArch PH 2/1998
“Ukraine Has Largest Minefield In The World, Prime Minister Says,” Radio Free Europe/Radio Liberty, January 8, 2023.
“Ukraine says Russia dropped over 51,000 guided bombs since start of war,” Reuters, January 9, 2025.
“Why the US Has Only One Factory for Cannon Barrels,” Business Insider, YouTube. As of August 27, 2025: https:// www.youtube.com/watch?v=xWOCxtfN4js
“Осколки в руках и ногах. Какие ранения и травмы чаще
Украине,” Vertska, June 25, 2024. As of August 23, 2025: https://verstka. media/bolee-5-procentov-potencialnospasaemyh-voennyh-umirayut-na-poleboya-ne-poluchiv-medpomosch
“Понад 430 000 солдатів, майже 9 000 ББМ, 3 700 танків – найбільші втрати російської армії в 2024 році,” Ukrainian Ministry of Defense website, January 3, 2025. As of July 23, 2025: https://mod. gov.ua/news/ponad-430-000-soldativmajzhe-9-000-bbm-3-700-tankiv-najbilshivtrati-rosijskoyi-armiyi-v-2024-roczi
“Потери России во время спецоперации составили 498 человек. Украиныболее 2,8 тыс.,” Tass.ru, March 2, 2022.
“ПОТЕРИ УКРАИНЫ ЗА ВРЕМЯ СПЕЦИАЛЬНОЙ ВОЕННОЙ ОПЕРАЦИИ (СВО),” Mskvremya.ru, August 26, 2025.
“Шойгу: потери ВСУ с начала спецоперации составили полмиллиона военных,” Vedomosti.ru, April 23, 2024.
Abramov, Nikita, “Песков
Lenta.ru, April 17, 2024. Adams, Paul, “Kill Russian soldiers, win points: Is Ukraine’s new drone scheme gamifying war?,” BBC, mid-July, 2025.
American Battlefield Trust, Civil War
Casualties, The Cost of War: Killed, Wounded, Captured, and Missing, updated as of September 15, 2023. As of August 14, 2025: https://www.battlefields.org/ learn/articles/civil-war-casualties
Anderson, Benjamin O., Nino Berdzuli, Andrew Ilbawi, Devora Kestel, Hans P. Kluge, and Rudiger Krech, “Health and cancer risks associated with low levels of alcohol consumption,” The Lancet: Public Health, Vol. 8, No. 1, January 2023, pp. 6-7.
Andrejsons, Kristaps, “Russia’s ConvictSoldiers Have Their Own Brutal Rules,” FP, May 16, 2023.
Antal, John F., 7 Seconds to Die: A Military Analysis of the Second Nagorno-Karabakh War and the Future of Warfighting, Havertown, Penn.: Casemate Publishers, 2022.
Axe, David, “Russia Needs Fresh Artillery Barrels, Bad. It’s Yanking Them Off Old Guns By The Thousand,” Forbes, January 12, 2024.
_____, “ ‘They’re Just Choosing Where To Die.’ When Russia’s Poorly-Trained Stormtroopers Retreat, Russian Barrier Troops Gun Them Down,” Forbes, January 18, 2024.
_____, “Artillery Blasts Away At A Rate Of Millions of Shells A Year,” Forbes, February 28, 2025.
_____, “New Guns, More Ammo: Ukraine’s Artillery Blasts Away At A Rate Of Millions of Shells A Year,” Forbes, February 28, 2025.
Ayres, Leonard P., The War With Germany: A Statistical Summary, Washington, D.C.: Government Printing Office, 1919.
Bacon, Edward, “Soviet Military Losses in World War II,” The Journal of Slavic Military Studies, Vol. 6, No. 4, 1993, pp. 616-633.
Bartles, Charles, “Russia Fields the ‘Zemlediliye’ Engineering System for
Remote Mining,” U.S. Army TRADOC G-2, April 2, 2022.
Bartosiewicz, Miłosz, and Piotr Żochowski, “Putin’s foreign legion: Foreigners fighting in the war with Ukraine,” OSW, February 20, 2024.
Beebe, Gilbert W., and Michael E. DeBakey, Battle casualties: Incidence, mortality, and logistic considerations, Springfield, Ill.: Thomas, 1952.
Bellamy, Ronald F., and Russ Zajtchuk, “Assessing the Effectiveness of Conventional Weapons,” book chapter in: Ronald F. Bellamy and Russ Zajtchuk, eds., Conventional Warfare: Ballistic, Blast, and Burn Injuries, Washington, D.C.: Center of Excellence in Military Medical Research and Education, 1989.
Ber, Jakub, “From Popasna to Bakhmut. The Wagner Group in the Russia-Ukraine War,” OSW, April 28, 2023; “The price of Bakhmut. We reveal the staggering toll of Russia’s bloodiest battle since WW2 and Wagner’s inmates recruited to fight it,” Mediazona, June 10, 2024. As of August 26, 2025: https://en.zona.media/ article/2024/06/10/wagner-trl
Berman, Eli, Joseph H. Felter, and Jacob N. Shapiro, Small Wars, Big Data: The Information Revolution in Modern Conflict, Princeton, N.J.: Princeton University Press, 2018.
Blanken, Leo J., Hy Rothstein, and Jason J. Lepore, Assessing War: The Challenge of Measuring Success and Failure, Washington, D.C.: Georgetown University Press, 2015.
Blanton, DeAnne, “Confederate Medical Personnel,” Prologue Magazine, Vol. 26, No.1, Spring 1994, no pages given.
Bodart, Gaston, Losses of Live in Modern Wars, Oxford, UK: Humphrey Milford, 1916.
Boot, Max, “Weapons of War: The Race Between Russia and Ukraine,” Council on Foreign Relations, April 24, 2024.
Bruckner, Rasmus, Hauke R. Heekeren, Matthew R. Nassar, “Understanding
learning through uncertainty and bias,” Communications Psychology, Vol. 3, No. 24, 2025.
Butusov, Yurii, “Best UAV units –eliminators of occupiers, January ranking,” Censor.net, February 17, 2025. As of August 26, 2025: https://censor.net/en/ resonance/3536303/the-rating-of-uav-units Callahan, Alice, “How Red Wine Lost Its Health Halo,” The New York Times, February 17, 2024.
Cancian, Mark F., and Chris H. Park, “Can South Korean 105-milimeter Ammunition Rescue Ukraine?,” Center for Strategic and International Studies, March 22, 2024.
Champion, Howard R., and Mary M. Lawnick, Coding the WDMET Database, Fort Detrick, Md.: U.S. Army Medical Research and Materiel Command, 2006.
Chulov, Martin, and Helen Pidd, “Defector admits to WMD lies that triggered Iraq war,” The Guardian, February 15, 2011.
Connable, Ben, Embracing the Fog of War, Santa Monica, Calif.: RAND Corporation, 2012.
_____, Iraqi Army Will to Fight: A Willto-Fight Case Study with Lessons for Western Security Force Assistance, Santa Monica, Calif.: RAND Corporation, 2022, p. 40 and others.
_____, “Russians Do Break: Historical and Cultural Context for a Prospective Ukrainian Victory,” War On The Rocks, September 25, 2024.
_____, Ground Combat: Puncturing the Myths of Modern War, Washington, D.C.: Georgetown University Press, 2025.
Connable, Ben, and Martin Libicki, How Insurgencies End, Santa Monica, Calif.: RAND Corporation, 2010.
Connable, Ben, Jason Campbell, Bryce Loidolt, and Gail Fisher, Assessing Freedom of Movement for Counterinsurgency Campaigns, Santa Monica, Calif.: RAND Corporation, 2012.
Connable, Ben, “Learning from the
Vietnam-Era Strategic Assessment Failure,” book chapter in: Andrew Williams, James Bexfield, Fabrizio Fitzgerald Farina, and Johannes de Nijs, eds., Innovation in Operations Assessment: Recent Developments in Measuring Results in Combat Effectiveness, Norfolk, Vir.: Allied Transformation Command, North Atlantic Treaty Organization, 2013.
Connable, Ben, Michael J. McNerney, William Marcellino, Aaron Frank, Henry Hargrove, Marek N. Posard, S. Rebecca Zimmerman, Natasha Lander, Jasen J. Castillo, and James Sladden, Will to Fight: Analyzing, Modeling, and Simulating the Will to Fight of Military Units, Santa Monica, Calif.: RAND Corporation, 2018.
Connable, Ben, Stephanie Young, Stephanie Pezard, Andrew Radin, Raphael S. Cohen, Katya Migacheva, and James Sladden, Russia’s Hostile Measures: Combating Russian Gray Zone Aggression Against NATO in the Contact, Blunt, and Surge Layers of Competition,” Santa Monica, Calif.: RAND Corporation, 2020.
Cordesman, Anthony H., and Abraham R. Wagner, The Lessons of Modern War, Volume II: The Iran-Iraq War, Boulder, Colo.: Westview Press, 1990.
Correll, John T., “Casualties,” Air Force Magazine, June 2023, pp. 48-53.
Cranny-Evans, Sam, “Russia’s Artillery War in Ukraine: Challenges and Innovations,” Royal United Services Institute, August 9, 2023.
Crosby, Alfred W., The Measure of Reality: Quantification and Western Society, 1250-1600, Cambridge, UK: Cambridge University Press, 1996.
Daly, Jess, “Ukraine’s New US Weapons That Strike Russian Troops from Below,” United24 Media, January 30, 2025.
Danilov, Vladimir, “Сводка
Pulse19, June 21, 2025.
Davis, Jack, “Sherman Kent and the Profession of Intelligence Analysis,” occasional paper, Vol. 1, No. 5, 2002. As
of August 29, 2025: https://www.cia.gov/ resources/csi/static/Kent-Profession-IntelAnalysis.pdf
DeBruyne, Nese F., and Anne Leland, American War and Military Operations Casualties: Lists and Statistics, Washington, D.C.: Congressional Research Service, January 2, 2025.
Denisova, Kateryna, “Ukrainian long-range strikes cut Russia’s shell fire rate by nearly half, Syrskyi says,” Kyiv Independent, April 9, 2025. As of August 28, 2025: https:// kyivindependent.com/ukraines-deepstrikes-cut-russias-ammunition-use-byhalf-military-chief-says/
Duarte, Pablo, Marcel Freidinger, and Andreas Hoffmann, “Military Casualties and Exchange Rates During the First World War: Did the Eastern Front Matter?,” The Economic History Review, Vol. 72, No. 4, November 2019, pp. 1312-1334.
Duelfer, Charles, Comprehensive Report of the Special Advisor to the DCI on Iraq’s WMD, Langley, Vir.: Central Intelligence Agency, 2004.
Dupuy, Trevor N., The Evolution of Weapons and Warfare, Fairfax, Vir.: Hero Books, 1984, p. 171.
Dupuy, Trevor N., John R. Brinkerhoff, C. Curtiss Johnson, and Peter J. Clark, Handbook on Ground Forces Attrition in Modern Warfare, Historical Evaluation & Research Organization, Fairfax, Vir., September 1986. As of August 20, 2025: https://apps.dtic.mil/sti/tr/pdf/ ADA278728.pdf
Dzutsati, Valery, “Russian Casualties in Ukraine Continue to Rise,” Eurasia Daily Monitor, July 16, 2024.
Eckhardt, William, “War-related Deaths Since 3000 BC,” Bulletin of Peace Proposals, Vol. 22, No. 4, December 1991, pp. 437-443.
Edmans, Alex, May Contain Lies: How Stories, Statistics, and Studies Exploit Our Biases—And What We Can Do About It, Oakland, Calif.: University of California Press, 2024.
Ellis, Frank, The Stalingrad Cauldron: Inside the Encirclement and Destruction of the 6th Army, Lawrence, Kan.: University Press of Kansas, 2013.
Eskridge, Susan L., Caroline A. Macera, Michael G. Galarneau, Troy L. Holbrook, Susan I. Woodruff, Andrew J. MacGregor, Deborah J. Morton, and Richard A. Shaffer, “Injuries from combat explosions in Iraq: Injury type, location, and severity,” Injury, Vol. 43, 2012, pp. 1678-1682.
Evans, Nathan J., Babette Rae, Maxim Bushmakin, Mark Rubin, and Scott D. Brown, “Need for closure is associated with urgency in perceptual decisionmaking,” Memory and Cognition, No. 45, 2017, pp. 1193-1205.
Falk, Dean, and Charles Hildebolt, “Annual War Deaths in Small-Scale versus State Societies Scale with Population Size Rather than Violence,” Current Anthropology, Vol. 58, No. 6, December 2017, pp. 805-813.
Flotow, M., “J. David Hacker’s ‘A CensusBased Count of the Civil War Dead,’” The Journal of the Abraham Lincoln Association, Vol. 42, No. 2, 2022.
Forczyk, Robert, Stalingrad 1942-43 (3): Catastrophe: the Death of 6th Army, Oxford, UK: Osprey, 2022.
Garcia, Richard L., and Colin Colley, Russian Minefield Tactics Pose Challenge to Mobility, Red Diamond newsletter, U.S. Army TRADOC G-2, November 13, 2024.
Garrick, James G., Naval Battle Casualty Study, Springfield, Vir.: National Technical Information Service, December 1973.
Gartner, Scott Sigmund, and Marissa Edson Myers, “Body Counts and ‘Success’ in the Vietnam and Korean Wars,” Journal of Interdisciplinary History, Vol. 25, No. 3, Winter 1995, pp. 337-395.
Gates, A.I., “The Unreliability of M.A. and I.Q. Based on Group Tests of General Ability,” Journal of Applied Psychology, Vol. 7, No. 1, 1923, pp. 93-100.
Gauvin, Ray, and Darrell Lane, “Seeking to Protect The Force: Experiences With The
Wound Data and Munitions Effectiveness Team,” The West Point Center for Oral History, April 24, 2023. As of August 28, 2025: https://www.westpointcoh.org/ interviews/seeking-to-protect-the-forceexperiences-with-the-wound-data-andmunitions-effectiveness-team
Gay, Victor, and Pauline Grosjean, “Morts Pour la France: A Database of French Fatalities of the Great War,” Explorations in Economic History, Vol. 90, October 2023.
German, J.B., and R.L. Walzem, “The Health Benefits of Wine,” Annual Review of Nutrition, No. 20, 2000, pp. 561-593.
Gibler, Douglas M., and Steven V. Miller, “The Militarized Interstate Events (MIE) dataset, 1816-2014,” Conflict Management and Peace Science, Vol. 41, No. 4, pp. 463-481.
Glantz, David M., and Jonathan M. House, Endgame at Stalingrad, Book Two: December 1942-February 1943, the Stalingrad Trilogy, Volume 3, Lawrence, Kan.: University Press of Kansas, 2014, pp. 600-601.
Gleditsch, Nils Petter, Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg, and Havard Strand, “Armed Conflict 1946-2001: A New Dataset,” Journal of Peace Research, Vol. 39, No. 5., 2002.
As of August 21, 2021, data are available for download at: https://ucdp.uu.se/ downloads/index.html#armedconflict
Goldhamer, Herbert, The Soviet Soldier: Soviet Military Management at the Troop Level, Santa Monica, Calif.: RAND Corporation, September 1975.
Grey, Stephen, John Shiffman, and Allison Martell, “Years of miscalculations by U.S., NATO led to dire shell shortage in Ukraine,” Reuters, July 19, 2024.
Gronroos, Noelle N., James M. Zouris, and Amber L. Wade, “Odds of Hospitalization Among Marine Corps Personnel by Military Occupational Specialty and Causative Agents During OEF and OIF,” Military Medicine, Vol. 174, No. 7, 2009, pp. 715-720.
Hacker, J. David, “A Census-Based Count
of the Civil War Dead,” Civil War History, Vol. 57, No. 4, December 2011, pp. 307348.
Hambling, David, “How Is Ukraine Destroying So Much Russian Artillery,” Forbes, July 17, 2024.
_____, “Ukraine’s Elite Drone Unite Is A Lethal, Fast Growing Tech Start Up,” Forbes, April 14, 2025.
Hambling, David, “Why Is Russia Sending Assault Troopers Into Battle On Crutches?,” Forbes, May 06, 2025.
Hecht, Eado, “Drones in the NagornoKarabakh War: Analyzing the Data,” Military Strategy Magazine, Vol. 7, No. 4, 2022, pp. 31-37.
Heitz, Richard P., “The Speed-Accuracy Tradeoff: History, Physiology, Methodology, and Behavior,” Frontiers in Neuroscience, Vol. 8, June 2014, pp. 1-19.
Heuer, Richards J., Jr., Psychology of Intelligence Analysis, Langley, Vir.: Center for the Study of Intelligence, U.S. Central Intelligence Agency, 1999. As of August 29, 2025: https://www.ialeia.org/docs/ Psychology_of_Intelligence_Analysis.pdf
Hiro, Dilip, The Longest War: The IranIraq Military Conflict, London, UK: Grafton Books, 1989.
Hirschfeld, Gerhard, Gerd Krumeich, and Irina Renz, eds., Enzyklopädie Erster Weltkrieg: Aktualisierte und erweiterte Studienausgabe, Paderborn, Germany: Ferdinand Schöningh, 2014, pp. 664-665.
Hoencamp, Rigo, Eric Vermetten, Edward C.T.H. Tan, Hein Putter, Luke P.H. Leenen, and Jaap F. Hamming, “Systematic review of the prevalence and characteristics of battle casualties from NATO coalition forces in Iraq and Afghanistan,” Injury, Vol. 45, 2014, pp. 1028-1034.
Holcomb, John B., Lynn G. Stansbury, Howard R. Champion, Charles Wade, and Ronald F. Bellamy, “Understanding Combat Casualty Care Statistics,” The Journal of Trauma, Injury, Infection, and Critical Care, Vol. 60, No. 2, 2005, pp. 397-401.
Infantry in Battle, second edition, Richmond, Vir.: Garrett & Massie, 1939.
Jdanov, Dmitri A., Dana A. Glei, and Domantas Jasilionis, “A modeling approach for estimating total mortality for Italy during the First and Second World Wars,” Genus, Vol. 66, No. 1, 2010, pp. 17-36.
Johnson, Rob, The Iran-Iraq War, London, UK: Palgrave-Macmillan, 2011.
Kahneman, Daniel, Thinking, Fast and Slow, New York, NY: Farrar, Straus, and Giroux, 2011.
Kasmirov, R.R., I.M. Samokhvalov, A.A. Zavrazhnov, V.V. Kudryashov, S.A. Kovalenko, and I.A. Tolmachev, “Causes of Death of Russian Servicemen in Modern Warfare,” Military Medical Journal (or Voenno-medicinskij žurnal), Vol. 345, No. 8, 2024, pp. 11-16.
Kelley, Robert E., “Twenty years ago in Iraq, ignoring the expert weapons inspectors proved to be a fatal mistake,” Stockholm International Peace Research Institute, March 9, 2023.
Kent, Sherman, “Words of Estimative Probability,” second edition, Studies in Intelligence, Vol. 8, No. 4, 1964, pp. 49-65. As of August 29, 2025: https:// www.cia.gov/readingroom/docs/CIARDP93T01132R000100020036-3.pdf
Khomenko, Ivan, “Ukraine Launches PointBased Rewards for Drone Operators,” United24 Media, April 30, 2025.
Kirichenko, David, “Ukraine’s drone wall is Europe’s first line of defense against Russia,” Atlantic Council, July 2, 2025.
_____, “Madyar: Ukraine’s Birdman Hunting Worms From the Sky,” Kyiv Post, August 11, 2025.
Knight, Ben, “Global conflicts: Death toll at highest in 21st Century,” DW, June 28, 2023.
Kobak, Dmitry, Alexey Bessudnov, Alexander Ershov, Tatiana Mikhailova, Alexey Rakhsha, “War Fatalities in Russia 2022-2023 Estimated Via Excess Male
Mortality: A Research Note,” Demography, Vol. 62, No. 2, April 1, 2025, pp. 335-347.
Kofman, Michael, podcast with Amos Fox, “Michael Kofman on the Russo-Ukraine War and the State of Military Thought,” Revolution in Military Affairs podcast, August 4, 2025.
Kohanets, Roman, “Russia’s War Creates ‘Caste System’ Valuing Soldiers Unequally by Region,” United24 Media, August 15, 2025.
Kosoy, Daniel, “Russia Is Sending Its Wounded Soldiers Into Meat Grinder Assaults, Even on Crutches,” United24 Media, February 17, 2025.
Kostenko, Roman, interview with Ihor Dolgov, Ukrinform, undated, posted on April 7, 2025. As of July 21, 2025: https://www.youtube.com/ watch?v=tzfwmQZW8Ag.
Kramer, Andrew E., “Amid Uncertainty About U.S. Support, Ukraine Pins Its Hopes on Innovation,” The New York Times, April 28, 2025.
Kruglanski, Arie W., and Donna M. Webster, “Motivated Closing of the Mind: ‘Seizing’ and ‘Freezing,’” Psychological Review, Vol. 103, No. 2, 1996, pp. 263283.
Kube, Courtney, “Russia and Ukraine are firing 24,000 or more artillery rounds a day,” NBC News, November 10, 2022.
Lacina, Bethany, and Nils Petter Gleditsch, “Monitoring Trends in Global Combat: A New Dataset of Battle Deaths,” European Journal of Population, Vol. 21, No. 2/3, June 2005, pp. 145-166.
Lada, John, and Frank A. Reister, eds., Medical Statistics in World War II, Office of the Surgeon General, Washington, D.C.: Department of the Army, 1975. Laurent, Helena, “Ethiopia: The daunting task of reporting the Tigray conflict,” DW, January 27, 2022. Also see the work of the Tigray War Project at Ghent University and Every Casualty Counts, with some data primarily on civilian casualties available at, as of August 21,
2025: https://www.ethiopiatigraywar.com/ incidents.php
Lawrence, Christopher, War By Numbers: Understanding Conventional Combat, Lincoln, Neb.: Potomac Books, 2017.
Lawry, Lynn Lieberman, Amandari Kanagaratnam, Ashleigh Roberds, Jessica Korona-Bailey, Luke Juman, Miranda Janvrin, Zoe Amowitz, Tiffany E. Hamm, John Maddox, Oleh Berezyuk, and Tracey Perez Koehlmoos, “A qualitative assessment of disease and non-battle injuries in Ukraine since the Russian invasion,” Conflict and Health, Vol. 19, No. 19, 2025.
Ledford, Frank F., Jr., “Medical Support for Operation Desert Storm,” The Journal of the U.S. Army Medical Department, January-February 1992, pp. 3-6.
Leitenberg, Milton, Deaths in Wars and Conflicts in the 20th Century, occasional paper, 3rd edition, Ithica, N.Y.: Cornell University, 2006.
Levine, Robert, “Principles of Intelligence Analysis,” Studies in Intelligence, Vol. 65, No. 4, December 2021, extracts.
Lewis, Michael, Moneyball: The Art of Winning an Unfair Game, New York, NY: W.W. Norton & Company, 2003.
Lillis, Katie Bo, Natasha Bertrand, Oren Liebermann, and Haley Britzky, “Exclusive: Russia producing three times more artillery shells than US and Europe for Ukraine,” CNN, March 11, 2024.
Lima, Michael K., “Munitions for Ukraine: Observations and Recommendations,” U.S. Army, April 23, 2024.
Livermore, Thomas L., Numbers and Losses in the Civil War in America, 186165, Cambridge: Houghton, Mifflin and Company, 1900.
Lopukhovsky, Lev, and Boris Kavalerchik, The Price of Victory: The Red Army’s Casualties in the Great Patriotic War, South Yorkshire, UK: Pen & Sword Books Ltd., 2017.
Lord Moran (Charles McMoran Wilson),
Anatomy of Courage, reprint, New York, N.Y.: Carroll & Graf, 2007 (1945).
Love, Albert G., The Medical Department of the United States Army in the World War, Statistics, Part Two: Medical and Casualty Statistics, Washington, D.C.: Government Printing Office, 2025.
_____, War Casualties, Army Medical Bulletin Number 24, Carlisle Barracks, Penn.: U.S. Army Medical Field Service School, 1931.
Lythgoe, Trent J., “Coup d’oeil and Cognition: How to Build Adaptive Tactical Experts,” Military Review, March-April 2003, pp. 95-107.
Mabry, Robert L., John B. Holcomb, Andrew M. Baker, Clifford C. Cloonan, John M. Uhorchak, Denver E. Perkins, Anthony J. Canfield, and John H.Hagmann, “United States Army Rangers in Somalia: An Analysis of Combat Casualties on an Urban Battlefield,” The Journal of Trauma Injury, Infection, and Critical Care, Vol. 49, No. 3, 2000, pp. 515-529.
Manthey, Jakob, Kevin Shield, and Jurgen Rehm, “Alcohol and Health,” The Lancet, Vol. 400, No. 10365, November 2022, pp. 1764-1765.
Marcelo, Joan, Jeffrey L. Jensen, Leonid Peisakhin, and Haoyu Zhai, “New estimates of US Civil War mortality from full-census records,” PNAS, Vol. 121, No. 48, 2024, pp. 1-3.
McIntyre, Joshua, “Syrian Civil War: A Systematic Review of Trauma Casualty Epidemiology,” BMJ Military Health, Vol. 166, No. 4, 2019, pp. 261-265.
McClellan, Edwin M., The United States Marine Corps in the World War, Washington, D.C.: Headquarters, U.S. Marine Corps, 1920.
McRandle, James H., and James Quirk, “The Blood Test Revisited: A New Look at German Casualty Counts in World War I,” The Journal of Military History, Vol. 70, No. 3, July 2006, pp. 667-701..
Miller, Lester L., Jr., Military Casualty
and Statistical Data Associated with Wars, a Two-Part Bibliography, Fort Sill, Ok.: January 1978. As of August 19, 2025: https://apps.dtic.mil/sti/tr/pdf/ ADA047811.pdf
Miller, Sergio, “The Russian Army Death Cult,” Wavell Room, May 14, 2024; and “Russian soldiers fighting in Ukraine given ‘suicide instructions’ – intel,” UKRINFORM, April 12, 2024.
Murdoch, Stephen, IQ: A Smart History of a Failed Idea, Hoboken, N.J.: John Wiley & Sons, Inc., 2007.
Myre, Greg, “The many ways Ukraine carries out unprecedented drone attacks,” National Public Radio, June 7, 2025.
Neel, Spurgeon, Medical Support of the U.S. Army in Vietnam, 1965-1970, Washington, D.C.: Department of the Army, 1991.
New Zealand Institute of Intelligence Professionals, Intelligence Practitioner’s Handbook, August, 2023. As of August 29, 2025: https://nziip.org.nz/wp-content/ uploads/2024/01/nziip-practitionershandbook.pdf
Noonan, David, Those We Forget: Recounting Australian Casualties of the First World War, Victoria, Australia: Melbourne University Publishing, 2014.
Ohtani, Eri, Haruyki Makishima, and Kazuhiro Sakaue, “The recovery and repatriation of the remains of Japanese war dead and the roles of physical anthropologists,” Forensic Science International, Vol. 324, July 2021.
Open Source Centre and Royal United Services Institute, Ore to Ordnance: Disrupting Russia’s Artillery Supply Chains, undated.
Otake, Tomoko, “The long search for Japan’s lost soldiers,” The Japan Times, August 15, 2025.
Palinkas, L.A., and P. Coben, Combat Casualties Among U.S. Marine Corps Personnel in Vietnam: 1964-1972, San Diego, Calif.: Naval Health Research
Center, 1985.
Paul, Christopher, Colin P. Clarke, and Beth Grill, Victory Has a Thousand Fathers: Detailed Counterinsurgency Case Studies, Santa Monica, Calif.: RAND Corporation, 2010.
Peck, Michael, “Breaking Russia’s Triple Chokehold,” CEPA, April 16, 2025.
_____, “Glide Bombs: The Russian Wonder Weapon?,” CEPA, April 9, 2024.
Perdue, Mitzi, “An Explosive Choice: Landmines and Ukraine,” CEPA, November 22, 2024.
Plante, Trevor K., “Researching Confederate Marines in the Civil War,” Prologue Magazine, Vol. 33, No. 4, Winter 2001, no pages given.
Powell, Colin S., U.S. Secretary of State Colin Powell Addresses the U.N. Security Council, February 5, 2003.
Powers, Bryan, “The Battle for Bakhmut: When Is a Battlefield Loss a Strategic Victory?,” U.S. Army, March 31, 2025.
Prysiazhniuk, Oleksandr, Roman Palyvoda, Yurii Chepurnyi, Tetiana Pavylychuk, Denis Chernogorskyi, Igor Fedirko, Yaroslav Sazanskyi, Danylo Kalashnikov, and Andrii Kopchak, “Warrelated maxillofacial injuries in Ukraine: a retrospective multicenter study,” Archives of Craniofacial Surgery, Vol. 26, No. 2, 2024, pp. 51-58.
Radford, Benjamin J., Yaoyao Dai, Niklas Stoehr, Aaron Schein, Mya Fernandez, and Hanif Sajid, “Estimating troop losses on both sides in the Russia-Ukraine War,” analysis blog, The Loop, undated.
Rapin, Ami-Jacques, Shrewdness, coup d’oeil, and genius: the cognitive attributes of the consummate general (Greek antiquity, Byzantine era, modern times, preprint edition, Lausanne, Switzerland: University of Lausanne, 2023.
Rayment, Sean, and Thomas Godfrey, “Killer Drones: Combat drones responsible for almost 80% of all deaths in blood-
soaked Ukraine conflict, Sun on Sunday probe reveals,” The Sun, April 12, 2025.
Razoux, Pierre, with Nicholas Elliott, tr., The Iran-Iraq War, Cambridge, Mass.: The Belknap Press of Harvard University Press, 2015.
Reese, Roger R., Why Stalin’s Soldiers Fought: The Red Army’s Military Effectiveness in World War II, Lawrence, Kan.: University of Kansas Press, 2011.
Reister, Frank A., Battle Casualties and Medical Statistics: U.S. Army Experience in the Korea War, Washington, D.C.: Department of the Army, 1973.
Renaud, Jeff, “Western University-led research debunks the IQ myth,” press release, Western University, December 19, 2012.
Rivero, Jorge, “Innovating under fire: Lessons from Ukraine’s frontline drone workshops,” Modern War Institute, March 25, 2025.
Roberts, Evan, and Alexandra Burda, Correlates and Consequences of American War Casualties in World War I, working paper, University of Minnesota, July 2018. As of August 19, 2025: https://pop.umn. edu/sites/pop.umn.edu/files/roberts_ working_paper.pdf
Roberts, Pat, and John D. Rockefeller IV, Report on the U.S. Intelligence Community’s Prewar Intelligence Assessments on Iraq, report to the U.S. Senate Select Committee on Intelligence, U.S. Senate, undated.
Rosa-Hernandez, Gabriela Iveliz, and Patrick Enochs, We Need a Medic! The Russian Military Medicine Experience in Ukraine, Arlington, Vir.: Center for Naval Analyses, November 2024.
Rubincam, Catherine, “Casualty Figures in the Battle Descriptions of Thucydides,” Transactions of the American Philological Association (1974-2014), 1991, Vol. 121, pp. 181-198.
Ruokonen, Antti, “Ukraine’s Artillery Shell Shortfall,” Lawfare, April 3, 2024.
Saballa, Joe, “Ukraine Slowly Matching Rate of Russia’s Artillery Fires: Defense Official,” The Defence Post, October 2, 2024.
Saghafi-Nia, Masoud, Seyed Masoud Khatami, Nahid Nafissi, and Yadollah Rezaei, “Triage of war-injured troops in the Iran-Iraq War,” Journal of Research in Medical Sciences, Vol. 13, No. 1, 2008, pp. 8-11.
Santora, Marc, Lara Jakes, Andrew E. Kramer, Marco Hernandez, and Liubo Sholudko, “A Thousand Snipers in the Sky: The New War in Ukraine,” The New York Times, March 3, 2025.
Schroden, Jonathan, “Why Operations Assessments Fail—It’s Not Just the Metrics,” Naval War College Review, Vol. 64, No. 4, 2011.
Schunn, Christian D., and J. Gregory Trafton, “The Psychology of Uncertainty in Scientific Data Analysis,” book chapter, Applied Psychologies of Science. As of July 27, 2025: https://www.lrdc.pitt.edu/ Schunn/papers/SchunnTrafton-uncertainty. pdf
Sharp, U.S.G., and William C. Westmoreland, Report on the War in Vietnam as of 30 June 1968, Washington, D.C.: U.S. Government Printing Office, 1968.
Shen-Gunther, Jane, Richard Ellison, Charles Kuhens, Christopher J. Roach, and Steve Jarrard, “Operation Enduring Freedom: Trends in Combat Casualty Care by Forward Surgical Teams Deployed to Afghanistan,” Military Medicine, Vol. 176, No. 1 (or 67), 2011, pp. 67-78.
Silberman, Laurence H., and Charles S. Ross, The Commission on the Intelligence Capabilities of the United States Regarding Weapons of Mass Destruction, report to the President of the United States, Washington, D.C.: Government Printing Office, March 31, 2005.
Sinyuk, Mikola, V. Polishchuk, P. Yuschak, and I. Burachok, “Management of warrelated facial wounds in Ukraine: the Lviv military hospital experience,” BMJ Military
Health, Vol. 171, 2025, pp. 12-15.
Sokolov, Boris, “How to Calculate Human Losses During the Second World War,” The Journal of Slavic Military Studies, Vol. 22, 2009, pp. 437-458.
Staker, Hailey, “28th BW reflects on Odyssey Dawn launch,” press report, 28th Bomb Wing Public Affairs, U.S. Air Force, March 28, 2016. As of July 27, 2025: https://www.afgsc.af.mil/News/ArticleDisplay/Article/706087/28th-bw-reflectson-odyssey-dawn-launch/
Studeman, Bill, “10 Philosophies/Principles of Intelligence,” briefing, September 2013. As of August 29, 2025: https:// www.dni.gov/files/documents/RG/10%20 Philosophies.pdf
Sunshine, Ian, Analysis of 500 US Army Combat Fatalities in Vietnam, July 1967 to November 1968, Edgewood Arsenal, Md.: Department of the Army, September 1970.
Szmitko, Paul E., and Subodh Verma, “Red Wine and Your Heart,” Circulation, Vol. 111, No. 2, pp. 10-11, 2005.
Tanghe, Mila, “Russia Assails Ukraine’s Drone Wall,” CEPA, June 4, 2025: https:// cepa.org/article/russia-assails-ukrainesdrone-wall/
Tarasulo, Ytzhak, “A Profile Of the Soviet Soldier,” Armed Forces & Society, Vol. 11, No. 2, Winter 1985, pp. 221-234.
Thayer, Thomas C., ed., A Systems Analysis View of the Vietnam War, 19561972, Volume 6: Casualties and Losses, Washington, D.C., Office of the Secretary of Defense, February 17, 1975.
Tucker, Patrick, “The West is underestimating Ukraine’s artillery needs,” Defense One, February 22, 2024.
U.S. Army Center of Military History, United States Army in the World War, 1917-1919: Reports of the Commander-in-Chief, Staff Sections and Services, Washington, D.C., 1991.
U.S. Army, Statistical Data on Army Troops Wounded in Vietnam, January 1965-June
1970, Medical Statistics Agency, Office of the Surgeon General (undated).
U.S. Army Worldwide Equipment Guide, or WEG, available as of August 27, 2025 at: https://odin.tradoc.army.mil/WEG
U.S. Government Accountability Office, Operation Desert Storm: Full Army Medical Capability Not Achieved, Washington, D.C., August, 1992.
U.S. Joint Chiefs of Staff, Doctrine for the Armed Forces of the United States, Joint Publication 1, Washington, D.C.: Joint Staff, as of July 12, 2017.
U.S. Marine Corps, Intelligence, MCDP 2, Washington, D.C.: Headquarters, United States Marine Corps, 1997.
Vagts, Alfred, “Battle and Other Combatant Casualties in the Second World War,” parts I and II, part 1 in The Journal of Politics, Vol. 7, No. 3, August, 1945, pp. 256-294.
van Brugen, Isabel, “Russian Soldier Appears To Be Unscathed After Ukraine Drone Headbutt—Video,” Newsweek, August 15, 2024. As of August 28, 2025: https://www.newsweek.com/ russian-soldier-ukraine-drone-videokharkiv-1939721
van Hoogdalem, Anouk, and Anna MT Bosman, “Intelligence tests and the individual: Unsolvable problems with validity and reliability,” Methodological Innovations, Vol. 17, No. 1, pp. 6-18.
Vershinin, Alex, “The Return of Industrial Warfare,” Royal United Services Institute, June 17, 2022.
Vidavular, Ramesh, Hajime Otani, Pawan K. Singal, and Nilanjana Maulik, “Significance of wine and reservatrol in cardiovascular disease: French paradox revisited,” Experimental & Clinical Cardiology, Vol. 11, No. 3, Fall 2006, pp. 217-225.
Vincent, James, Beyond Measure: The Hidden History of Measurement from Cubits to Quantum Constants, New York, NY: W.W. Norton & Company, Inc., 2022.
Walsh, Nick Paton, “Injured Russian
soldiers are being returned to the frontline,” CNN, February 22, 2025, https://www. youtube.com/shorts/-Jfj3m5YDLA
Watling, Jack, “The Peril of Ukraine’s Ammo Shortage,” Time, February 19, 2024.
Watling, Jack, and Nick Reynolds, Meatgrinder: Russian Tactics in the Second Year of Its Invasion of Ukraine, Royal United Services Institute, May 19, 2023.
Watling, Jack, and Nick Reynolds, Tactical Developments During the Third Year of the Russo-Ukrainian War, London, UK: Royal United Services Institute, 2025, p. 10. As of July 23, 2025: https://static.rusi.org/ tactical-developments-third-year-russoukrainian-war-february-2205.pdf
Watling, Jack, and Nick Reynolds, Tactical Developments During the Third Year of the Russia-Ukrainian War, Royal United Services Institute, February 2025.
Watling, Jack, Oleksandr V. Danylyuk, and Nick Reynolds, Preliminary Lessons from Ukraine’s Offensive Operations, 2022-2023, Royal United Services Institute, July 2024.
Weldemichel, Teklehaymanot G., “Tigray war: Modern geographies of mass violence and the invisibilization of populations,” Political Geography, Vol. 118, April 2025, no page numbers.
West, Bing, “Has High Tech Made Artillery Obsolete?,” Strategika, August 30, 2024.
Whalen, Robert Weldon, “War Losses (Germany),” in: International Encyclopedia of the First World War, Version 1.0, October 2014. As of August 17, 2025: https:// encyclopedia.1914-1918-online.net/article/ war-losses-germany/#easy-footnotebottom-4-1088520
Windschitl, Paul D., and Gary L. Wells, “Measuring Psychological Uncertainty: Verbal Versus Numeric Methods,” Journal of Experimental Psychology, Vol. 2, No. 4, 1996, pp. 343-364.
Woodford, Shawn, “‘Love’s Tables’: U.S. War Department Casualty Estimation in World War II,” The Dupuy Institute, March
19, 2018. As of August 19, 2025: https:// dupuyinstitute.org/2018/03/29/lovestables-u-s-war-department-casualtyestimation-in-world-war-ii/
Yamaguchi, Mari, “75 years later, 1 million Japanese war dead still missing,” Associated Press, August 12, 2020. Yan, Xu, and Li Xixo-Bing, “The Chinese Forces and Their Casualties in the Korean War: Facts and Statistics,” Chinese Historians, Vol. 6, No. 2, 1993, pp. 45-48.
Zarei, Mohammad, Mahmoud Farzan, Mohammad Javad Deghani Firoozbadi, Furqan Mohammed Yaseen Khan, and Mehdi Tavokoli, “Extremity war injuries: A retrospective study of the Iran-Iraq War,” Archives of Trauma Research, Vo, 8, No. 1, January 2019, pp. 28-32.
Zoria, Yuri, “Elite Ukrainian drone unit becomes combat brigade, triples size in major expansion,” Euromaidan Press, November 12th, 2024.
Zouris, James M., G. Jay Walker, Judy Dye, and Michael Galarneau, “Wounding Patterns for U.S. Marines and Sailors during Operation Iraqi Freedom, Major Combat Phase,” Military Medicine, Vol. 171, No. 3, 2006, pp. 246-252.
Picture: Vadym Pliashechko / State Border Guard Service of Ukraine / the Collection of war.ukraine.ua
© Battle Research Group Foundation, Inc, 2025 Battle Research Group Foundation, Inc., does not give permission to any entity, corporation, nonprofit, or individual to in any way scrape, capture, or reuse this material, and we explicitly deny the use of this material for use in training artificial intelligence of any type and in any form.
This publication is available at: battleresearchgroup.org/publications
