Game data science el-nasr - Download the ebook now to never miss important information

Page 1


https://ebookmass.com/product/game-data-science-el-nasr/

Instant digital products (PDF, ePub, MOBI) ready for you

Download now and discover formats that fit your needs...

Principles of Data Science Sinan Ozdemir

https://ebookmass.com/product/principles-of-data-science-sinanozdemir/

ebookmass.com

Smarter data science: succeeding with enterprise-grade data and ai projects Fishman

https://ebookmass.com/product/smarter-data-science-succeeding-withenterprise-grade-data-and-ai-projects-fishman/

ebookmass.com

Distrust: Big Data, Data-Torturing, and the Assault on Science Gary Smith

https://ebookmass.com/product/distrust-big-data-data-torturing-andthe-assault-on-science-gary-smith/

ebookmass.com

Obstetrics & Gynecology (Diagnostic Medical Sonography Series) 4th Edition, (Ebook PDF)

https://ebookmass.com/product/obstetrics-gynecology-diagnosticmedical-sonography-series-4th-edition-ebook-pdf/

ebookmass.com

Curriculum Development and Evaluation in Nursing Education

Paperback u2013

https://ebookmass.com/product/curriculum-development-and-evaluationin-nursing-education-paperback/

ebookmass.com

How to Heal a Workplace Kerry Howard

https://ebookmass.com/product/how-to-heal-a-workplace-kerry-howard/

ebookmass.com

Gender, Authorship, and Early Modern Women’s Collaboration (Early Modern Literature in History) 1st ed. 2017 Edition

https://ebookmass.com/product/gender-authorship-and-early-modernwomens-collaboration-early-modern-literature-in-history-1sted-2017-edition-patricia-pender/ ebookmass.com

Microeconometrics Using Stata: Revised Edition

https://ebookmass.com/product/microeconometrics-using-stata-revisededition/

ebookmass.com

Homer and the Poetics of Gesture Alex C Purves

https://ebookmass.com/product/homer-and-the-poetics-of-gesture-alex-cpurves/

ebookmass.com

https://ebookmass.com/product/etextbook-978-1285866383-small-businessmanagement-entrepreneurship-and-beyond/

ebookmass.com

GAMEDATA SCIENCE

MAGYSEIFEL-NASR TRUONGHUYNGUYENDINH ALESSANDROCANOSSA ANDERSDRACHEN

GreatClarendonStreet,Oxford,OX26DP, UnitedKingdom

OxfordUniversityPressisadepartmentoftheUniversityofOxford. ItfurtherstheUniversity’sobjectiveofexcellenceinresearch,scholarship, andeducationbypublishingworldwide.Oxfordisaregisteredtrademarkof OxfordUniversityPressintheUKandincertainothercountries

©MagySeifEl-Nasr,TruongHuyNguyenDinh,AlessandroCanossa,andAndersDrachen2021

Themoralrightsoftheauthorshavebeenasserted FirstEditionpublishedin2021 Impression:1

Allrightsreserved.Nopartofthispublicationmaybereproduced,storedin aretrievalsystem,ortransmitted,inanyformorbyanymeans,withoutthe priorpermissioninwritingofOxfordUniversityPress,orasexpresslypermitted bylaw,bylicenceorundertermsagreedwiththeappropriatereprographics rightsorganization.Enquiriesconcerningreproductionoutsidethescopeofthe aboveshouldbesenttotheRightsDepartment,OxfordUniversityPress,atthe addressabove

Youmustnotcirculatethisworkinanyotherform andyoumustimposethissameconditiononanyacquirer

PublishedintheUnitedStatesofAmericabyOxfordUniversityPress 198MadisonAvenue,NewYork,NY10016,UnitedStatesofAmerica

BritishLibraryCataloguinginPublicationData Dataavailable

LibraryofCongressControlNumber:2021932524

ISBN978–0–19–289787–9(hbk.) ISBN978–0–19–289788–6(pbk.) DOI:10.1093/oso/9780192897879.001.0001

Printedandboundby CPIGroup(UK)Ltd,Croydon,CR04YY

LinkstothirdpartywebsitesareprovidedbyOxfordingoodfaithand forinformationonly.Oxforddisclaimsanyresponsibilityforthematerials containedinanythirdpartywebsitereferencedinthiswork.

FOREWORD

Intuitionvs.analytics

Ihaveworkedwithsomedeeplybrilliantpeople,andwhatamazesmemost istheintuitivegracewithwhichtheycanmakethemyriadmicrodecisions thatgointobuildingaproduct.Thewayourintuitionoperatesondatawe havealreadyinternalizedisamazing.

Butintuitionis,bynecessity,oftenflawed:itissometimesmyopic,basedon pastexperiencesandthepeoplewehavespentthemosttimewith,andmisses factsthatstickoutofthedatalikesorethumbs.

Wehaveallmetthedesignerwhorefusestoevenconsiderinsightsfromdata, because“Picassodidn’tneedanalytics.”Butartanddataarenotinopposition, andfightingblindfoldedisasfoolhardyasitispoetic.

Generosity,collaboration,andtooling

WhenIcametothegamesindustrynearlytwodecadesago,myfavoriteaspect ofitwashowcollaborativeallthesecrazysmartandcreativepeoplewere. Everyonewassharingbestpracticesfromarticlesinthe Gems bookseries, toconferences,totheirblogs.Industrylegendhasitthatevencompeting companiessometimessharedusefulbitsofcodewitheachother.

Thisallenabledsomanyofustomakegreatgamesandgreatgameengines andultimatelymovedtheindustryforwardatthebreakneckpacethathasmade itsofunandcrazyandwildandcompetitiveandinnovativeandfrustrating— andalsothebiggestformofentertainmentbyfar.

Aswegrewupandmovedonline,datastartedpouringin—datawithoutend, andwithoutmeaning.Butwearesmartandresourceful,andtheindustryhas beenadaptingandinventingmethodstounderstandandactonthistorrent.

Yet,ithasbeenfrustratingthatthesharingofgameanalyticsmethodshas laggedbehindotherareas.Toomanywheelsgotreinvented,andthedemocratizationofmethodsandtoolsisnotquitethere.

Ihopethatthisbookinspiresagenerationoftoolmakerstobringtheindustry forward.Datawithoutcodeis justaPowerPoint.

Getsmart

Finally,whetherornotyouintendtomakeuseofthemethodsdescribedinthis book,youcanbesurethatthey will beappliedtoyou.And,inanagewhereyour dataisneverreallyonlyyours,thebestwaytosafeguardyourdigitalidentityis towiseupabouthowitwillbeinterpretedandoperationalized.

GameDataScience providesanin-depthoverviewofthetechniquesand methodsusedtoextractintelligencefromthedatathatcanbegatheredfrom playinggames.Theauthorsareallpioneersinthefieldofgameanalytics,and theyoffertheirexperienceandinsighttobringanyoneuptospeedwithcuttingedgepractices.

Thatbyitselfisexciting.

PREFACE

Thisbookwasdevelopedtogivereadersanintroductiontothepractical sideof gamedatascience.Beforewediscusswhatthatmeans,let’sdelve abitdeeperintowhat“gamedatascience”means.Gamedatascience isatermthatweusetodenoteaprocesscomposedofmethodsandtechniques bywhichananalystoradatascientistcanmakesenseofdatatoallowdecision makersinagamecompanytomakeinformeddecisions.Thetypeofdataused, andstakeholdersinvolvedcanvaryfromcompanytocompany.Forexample, analystsatRiotusegameplaydatacollectedfromplayerstomakedecisions aboutthedesignofthegame.Insuchcases,gamedatascientistswillanalyze datatocomeupwithpatternsthattheycancommunicatetothedesignteam, allowingthemtoadjusttheirdesigns.However,thisisnottheonlyreasonto analyzedatafromgames.Besidesgameplaydata,therearemanyotherusesof gamedata,suchassystem-levelanalysis,marketinganalysisorsegmentation, oradjustingworkflowplans.Inourearlierbook, GameAnalytics:Maximizing theValueofPlayerData,wediscussedsomecasestudiesofsuchdifferentand varieduses.Ifinterested,readersshouldconsultthisbookformoreinformation onhowdatafromgamescanbeusedandthedifferentstakeholdersinvolved. Wewillcoversomeaspectsofthisagainin Chapter1 ofthisbook,butthese topicswillnotbediscussedindepth.

Whenyouareagamedatascientistlookingatgamedata,youmayhave manygoalsand/orquestionsinmind.Someofthesequestionsmaybeaboutthe following:engagement(howcanweengagenewusers?Howlongisanaverage playsession?);retention(howmanyreturningusersarethereinthegame?How muchtimedouserstypicallystayinthegamebeforetheyquit?);gamedesign (didplayerslearnthegamemechanicsordidtheystrugglewithsome?Were thereanydominantstrategiesinthegame?);gamedevelopment(arethereany

apparentbugsinthegame?);money(e.g.,howmuchmoneydidthegamemake? CanIprojecthowmuchmoneythegamewillcontinuetomakeinthefuture?); andsoon.Allthesetypesofquestionsareanswerablethroughgamedatascience techniques.

Industryandacademicresearchershaveworkedhard(andarestillworking) todevelopasetofmethodsthatcanbeusedtoanswerthesequestions. Thesemethodsandtechniquesareborrowedandadaptedfrommanyfields ofstudy,including MachineLearning(ML),datamining,SocialNetwork Analysis(SNA),ArtificialIntelligence(AI), and visualization.Itisimportant torealizethateachofthesefieldshavetheirowncommunityandareactive fieldsofresearchwithseveralvenueswheresuchresearchispublishedand presented.Readerswhoareinterestedinthebasicandadvancedtechniquesin anyoftheseareasareadvisedtoconsultthesecommunitiesandattendconferencesintheseareas,suchas AAAI(AmericanAssociationfortheAdvancementofArtificialIntelligence),InfoVis(InformationVisualization),VAST (VisualAnalyticsScienceandTechnology),ICML(InternationalConference onMachineLearning),andKDD(KnowledgeDiscoveryandDataMining). Therearealsosomegamespecificconferencesthatoftendealwithanalytictechniques,suchas FDG(FoundationofDigitalGames),IEEECOG(Conference onGames),and AAAIAIIDE(ArtificialIntelligenceandInteractiveDigital Entertainment).

Inthisbook,wewilldiscussbasictechniquesthathavebeenappliedin gamedatasciencesofarfromthesefields.Butthegamedatasciencefield isstillinitsinfancy.Researchersandanalystsinthefieldarefollowingsuch conferencescloselytoacquiremoreadvancedtechniquestousetosolveopen problemsorquestions.Thus,oncereadersgetacquaintedwiththebasics,they areencouragedtolookatconferencepapersinthefieldformoreadvanced techniques.

Itisalsoimportanttorealizethateachoneofthemethodsandtechniquesthat wewilldiscussinthisbookisessentiallyborrowedfromdifferentfields.Therefore,understandingthemwillrequiresometheoreticalfoundations,whichwe willintroducethroughoutthebook.Forexample,machinelearningtechniques areoftenbasedon probabilitytheory and informationtheory.Thus,thebook willincludeanintroductiontothesetheories.Wewilldiscussthemwhenthey areapplicabletothemethodsused,andwewillalsoincludeamorepractical exampletoshowhowtheyareapplied.However,thebookwillnotgoindepth inallareas,andreadersmayneedtoconsultfurtherreadingsandreferences togetmorein-depthknowledgeofareascovered.Eachchapterwillcontain

severalpointersforfurtherreadingsandreferencesthatinterestedreaderscan consult.

IntendedAudience

Theintendedreadersofthisbookarestudentswhowanttolearnaboutgame datasciencetechniquesandhowtheyareapplied.Additionally,thebookisalso targetedtothegameindustry:itcaninfacthelpwiththecommunicationacross teams,forexample,gameandsystemsdesignersoftenneedtointerfacewith dataanalyststoformulateappropriateandfeasibleresearchquestions.Thereare noprerequisitestounderstandingthetechniquesinthisbook,althoughmany ofthemethodsdiscussedrequiresomeunderstandingofprobabilitytheoryor otherstatisticaltechniques,whichwewillbeintroducingthroughoutthebook. Thus,thebookshouldbeself-containedandaccessibleforanystudentatthe graduateoradvancedundergraduatelevel.

LabsandSupplementaryMaterials

Alllabsanddataareavailableassupplementarymaterialsatthebook’scompanionwebsite:www.oup.co.uk/companion/GameDataScience.Pleasenavigateto thiswebsiteandthenregistertogainaccesstothelabanddatamaterials discussedthroughthisbook.

ACKNOWLEDGMENT

Wearegratefultosomanypeoplefortheirhelpandsupportaswe embarkedonthejourneyofwritingthisbook.Wefirstwantto thankourfamilies,ourpartners,children,parents,andsiblings whohavealreadyheardalotaboutthisbookaswetirelesslyworkedonit.

Wearealsoindebtedtothemanyamazinggraduatestudentswhohave workedonmanychaptersofthisbook,includingEricaKleinman,Sabbir Ahmed,AndyBryant,MadkourAmrAbdelRahman,NathanPartlan,Luis FernandoLarisPardo,EvelynTan,ValerioBonomettiandOzanVardal.We wouldliketopraisethemfortheirhardworkandfeedbackaswellasthe continuedsupportandhelpaswepolishedthechaptersinvolved.And,weare alsoindebtedtoDr.PaolaRizzoforherreviewandfeedback.

Agreatthankstoourcopyeditorswhohavehelpeduseditthisbookandgave usalotofadviceonstructure,grammarandvision.GreatthankstoMiranda Adkins,SimranDhaliwal,andFergusScott,writingadvisorsandMaster’sstudentsintheComputerScienceAlignProgramatKhouryCollegeofComputer Sciences;IanMagnusson,teamleader,writingadvisorandMaster’sstudentin theComputerScienceAlignProgramatKhouryCollegeofComputerSciences; lastbutnotleastgreatthankstoJaneKokernak,seniorcommunicationsadvisor atKhouryCollegeofComputerSciences.

Further,wewouldliketoacknowledgeourcolleaguesacrossindustryand academiawhohelpedoutinnumerousways.Weareparticularlygratefulto YusufPisanandFoaadKhosmoodfortheirfeedbackandsuggestededitsfor Chapter11ofthebook.Wealsowanttorecognizeourpartnersandcolleagues atSquareEnix,UbisoftMassive,UbisoftMontreal,King,Nordeus,EA,ESL, RiotGames,Bungie,Microsoft,andmanyothers,whohavechampionedjoint

xii | ACKNOWLEDGMENT

explorationandresearchingamedatascienceacrossacademiaandindustry, andacrossdisciplinaryfields,forclosetotwodecadesnow.Gamedatascience hasgrownintoastronglycollaborativedomainandwearegratefultoeveryone whohasbeenpartofthejourneysofar,orwhowilljointhejourneyinthe future.

HOWTOREADTHEBOOK

Thebookisdividedintochaptersthatfolloweachother,soreaders shouldbeabletostartfromthebeginningofthebookandread through.Eachchapterwillincludethefollowing:

• Theoreticalfoundations. Asdiscussedabove,allmethodsusedwill beprecededwithatheoreticaldiscussionoftheirfoundationstoallow readerstounderstandthemfully.

• PracticalimplementationsandLabs. Thebookwillincludepracticalintroductionstoallalgorithmsdiscussed.Almostallchapters, withtheexceptionofoneortwochapters,willincludelabstoallow readerstofollowalongandapplythetechniquesdiscussed.Alllabs willalsocomewithgamedatatoallowreaderstounderstandhow toapplythealgorithmstogamedata.Alllabsanddataareavailableassupplementarymaterialsatthebook’scompanionwebsite: www.oup.co.uk/companion/GameDataScience.Pleasegothiswebsite andregistertogainaccesstothelabanddatamaterialsdiscussed throughthisbook.

• Applicabilitytogamedatascience. Insomesectionswewilldiscussa generaltechniqueoralgorithmandthenfollowwithamorein-depth discussionofhowthesemoregeneralalgorithmsapplytogames.It shouldbenotedthatgamedatascienceisstillagrowingfield,and notallmethodsdiscussedintheAIcommunityhavebeenappliedto games,someforgoodreasons.Wewilldiscusstheseastheycomeup.

• Casestudies. Itisimportanttoshowcasestudiesoftheuseofthe techniquesdiscussedeitherintheindustryoracademicresearch.We willincludethesethroughthebookchapters.Buttheywillbecome

moreapparentinthelaterpartsofthebookwhenwegettomore complexconceptsormethods.

• Exercises. Asanytextbook,thistextbookwillincludesomeexercises attheendofeachchaptertohelpsolidifytheconcepts.

GameDataScience: AnIntroduction

Youmayhaveheardoftheterm gameanalytics or gamedatascience. Infact,youmayhaveevenpickedupthisbookduetotheuseofthe terminindustryoracademiccircles. Gamedatascience hasbecome acornerstoneofgamedevelopmentinaveryshortperiodoftime.Infact, backinthe1990s,noonewouldhavethoughtthatgamedatawouldbecome afieldofstudyandinnovationingameresearchandindustry.Backinthe 1990s,wewerestillworkingondevelopingbettergraphics,developmenttools, anddesignpractices.Fastforwardtonow,gamedatascienceisemergingasa veryimportantfieldofstudyduetotheemergenceofsocialgamesembedded inonlinesocialnetworks.Theubiquityofsocialgamesgivesaccesstonew datasourcesandhasanimpactonimportantbusinessdecisions,giventhe introductionoffreemium1businessmodels.

Gamedatascienceisabroaddomaincoveringallaspectsofcollecting, storing,analyzingdata,andcommunicatinginsights.Itcansupportanyaspect ofdesignanddevelopment,anditisnot only aboutplayerbehavior,although thatiscertainlyanimportantpartoftheprocess.Withamaturedatascienceframeworkinplace,companieshavetheinstrumentstogainobjective knowledgeaboutworkflowsandcompetitors,understandtheircommunities andplayers,improvedevelopmentprocesses,increaseretentionandrevenue,

1 Freemium isamonetizationstrategywherethebareboneserviceisprovidedforfreebut customersareexpectedtopayforadditionalelementssuchasvanityitems,in-gamecurrency,and fastercooldowns.

GameDataScience.MagySeifEl-Nasr,TruongHuyNguyenDinh,AlessandroCanossa,andAndersDrachen,Oxford UniversityPress.©MagySeifEl-Nasr,TruongHuyNguyenDinh,AlessandroCanossa,andAndersDrachen(2021). DOI:10.1093/oso/9780192897879.003.0001

andbuildcapacitytooffergamesforfreetocustomers,asweshalltalkabout morebelow.

Gamedatasciencefundamentallyaimstoadddata-drivenevidencetosupportdecision-makingacrossoperational,tactical,andstrategiclevelsofgame development,andthisiswhyitissovaluable.Itallowsresearchersandthe industrytomoveawayfromguessworkandmakedecisionsbasedoncarefully collected,curated,andanalyzeddata.

Gamedatascienceisthesubjectofthisbook.Afterreadingthisbook,you shouldhaveaclearunderstandingofthecurrentstandardmethodsandtools usedtoanalyzedatacollectedfromgames.Astheknowledgeandpractices ingamedatascienceareexpandingrapidly,theideas,methods,andtools presentedinthisbookwillalsolikelyexpandasnewsolutionsbecomeavailable.Thisbookprovidesanintroductiontothefoundationalapproachesand theoriesthatwillhelpyouunderstandcurrentandfutureapproachesofgame datascience.

Withthisintroductorychapter,youbeginyourjourneyinthefieldofgame datascience.Inparticular,thischapterwillprovideahigh-levelpanoramic introductiontotheprocessesusedtoanalyzeandmakesenseofgamedataand suggestactionableinformationwiththescientificmethodasabaseprocess. Unlikeotherchaptersinthisbook,thisopeningchapterdoesnotcontain practicallabs.Thematerialdiscussedisconceptual,providingyouwiththe basicsasyouembarkonthejourneyofunderstandingandpracticinggame datascience.

1.1Whatisgamedatascience?

Fundamentally,gamedatascienceistheprocessofdiscoveringandcommunicatingpatternsindatawiththepurposeofinformingdecision-makingin differentdomains,suchasbusinessordesign,inthecontextofgames.Assuch, gamedatascienceincludesmanytypesofanalyses,suchassummarizingthe numberofactiveplayerswithinacertaintimeunit,predictingwhenplayers willstopplayingagame,orevaluatingtheperformanceofservers.

Inourpreviousbook, GameAnalytics (SeifEl-Nasr,Drachen,andCanossa, 2013),weusedthetermgameanalyticsratherthangamedatascienceto denotetheprocessofanalyzingandapplyingdatacollectedthroughoutthe developmentprocess.Here,weadoptedgamedatascienceratherthangame analyticsforseveralreasons,mostimportantly,becauseanalyticsinmany communitiesrelatestobusinessintelligenceormakingdecisionsaboutbusiness

aspectsusingdata.Therefore,thereissometimesaconfusionaboutwhether gameanalyticsrefersonlytotheapplicationofdatasciencetoinformdecisionmakingfortraditionalbusinesspurposesorifitalsocoverstheapplicationof datasciencetoinformdesignprocesses.Becausetheapplicationofdatascience toinformdesignisalargepartofthisbook,we,therefore,willusethebroader andmoreinclusivetermofgamedatascience.Thewayweusethistermdenotes thebreadthofthefieldofknowledgediscoveryusingdatacollectedthrough thegamedesign,development,andpost-launchproductionprocesses.

Gamedatascience,thus,overlapssubstantiallywithotherdata-informed processesingamedevelopment,includingGamesUserResearch(GUR)2, businessintelligenceasitisappliedinthegamesindustry,andmarketingand brandresearch.Whilethereismuchongoingdiscussioninthecommunity aboutwhatexactlygamedatascienceisandisnot,inthisbook,wewilladopt aninclusiveviewpoint,ratherthantryingtosetlimitsaroundtheterm.

Tosummarize,gamedatascienceisthetermweusecollectivelyforthe processofprovidingdata-drivenevidencefordecisionsmadeatvariousparts ofthegamedesign,development,andproductionprocesses.Youcanapplythe toolsandtechniquesofgamedatascienceacrossvirtuallyanyaspectofthegame designanddevelopmentprocesses.

1.2Whatisgamedata?

Agreatvarietyofdatacanbecollected,stored,analyzed,andleveragedto gatherintelligencethroughoutthelifetimeofagametitleorgamecompany. Typicalsourcesofdataincludebehavioraldatafromgames,informationfrom advertisingpartnersandotherthirdparties(i.e.,socialmediaplatforms),and datacollectedfrominfrastructure(suchasservers),thedevelopmentprocess itself,marketing,anduserresearch.

Thesevariedsourcesofdatacanbeusedinmanypartsoftheproduction processtoinformgamedesignanddevelopment,includingunderstanding oroptimizingdevelopers’workflowduringproduction,optimizingserver performanceafterrelease,andtestingtoidentifybugsorplayerengagement. Whileevaluationoftechnicalinfrastructureandplatformcompatibilitycan

2GamesUserResearch(GUR)isafieldofstudythatfocusesonunderstandinguserbehaviors, needs,andmotivationsbyanalyzinghowthedesignofacertainapplicationorgameimpactsits audience.Asyouwillseeinthehistorysectionwithinthischapter,researchersworkinginthisarea arealsotightlycoupledwithgameanalystsassomeoftheprocessesusedbygamesuserresearchers alsousegamedata.

providesubstantialdatasetsthatareimportanttotheoperationofagame,in thisbook,wewillnotfocusonthistopicastheintersectionbetweensoftware engineeringanddatasciencedeservesitsownbook.

Inthisbook,wewillfocuson playerdata.Thedataexamplesandpractical exercisesyouwillfindthroughoutthisbookwilluseplayerdata.Thisisbecause playerdatais,byfar,themostcommonlyusedandavailablesourceofdatain gamedatascience.Therearedifferentformsofplayerdata,includingbehavioral datacollectedinrealtimeasplayersplaythegame,andplayerpreferenceor statistics,suchashowmanygamestheyplayedandtheirranksorscores.The behavioraldatacollectedinrealtimeisoftencalled behavioraltelemetry.

Behavioraltelemetry,inamoregeneralsense,isdatathatweconstantly leaveastrailsthroughalltheactionsweperforminourdailylife:borrowing booksfromalibrary,visitingwebsites,purchasingahouse,workingasa middlemanager,orvacationinginSoutheastAsia.Whetherwedriveacar, amotorcycle,orarickshaw,almostanyactionwetakeinthepublicspace canrepresentasyllableofalongersentencethatcontributestocomposingthe narrativeofourlives.Thedigitaltrailsweleavebehindareeveneasiertocollect. Thewayweuseourphonescreatesaconstantlyevolvingrepresentationofwho weare.

Telemetry basicallymeansdatacollectedfromafar.Inthecontextofgames,as peopleplayagame,wecancollectdataaboutwhattheydointhegame,down tothepressofabuttonormovementofamouse,ifsodesired.Thistypeof userdataiscommonlycollectedacrosstheITsector.Theprocessofcollecting andstoringtelemetrydataiseasierthaneverduetocheapandlargestorage solutions,pervasivedeviceconnectivity,andinstrumentationofsoftwareand hardware.

Withindigitalgames,thetrailscanbesodetailedandcomplexthatthey revealaspectsofplayerpersonalities,motivations,andexperiencesthrough theactionsanddecisionstakenwhenplaying,declaredorinferredpreferences, movementpatterns,andtherelationshipsplayersbuild. Behavioraltelemetry is, withinthescopeofbehavioraldata,themostcommonsourceofinformation wehaveandcertainlythemostvoluminous.Behavioraldataallowsustomove beyondfindingpatternsindatatobegindrawinginferenceaboutthemeaning behinddigitalactions.Understandingwhyplayersdoparticularthingsor behavethewaytheydoisvaluable.Itcanbereadilyappliedtoevaluatingand informingdesign,userexperience,andmonetization.

Thegamesindustryhasinvested,especiallyinrecentyears,considerable effortstoestablishexpertise,implementtools,andbuildprocessesthatcan

leveragetheknowledgeextractedfromanalyzingthetrailsofdatathatplayers leavebehind.Themethods—thetoolsetofa gamedatascientist—inmany waysleveragetheknowledgeandmethodsthatalreadyexist,pioneeredin theriseofbigdata,datascience,andArtificialIntelligence(AI).However,it isimportanttorealizethatgamedatascienceoftenendsupdrawingupon knowledgeinfields,suchasdesign,psychology,sociology,informationsystems, userexperience,oruserresearch,whenitcomestoinformingwhatanalysisto runonplayerdata,howtointerprettheresultsofsuchanalyses,and,perhaps morecrucially,howtotranslatetheresultsintoaction.

Thoughknowledgeandanalyticalapproacheshavegrownrapidly,atthe timeofwritingthisbook,gamedatascienceisinmanywaysstillinitsinfancy. Therearenosetstandardsordefinitionsofmetrics,andmuchoftheavailable knowledgeislockedawayduetotheinherent(proprietary)valueindata.On thepositiveside,thismeansthatnowisanexcitingtimetoworkingamedata science.Italsomeansthatthereisanongoingchallengeindevelopingtoolsand methodsthatcanleverageexpertknowledgetoanalyzeandmakesenseofsuch vastamountsofdataandensurethatnewknowledgeinformsdecision-making thattranslatesintoaction.

1.3Advantagesofgamedatascience

Thebenefitsandadvantagesofintegratinggamedatascienceingamedevelopmentaremanyandfar-reaching.Withamaturedatascienceframework inplace,companieshavetheinstrumentstogainobjectiveknowledgeabout workflowandserverworkloadaswellasgainknowledgeabouttheirplayers, gatherinsightsintowhichelementsofacertaingamearemostpopular,and figureoutatwhatpointplayersstopplaying.Inadditiontoinsightsintodesign, thegamesindustryutilizesknowledgegainedfromdatatoincreaserevenue andimproveplayerexperience.Together,thesetwoissuesdrivebusinessand developmentdecisionssincethevectorsformonetizationandplayerexperience arealigned.Abetteruserexperienceturnsintohighersalesandhigherplayer retention.

Intherealmofacademic gameresearch and seriousgames3,theapplication ofgamedatasciencehasgainedsubstantialmomentum,asitallowscompanies

3Atermusedtodescribegamesdevelopedforpurposesotherthanentertainment,suchas training,promotinghealth,citizenscience,orpsychologicalexperiments.

andresearcherstoanalyzetherelationshipbetweenplayeroruserbehavior andtheoutcomesofsuchbehavior,e.g.,increasedawarenessofatopic,health benefits,orlearning.Thediscoveriesbeingmadeusingdata-driventechniques, suchasinthefieldoflearninganalytics,havemajorimplicationsforeducation andhealth.Citizenscienceandcrowdsourcinggamesalsorelyonsuchmethods toincreaseawareness,retention,andmotivation.

1.4Thehistoricalcontextforgame datascience

Gamedatascienceisinmanywaysarelativelyyoungdomain—especially viewedthroughthelensofacademicresearch.However,theapplicationofdata sciencemethodstodatafromgamesorfromgamecompanieshasexpandedso fastandevolvedsorapidlythatitiseasytooverlookthefactthat,adecadeago, usingmachinelearningalgorithmsongamedatawaslargelyunheardof.The historyofgamedatasciencecanthusbethoughtofasbeingshallowbutbroad.

Ingeneral,thereareseveralchallengestomappingthehistoryofgame datascience.First,thesubstantialamountsofknowledgegeneratedarenot recordedanywherethatispubliclyavailable.Companiesinvestresourcesin businessintelligence,andtheresultsareoftentreatedasconfidentialdueto theirbusinessvalue.Similarly,earlyacademicresearchintheareaispublished acrossadozenormoredomainsandthusisextremelyfragmented.Second, therehasbeenasubstantialparallelgrowthindifferentsectorsandcountries, andthusitishardtosaywhenaspecifictechnologywasdevelopedorhowit influencedthedevelopmentofthefield.Third,anyaccountofthehistorical perspectivewillnaturallybebiasedbythespecificareaoffocusorcommunity thattheauthorcomesfrom.

Tohighlightthechallengesindevelopingahistoricaloverviewofgame datascienceoraspectsofit,wehaveincludedanexercisespecificallyonthis topic(seeexercisesbelow).Inthissection,wewillfocusondiscussingsome ofthefactorsthatwethinkhasaffectedthegrowthofthefieldasweseeit, acknowledgingthatwehaveourownbiases.

Thereareseveralwavesofinnovationwithinthefieldoftechnologyand gamesthathavefacilitatedthedevelopmentofgamedatascience.Theobvious technologyinnovationsincludethedevelopmentofpersonalcomputers,the Internet,thedevelopmentofplatforms,suchasFacebookandSteam,thegrowth

ofserveranddatabasetechnologies,computingcapacity,andmachinelearning aswellastherecentdevelopmentsindeeplearning.Below,wediscusssomeof whatwethinkareimportantlandmarksthatledtothedevelopmentofgame datascienceasitstandstoday.

1.4.1TheriseoftheMMOG

Thereareaccounts,fromveryearlygametitles,ofplayerdatabeinggathered. However,priortotheintroductionoffirstMulti-UserDungeons(MUDs) andthenMassivelyMultiplayerOnlineGames(MMOGs),theapplicationof suchdataasanexternalprocess,towardinformingdesign,systems,virtual economies,andotheraspectsofthegameworld,hasbeenfragmentedatbest. WithMMOGs,suchas UltimaOnline,thereemergedaneedformonitoring apersistentgameworld,itsuserbase,andhowthatuserbasemighteven engageinout-of-gametrading(e.g.,selling UltimaOnline characters).MMOG economiesweredesignedandtested,andaccounts,suchastheonebySimpson (2000),showthatgamedatainformedpartofsuchdevelopment,albeitat asimplerlevelthanthekindsofeconomicanalysesthatareoftenrunon contemporaryMMOGs.

Ontheacademicside,earlyanalyticalworkonMMOGswasdevelopedin parallelwithsuchworkintheindustry.MMOGeconomiesandtheiranalysis weregivensubstantialvisibilitybyCastronova,whopublishedworkabout Everquest in2001documentinghowsyntheticworldsandtheireconomies operate,concludingthattheGrossNationalProduct(GNP)oftheseearlygame worldscouldrivalsomereal-worldcountries(Castronova,2001).Fromthis andothercontemporaryworks—e.g.,byWilliamsetal.(2011),Ducheneaut etal.(2006),Yee(2006),andDibbell(2006),aswellasthereleaseof Second Life andothervirtualworlds—broadpublicattentionhasemergedontheuse ofgameworldsandtheopportunitiestheyprovideastoolstoanalyzeplayer behavior.Thiswasturbo-chargedwiththereleaseof WorldofWarcraft andthe impressivesubscriptionnumbersitreached,bringingMassivelyMultiplayer OnlineRole-PlayingGames(MMORPGs)intopublicconsciousness,atleast intheWesternworld; Lineage and GuildWars,inAsiaandbeyond,also deservecredit.

Withtheemergenceofearlyanalyticalworkongamesduringtheyears 2003–2006,suddenly,manyresearchersrealizedthatvirtualworldsprovided fertilespacesforresearchacrosseconomics,behavioralscience,psychology, networklatency,andmore.Aroundtheseyears,someearlyworkssurfaced

acrossindustryandacademiathatshowcasedhowin-gameplayerbehavior couldbeanalyzedforvariouspurposes.Therewerealsomanyexamplesofhow playersthemselvesminedthegamesfordata,e.g.,tobuildonlineguidesand sitesaboutquestsorresourceharvesting.Ingeneral,thereexistedadegreeof dataaccessinMMOGsandinothergamesthatwasnotoftenseeninotherdataheavyITsectors.Itshouldbenotedthattheanalyticalmethodsatthetimewere stilllargelyconfinedtostatisticsand(simple)economicmodeling.

1.4.2Socialnetworkgames

Anotherangleonhowgamedatascienceemergedistheriseofonlinesocial networkplatforms,suchasFacebook.Theproliferationofsocialnetworksled totheemergenceofanewtypeofgame,the SocialNetworkGame (SNG) (Alsénetal.,2016).SNGscouldtapintosocialnetworkdataandusefree-to-play strategiestodrivemonetization,breakingwiththetraditionalretailsalesmodels.Duetotheabundanceofdataavailablefromthesocialnetworkplatforms, andtherequirementtomonitorin-gamebehaviorduetothemonetization strategyadopted,SNGshadabuilt-inimperativeforanalyzingplayerdata.This urgencybroughtanalytics(SeifEl-Nasretal.,2013)totheforefrontofthegames industrybyaround2007–2010.Termssuchas monetization,funnelanalysis, onboardingresearch,First-TimeUserExperience (FTUE),andothersstarted becomingcommonplace.In2011,oneofthefirstbooksaddressingthismarket waspublished,whichincludedalistofimportantmonetizationmetrics,such as DailyActiveUser (DAU)and AverageRevenuePerUser (ARPU)(Fieldsand Cotton,2011).

1.4.3Democratizingdatacollection

Afactorarisingbyaround2010onwardwasthedemocratizationofmetrics collection.Thankstotechnologicalinnovationsoutsidethegamesindustry andtheemergenceofnumerousstart-upcompaniesthatprovidedSoftware asService(SaaS)analyticsplatforms,suchasDeltaDNA,GameAnalytics, Swrve,NinjaMetrics,and,lateron,YokozunaDataandothers.Suchplatforms providedtoolsandcasestudiesshowingtheanalyticsprocess,whichunpacked theprocessofcollectionandanalysisofbehavioraltelemetry.Severalarticles publishedintechmagazinesdiscusshowcompanies,suchasWooga,Zynga, Microsoft,andUbisoft,usetelemetrydata,givingusmoreexamplesandcase studiesonhowananalyticsprocessisimplemented.

1.4.4GamesUserResearch(GUR)

AroundthesametimethatSNGsmadetheirfirstappearance,GURstartedto becomeamainpartofthegamedevelopmentprocess.ThehistoryofGUR isdocumentedinDrachen,Mirza-Babaei,andNacke(2018)andIsbisterand Schaffer(2008).Itisinterestingtonotethattheapplicationofbehavioral telemetrytoinformgamedesignwithinthecontextofAAA4gameswasdriven, toanextent,byuserresearch.Inthemid-2000s,Microsoft’sUserResearch divisiontookgameusertestingseriously,adaptingtechniquesfromthedomain ofHuman–ComputerInteraction(HCI)anddevelopingnewonestospecificallyworkintheuserexperience-focusedgameenvironments.Tothenew field’sbenefit,Microsoft,Bungie,andothercompaniesdiscussedtheirwork andmethods,e.g.,inThompson’sfamous Wired articlein2007(Thompson, 2007).Amayaetal.(2008)notablydetailedtheworkofMicrosoftUserResearch thatintegrateduserresearchwithautomatedrecordingofuserbehavior.Importantly,thisresearchandtheideasitpropagatedenhancedtheroleofbehavioral telemetryasausefulsourceofknowledge.Aroundthesametime,leadingup to2010,severalkeyblogposts,whitepapers,andpresentationsattheGame DevelopersConferenceshowcasedhowtheindustryatlargewasexploring gamedatascienceandbuildingnewtechnologies,methods,andideas.

AnimportantmilestoneinGURanditsinfluenceongamedatascienceis thestartoftheInternationalGameDevelopersAssociation’sGameResearch andUserExperienceSpecialInterestGroup(GRUXSIG)in2012.Thisspecialinterestgroupstartedorganizingandconnectinggamesuserresearchers acrossindustryandacademia,buildingsummits,andfacilitatingknowledge exchange.Thisincrediblywelcomingcommunityhadasignificanteffecton GURand,byextension,gamedatascience.Thegrouptodaycountsmore than2,100membersworldwideandhostsmultipleannualsummits(seeGRUX SIG,2015).

1.4.5GamesasaService

Moregamesdevelopedinthepastfewyearshavefocusedonbeingonline andpersistent,withdownloadablecontent,patches,andupdatesextending thelifetimeofthesegames.HavingaLiveOperations(LiveOps)teamfor

4AAAtitlesaregamesthattypicallyhaveahighermarkinganddevelopmentbudget.

Turn static files into dynamic content formats.

Create a flipbook