EXPECTED WORTH FOR 2 × 2 MATRIX GAMES By Michael R. … · COWLES FOUNDATION DISCUSSION PAPER NO....
Transcript of EXPECTED WORTH FOR 2 × 2 MATRIX GAMES By Michael R. … · COWLES FOUNDATION DISCUSSION PAPER NO....
EXPECTED WORTH FOR 2 × 2 MATRIX GAMES WITH VARIABLE GRID SIZES
By
Michael R. Powers, Martin Shubik, and Wen Wang
May 2016
COWLES FOUNDATION DISCUSSION PAPER NO. 2039
COWLES FOUNDATION FOR RESEARCH IN ECONOMICS YALE UNIVERSITY
Box 208281 New Haven, Connecticut 06520-8281
http://cowles.yale.edu/
1
ExpectedWorthfor2×2MatrixGameswithVariableGridSizes
MichaelR.Powers,1MartinShubik,2andWenWang3
May17,2016
Abstract
Weoffer adetailed examinationof a broad class of 2× 2matrix games as a first step towardconsideringmeasuresofresourcedistributionandefficiencyofoutcomes.Inthepresentessay,onlynoncooperativeequilibriaandentropicoutcomesareconsidered,andacrudemeasureofefficiencyemployed.Othersolutionconceptsandtheformalconstructionofanefficiencyindexwillbeaddressedinacompanionpaper.
JELClassifications:C63,C72,D61Keywords:2×2matrixgames,efficiency,coordination,worthofcoordination.
12×2MatrixGameswithCardinalPayoffs
Inthefolkloreandelementarypedagogyofgametheory,the2×2matrixgameplaysaspecialrole.
Severalofthesegamesbearwell‐knownnames,suchasthePrisoner’sDilemma,StagHunt,andBattle
of theSexes.Although thereareonly144strategicallydifferent2×2gameswithstrictlyordinal
preferences,oneoftenisinterestedinconsideringrelatedgameswithcardinalpreferences,whose
number is unbounded. The present paper is devoted to addressing applications in which it is
desirabletoexaminealargebutfinitesetof2×2gameswithcardinalpreferences.
1.1 OutcomeSets
Ageneric2×2gameisdescribedbythematrixshowninTable1.Here,therowplayerhasthetwo
strategies,“Up”and“Down”(correspondingtorowsi=1,2,respectively),whereasthecolumnplayer
1 Zurich Group Professor of Risk Mathematics, School of Business and Management, and Professor, Schwarzman
ScholarsProgram,TsinghuaUniversity.2SeymourKnoxProfessorofMathematicalInstitutionalEconomics(Emeritus),YaleUniversity,andExternalFaculty,
SantaFeInstitute.3InternationalBusinessSchool,NankaiUniversity.
2
has “Left” and “Right” (corresponding to columns j= 1,2, respectively). This yields four possible
payoffpairs(outcomes),(ai,j,bi,j),fori=1,2andj=1,2.
Table1:AGeneric2×2MatrixGame
Left(j=1) Right(j=2)
Up(i=1) a1,1,b1,1 a1,2,b1,2
Down(i=2) a2,1,b2,1 a2,2,b2,2
Theuniverseofallstrictlyordinalgamesiseasilydenumeratedbynotingthateachofthepayoff
vectors [a11,a12,a21,a22] and [b11,b12,b21,b22] must be permutations of the ordinal vector [1,2,3,4],
yielding a total of 4! × 4! = 576different outcomes.Thisnumber canbedividedby 2 to remove
duplicationsarisingfrominterchangingrows,andbyanother2toaccountforinterchangingcolumns,
leaving the canonical 144 strategically distinct games shown in Appendix 1. Topologically, the
outcomesetsofthesegamesmaybecharacterizedbyasmallersetof24distinctshapes,22ofwhich
aretwo‐dimensional(i.e.,gamesofopposition)andtheremaining2one‐dimensional(i.e.,gamesof
coordination).TheseshapesareshownindetailinAppendix1,wheretheyareassociatedwiththe
144games.
Inconsideringcardinalgames,weassumethatthepayoffpairs,(ai,j,bi,j),maybeexpressedinwell‐
definedunitsofmoneyorgold,withafixedminimalleveloffinenessthatcanbeperceivedand/or
traded.4Isthereanupperboundonhowlargeanindividualpayoffcanbe?Philosophically,onecould
argueineitherdirection;butforallpracticalpurposes,onecanimposealargeenoughupperbound
thatencompassesallpossibleobservationsforagivensociety.Wethereforeinvestigate
aclosedsetof2×2matrixgameswithpayoffsgivenbyelementsintheset ,
for k ∈ {1,2,...}, with a grid size of ∆ = 1/2k−1. Equivalently, one might choose the payoff set
,withagridsizeof∆=1.Althoughthelatterapproachoffersthesimplicityofan
easilycomprehensiblefixedgridsize,theformerprovidesbothaboundedmaximumpayoffsizeand
anintuitivelystraightforwardlimitingprocesstoassesstheimpactofgridsizeonplayerbehavior.
4Wedonotconcernourselveswithindividualpreferencesdirectly,butallowforthepossibilitythateachamountof
money/goldismappedontoindividualpreferencesinsomerisk‐aversemanner.
3
Intheremainderofthepaper,westudyvariouspropertiesofcardinalgamesarrangedinto144
categories associatedwith their corresponding strictlyordinal games.Our investigation relieson
both analytical and simulationmethods. In the latter case, we employ a computer program that
carriesoutthefollowingstepsforeachofgamesG=1,2,...,100,000,foragivenvaluek∈{1,2,...}:
1. Foreachofthefourcells,(i,j)=(1,1),(1,2),(2,1),(2,2),generatetwoindependentrandom
variables,ai,j∼Uniform andbi,j∼Uniform ,
wherethefourpairs(ai,j,bi,j)aremutuallyindependent.
2. Ifeitherai,j=ai0,j0orbi,j=bi0,j0forany(i,j)=(i0,j0),thenrejectthegameandreturntostep(1).
3. Definethecardinal2×2gameGbythefourpayoffpairs(ai,j,bi,j).
4. Separatelyorderthefourai,jandfourbi,jfromlowesttohighest,andlet̃ ai,j=rank(ai,j)∈{1,2,3,4}
and˜bi,j=rank(bi,j)∈{1,2,3,4}forall(i,j).
5. Definetheordinal2×2gameGbythefourpayoffpairs ,andmatchthisgameto
oneofthe144canonicalstrictlyordinalgames.
Bysymmetry,weknowthatthenumberofordinalgamesgeneratedforeachofthe144canonical
formswillbeapproximatelythesame.
1.2 MassProperties
Thegenerationofa largenumberofdistinctcardinalgames,eachassociatedwithoneof the144
canonicalordinalgames,providesthemeanstoconsiderthemasspropertiesofseveralapproaches
togameplay.Asolutionistheoutcome(orsetofoutcomes)derivedbytheselectionofastrategyby
eachofthegame’splayers,andmaybebaseduponawidearrayofindividualplayercharacteristics.
Forthepresent,however,wewilllimitconsiderationto(1)noncooperative,individuallyoptimizing
players,and(2)entropyplayersselectingeachroworcolumnrandomly,withprobability1/2.
4
2 JointMaximumPayoffs
Givenasetofcardinalgamesgeneratedrandomly,asabove,itisnaturaltoconsiderthedistribution
ofthejointmaximumpayoff, ,foragivenk∈{1,2,...},andeasytoseethatthesamplespaceofJMkisgivenbythesetofvalues .
2.1 Distributionfork=1
For the caseofk=1,5this sample space is simply the setof integers {5,6,7,8}, and it isuseful to
associatethesejointmaximawitheachofthreegamecategories:(1)gamesofcoordination,forwhich
JM1=8;(2)mixed‐motivegames,forwhichJM1∈{7,6};and(3)gamesofopposition,forwhichJM1=
5.Forthisbaselinecase,onecanworkoutthedistributionofthejointmaximumasinTable2,from
which it is clear that JM1 is negatively skewed,withmean,median, andmode of 6.875, 7, and 7,
respectively.
Table2:DistributionofJM1
Value #ofGames 8 36 7 60 6 42 5 6 Total 144
2.2 Distributionfork>1
Fork>1,thedistributionofJMkismorecomplex,butmuchcanbelearnedsimplybyconsideringthe
limitingcaseask→∞.Lettingai,jandbi,jbeindependentandidenticallydistributed(contin‐
uous)Uniform(0,4]randomvariablesforall(i,j),onecandefine ,andobservethatJM∞=max{X1,X2,X3,X4},whereX`∼i.i.d.Triangular[0,8];i.e.,
5Wenotethatfork=1,thesamplespaceoftherandomcardinalgamesisidenticaltothesetof144canonicalordinal
games.
5
x2/32
FX(x)=
−x2/32+x/2−1
forx∈[0,4].
forx∈(4,8]
Itthenfollowsthat
fory∈[0,4]
fory∈(4,8]
and
4) fory∈[0,4].
2) fory∈(4,8]
Thisprobabilitydensityfunction,plottedinFigure1,showsthatthedistributionofJM∞isnegatively
skewed,withmean,median,andmodegivenapproximatelyby5.6968,5.7436,and
5.8619,respectively.3
Figure1:ProbabilityDensityFunctionofJM∞
7436; and Mode = argmax{fJM∞(y)}=y∈[0,8]
.The above analysis reveals that the shape of the distribution of the jointmaximum remains
negativelyskewedforlargevaluesofk,justasitisfork=1.Onenoteworthydifference,however,is
6
thatgamesofcoordinationandgamesofoppositionbecomelessandlessprobable,approachingsets
ofmeasurezeroask→∞.
3 NoncooperativeEquilibrium
Theconceptofnoncooperativeequilibriumhasexistedintheeconomicliteraturesincetheworkof
AugustinCournot (1836),butwasmathematically fully formalizedandgeneralizedby JohnNash
(1952).Wedefinetheoutcomeofa2×2matrixgameasanorderedpairofstrategies,(sR,sC),inwhich
the first element denotes the rowplayer’smethodof selecting a row (i∈ {1,2}), and the second
elementdenotesthecolumnplayer’smethodofselectingacolumn(j∈{1,2}).6Wefurtherdefinea
noncooperativeequilibriumasanoutcomeinwhicheachplayerhasnomotivationtochangehisor
herstrategy,giventheindicatedstrategyoftheotherplayer.Restrictingattentiontopurestrategies,
inwhicheachplayer’sdecisionconsistsofafixed(asopposedtorandom)choiceofroworcolumn,
onecanseethat( )constitutesapure‐strategynoncooperativeequilibrium(PSNE)if
andonlyif
i∗=argmax{ai,j∗}i∈{1,2}
and
j∗=argmax{bi∗,j}.j∈{1,2}
A simple illustration of PSNE is given by the ordinal Prisoner’s Dilemma of Table 3. If both
prisonersremainsilent,theneachwillbegivenonlyaminorpenalty(of3);however,ifoneconfesses
andtheotherdoesnot,thentheformerreceivesaverylightpenalty(of4),whereasthelatterreceives
amoreseverepenalty(of1)thanthepenaltyifbothhadconfessed(of2).Forthisgame,itiseasyto
confirmthatthestrategypair(Down,Right)formsaPSNEwithpayoffs(2,2).
Table3:Prisoner’sDilemma
Left(“RemainSilent”) Right(“Confess”) Up(“RemainSilent”) 3,3 1,4
6Inamatrixornormal‐formgame,anymoveisequivalenttoastrategybecausetheplayersfacenocontingencies.
7
Down(“Confess”) 4,1 2,2
Oneof themost importantcontributionsofNash(1952)wastheextensionofnoncooperativeequilibrium from games with only pure strategies to games allowing each player to select aprobabilitydistributionoverhisorherpossiblechoices.Thus,insteadofjustthepure‐strategypairs,(sR,sC)=(i,j),wecanconsiderrandom‐strategypairs,(sR,sC)=(x,y),inwhichx∈(0,1)denotestherowplayer’s(non‐trivial)probabilityofselectingUp(i=1),andy∈(0,1)denotesthecolumnplayer’s(likewisenon‐trivial)probabilityofselectingLeft(j=1).Anoncooperativeequilibriumthatinvolvesrandomstrategiesisreferredtoasamixed‐strategynoncooperativeequilibrium(MSNE).Onecansolveforagame’sMSNEsfromthetwoconditions:
x∗=argmax{a1,1xy∗+a1,2x(1−y∗)+a2,1(1−x)y∗+a2,2(1−x)(1−y∗)}x∈(0,1)
(4.1)
y∗=argmax{b1,1x∗y+b1,2x∗(1−y)+b2,1(1−x∗)y+b2,2(1−x∗)(1−y)}. (4.2)y∈(0,1)
ThesimplegameofMatchingPennies,showninTable4,providesanintuitivelyreasonableuse
ofmixedstrategies.Forthemixed‐strategypair(x,y)=(1/2,1/2),eachplayerreceivesanexpected
payoffof0.However,ifeitherplayerselectedapurestrategy,thentheotherplayer’sbestresponse
wouldcausethefirstplayertolose1unitwithcertainty.7Althoughcointossesarecommonlyused
by individuals in certain decision‐making settings, the use ofmore complicatedmixed strategies
appearstodependheavilyonplayersophisticationandproblemcontext.
Table4:MatchingPennies
Heads Tails Heads 1,‐1 ‐1,1 Tails ‐1,1 1,‐1
3.1 PopulationofPSNEs
Inthestudyofmatrixgames,itisusefultoseparatePSNEsandMSNEsbecausetheformerdepend
onlyonpayoffordinalities,whereasthelatteraresensitivetocardinaldifferences.Forthesetof144
strictlyordinal2×2games,thedistributionofthenumberofPSNEsisgiveninTable5.
7Alternatively,onecouldconsideranordinalversionofthisgameinwhichthepayoffs−1and1arereplacedbythe
payoffranks1and2,respectively.Intheordinalgame,usingthemixedstrategy(x,y)=(1/2,1/2)giveseachplayerthesameopportunitytoreceive1or2.
8
Table5:DistributionofPSNEs
Value #ofGames0 181 1082 18
Total 144
3.2 PopulationofMSNEs
FromthelistinAppendix2,onecanseethatthesetof144canonicalgamescontains36gameswith
exactlyoneMSNE,andnogameswithmorethanoneMSNE.These36gamesmatchexactlywiththe
setofgameshavingeitherzeroortwoPSNEs,fromwhichitfollowsthatall144ordinalgamescontain
atleastonenoncooperativeequilibrium.
3.3 SomeEasyCalculations
Mostfeaturesandsolutionsof2×2matrixgames–whetherstrictlyordinalorcardinal–areeasyto
calculate. Indevelopingprocedures forsuchcalculations,dominantrowsand/orcolumnsplayan
importantrole.
Arow[column]inamatrixgameissaidtodominate(strictly)anotherrow[column]ifandonly
ifeachpayoffinthefirstrow[column]isgreaterthanthecorrespondingpayoffinthesecondrow
[column].Ina2×2game,itiswellknownthat:
• If a gamepossessesexactlyonedominant rowor column, then itmustpossessexactlyone
PSNE.
• Ifagamepossesseszerodominantrowsorcolumns, then itmustpossesszeroPSNEs(and
thereforeexactlyoneMSNE).
• Ifagamepossessesonedominantrowandonedominatingcolumn,thenitmaypossesseither
(a)exactlyonePSNEor(b)twoPSNEsandexactlyoneMSNE.
TheabovefactsenableustoconstructthefollowingalgorithmforcomputingallPSNEsandMSNEs:
9
1. Checkeachrowandcolumnforthedominanceproperty,andletd∈{0,1,2}denotethetotal
numberofdominantrows/columns.
2. Ifd=1,thensolvefortheuniquePSNEbyidentifyingthecellinthedominantrow[column]
thathasthegreaterpayoffforthecolumn[row]player.
3. Ifd=0,thensolvefortheMSNEexplicitlyas
.
4. If d = 2, then solve for each of the two PSNEs using the method described in step (1)
immediatelyabove,andsolvefortheMSNEasinstep(2)immediatelyabove.(Thetwo
PSNEsalsocanbefoundascornersolutionsofthesystemofequationsin(4.1)and(4.2).)
4 PopulationsofThreeSpecialGames
InAppendix3,weprovidetheC++programusedtogeneratecardinal(andstrictlyordinal)gamesas
describedinsteps(1)through(5)ofSubsection1.1.Appendix4containsvarioussamplemeansand
variancesassociatedwith100,000cardinalgamesgeneratedwithagridsizeof∆=1/256(i.e.,fork
=9).
Inthepresentsection,weconsiderthepopulationsofthreespecialgamesdiscussedwidelyinthe
behavioralscienceliterature:
• Prisoner’sDilemma(game1ofAppendix2);
• StagHunt(game41ofAppendix2);and• BattleoftheSexes(game10ofAppendix2).
The commonnames of these games attach considerable context to the abstract payoff structure,
whichmayormaynotbejustified.Inparticular,thereislittleindicationthatanyonewhodoesnot
know these common nameswould associate themwith the specific games involved (see, e.g., I.
10
PowersandShubik,1991).Nevertheless,thestructuralfeaturesofthesethreesettingsallowusto
illustrateseveralimportantaspectsof2×2games.
4.1 Prisoner’sDilemma
PossiblythemoststudiedofallgamesisthePrisoner’sDilemma,whosepopularityarisesatleastin
partbecauseitsordinalformistheonlygamewithinthecanonical144forwhich:(a)thereisaunique
PSNE, (Down,Right), that is strictlydominatedbyanother feasibleoutcome, (Up,Left); and (b)all
otheroutcomesareParetooptimal.Figure22ofAppendix1presentsthepayoffset forthiswell‐
knowngame.
InexploringthepopulationofcardinalPrisoner’sDilemmagamesforagivenchoiceofgridsize,
∆=1/2k−1, it ishelpful to thinkof theentiredomainofpossiblepayoffs, from thegamewith the
smallestpayoff values, inTable6, to thatwith the largestpayoff values, inTable7.Naturally, an
ordinaltreatmentofpreferencesrecognizesnodifferencebetweenthesetwogames.
Table6:Prisoner’sDilemma,SmallestPayoffs
Left RightUpDown
Table7:Prisoner’sDilemma,LargestPayoffs
Left Right
Up 4
Down 4
Askincreasesandthegridbecomesfiner,thepopulationofcardinalPrisoner’sDilemmagames
coversmoreandmoreoftheentiresquareinterval(0,4]2.Inallcases,thegamehasonlythesingle
PSNE(Down,Right),whosepayoffpairdependsontheparticularvaluesofai,jandbi,jgeneratedin
step(1)ofSubsection1.1.
Asabaseline,wenotethatfork=1,thePSNEpayoffpairisalways(2,2),whereasentropyplayers
do better on average, obtaining an expected payoff of (2.5,2.5). Figures 2 and 3 provide scatter
11
diagramsofpayoffpairsfrom100,000gamesfork=9and∆=1/256.Figure2showsthePSNEs,and
Figure3showsthecorrespondingentropicoutcomes.Ineachcase,thesample‐meanpayoffpairis
indicatedbyareddiamondnearthecenterofthepointcluster.
Figure2:Prisoner’sDilemma,NoncooperativeEquilibrium
Figure3:Prisoner’sDilemma,EntropyPlayers
4.2 StagHunt
TheStagHunt(Table8)possessestwoPSNEs,oneofwhich,(Up,Left),strictlydominatestheother,
(Down,Right).However, dependingonunderlyingassumptions, it canbe argued that the smaller
payoffpairsometimeswillbechosenbyrationalplayers.Inparticular,iftherow[column]player
believesthatthecolumn[row]playerwillchooseRight[Down]withprobabilitygreaterthan1/2,
thenheorshewillbemotivatedtochooseDown[Right].
Table8:StagHunt
Left Right
12
Up 4,4 1,3 Down 3,1 2,2
Fork=1,thetwoPSNEshavepayoffpairs(4,4)and(2,2),respectively,andtheMSNE,(x∗,y∗)=
(1/2,1/2),yieldsexpectedpayoffsof(2.5,2.5).Giventhevalueof(x∗,y∗),onecanseethattheMSNE
yieldsidenticalstrategiesandpayoffsasthegamewithentropyplayers.
Figures4and5providescatterdiagramsofpayoffpairsfrom100,000gamesfork=9and∆=
1/256.Figure4includesthePSNEsandMSNE,eachwithequalfrequency,andFigure5showsthe
corresponding entropic outcomes. As before, the sample‐mean payoff pairs are indicated by red
diamondsnearthecentersofthepointclusters.
Figure4:StagHunt,NoncooperativeEquilibrium
Figure5:StagHunt,EntropyPlayers
13
4.3 BattleoftheSexes
UnlikethePrisoner’sDilemmaandStagHunt,theBattleoftheSexes(Table9)isnotasymmetric
game(i.e.,ai,j6=bj,iforsome(i,j)).LiketheStagHunt,however,itpossessestwoPSNEs,(Up,Left)and
(Down,Right),and1MSNE,(x∗,y∗)=(1/2,1/2).
Table9:BattleoftheSexes
Left Right Up 4,3 2,2 Down 1,1 3,4
Fork=1, the twoPSNEshavepayoff pairs (4,3) and (3,4), respectively, and theMSNEgives
expectedpayoffsof(2.5,2.5).Thus,asinthecaseoftheStagHunt,thegamewithentropyplayers
yieldsidenticalstrategiesandpayoffsasnoncooperativeplayersusingtheMSNE.
Figures5and6providescatterdiagramsofpayoffpairsfrom100,000gamesfork=9and∆=
1/256.Figure5includesthePSNEsandMSNE,eachwithequalfrequency,andFigure6showsthe
correspondingentropicoutcomes.Onceagain, thesample‐meanpayoffpairsare indicatedbyred
diamonds.
Figure6:BattleoftheSexes,NoncooperativeEquilibrium
14
Figure7:BattleoftheSexes,EntropyPlayers
4.4 CorrelatedStrategiesandEfficiency
InexaminingtheresultsoftheStagHuntandBattleoftheSexes,weobservethattheexpectedpayoffs
oftheMSNEarestrictlylowerthanatleastonepairofPSNEpayoffsinbothgames.Thissuggestsa
potentialproblemofineffectivecoordination.Inotherwords,theplayersmayarriveata
noncooperativeequilibriumforwhichtheindividualand/orjointpayoffs(i.e.,ai∗,j∗,bi∗,j∗,and/or
ai∗,j∗+bi∗,j∗)arelessthanoptimal.Therefore,thevalueofbeingabletocoordinatestrategiesmaybe
substantial.
Table10offerssomequantificationofthedeficienciesattributabletoineffectivecoordinationfor
thethreegamesdiscussedabove.Foreachgame,thesecondcolumnpresentstheaverageofthejoint
payoffsoverallnoncooperativeequilibria,givingequalweighttoeachPSNEandMSNE.Forthe
Prisoner’sDilemma,thisis2+2=4;fortheStagHunt,itis[(4+4)+(2+2)+(2.5+2.5)]/3≈5.6667;andfortheBattleoftheSexes,itis[(4+3)+(3+4)+(2.5+2.5)]/3≈6.3333.Inthethird
column, we construct a simple efficiency measure by dividing the average joint payoff by the
maximumpossiblejointpayoffthatcanbeachievedbyeitherapure‐ormixed‐strategypairimposed
byanexogenousagency(custom,law,privateintermediation,etc.).Sincetheagentisexogenous,it
canexpandthedomainofmixedstrategiestocorrelatedstrategies,forwhichtheplayers’random
selectionsofUpandDownarestatisticallydependent.ForthePrisoner’sDilemma,thisyields3+3=
6;fortheStagHunt,ityields4+4=8;andfortheBattleoftheSexes,ityieldseither4+3=3+4=7
or[(4+3)p+(3+4)(1−p)]=7,wherethelattervaluecomesfromanycorrelatedmixedstrategy
15
thatchooses(Up,Left)and(Down,Right)withprobabilitiespand1−p,respectively.8Thefourthand
fifthcolumnspresentcorrespondingcalculationsforgameswithentropyplayers.
Table10:JointPayoffsfork=1
Game Avg.ofNEPayoffs NEEfficiency EntropyPayoffs EntropyEfficiency
Prisoner’sDilemma 4.0000 0.6667 5.0000 0.8333StagHunt 5.6667 0.7083 5.0000 0.6250
BattleoftheSexes 6.3333 0.9048 5.0000 0.7143
5 Discussion
5.1 EfficiencyAnalysisofAll2×2Games
Table 11 provides efficiency measures – as defined in the previous subsection – for the entire
populationofcardinalgamesgeneratedinsteps(1)through(5)ofSubsection1.1fork=1andk=9.
Thistableaddressesthenatureoftheoptimalityofindividualbehaviorwithinallpossible
2×2‐gamestructures,subdividedbythevaluesofJM1intheassociatedcanonicalordinalgame.
Thus,forclarity,wewouldnotethat:(a)Prisoner’sDilemmagamesareincludedinthecategoryof
JM1=6;(b)StagHuntgamesareincludedinJM1=7;and(c)BattleoftheSexesgamesareincluded
inJM1=8.
Table11:EfficienciesofAllGames
k=1 k=1 k=9 k=9
NE Entropy NE Entropy
JM1=8 0.9410 0.6250 0.7532 0.5000JM1=7 0.9127 0.7143 0.7305 0.5716JM1=6 0.9352 0.8333 0.7491 0.6669JM1=5 NA NA NA NAAverage 0.9300 0.7386 0.7445 0.5910
As is often true in game theory, behavioral paradoxes abound. We purposely indicate that
efficiencymeasuresare“notapplicable”forthecaseofJM1=5,becausethesegamesofoppositionare
qualitativelydifferentfromallothers.Formally,theefficiencymeasurescouldbedefinedas1.0,but
8WithintheconventionalBattleoftheSexesstoryline,themanandwomanwhoaretryingtodecidewhichmovietosee
couldchoosebetween“his”movieand“her”moviebytossingacoin(forwhichp=1/2).
16
suchcalculationswouldbemisleadingbecausethesegamesdonotreflect thecharacteristicsofa
society.Specifically, thereisnoroomforcooperation,coordination,oranyformofdiscourse,and
whateverone individualgains theother loses.Asnotedbefore,suchstructuraldystopiasbecome
increasinglyrareinthe“Flatland”(seeAbbott,1952[1884])ofmatrixgamesthatarisesfor large
valuesofk.
By definition, the category of JM1 = 8 comprises games of coordination. These games always
includeanoutcomewithpayoffpair(4,4),andsuchoutcomesmustbePSNEs.Inthiscase,thereason
whyefficiencyisnotexactly1.0isthatsomegameshavetwoPSNEsand1MSNE,andtheselowerthe
average.
Somewhatsurprisingly,thefall‐offintheefficiencyofnoncooperativeequilibriaask increases
from1to9isratherlargeforallgamecategories.Fork=1,theefficiencylossingamesofcoordination
(JM1=8)andmixed‐motivegames(JM1=7,6)isbetween6and9percent.Fork=9,thisgrowsto
between25and27percent,andresultsfork=10indicatethatk=9isveryclosetothelimit,with
differencesinefficiencyoflessthan0.01percent.(SeeTable12.)Partofthedecreaseinefficiencyis
attributabletothefactthatE[JMk]→E[JM∞]≈5.6968forlargek
,whichissubstantiallylessthanE[JM1]=6.875.Table12:Efficienciesfork=9andk=10
k=9 k=9 k=10 k=10
NE Entropy NE Entropy
JM1=8 0.7532 0.5000 0.7528 0.5004JM1=7 0.7305 0.5716 0.7310 0.5717JM1=6 0.7491 0.6669 0.7489 0.6672JM1=5 NA NA NA NAAverage 0.7445 0.5910 0.7447 0.5914
Wealsoconsiderthecompletelydifferentbehaviorofentropyplayers,whomaybeviewedas
“know‐nothing”or“zero‐intelligence”decisionmakers.Althoughitiseasytoconstructgames(e.g.,
thePrisoner’sDilemma)inwhichtheentropyplayers’expectedpayoffsof(2.5,2.5)aregreaterthan
thoseofnoncooperativeplayers,Table11showsthatforbothk=1andk=9,entropyplayersperform
worseonaveragethannoncooperativeplayers.Thisisbecausetheyareunabletotakeadvantageof
certaingamestructures,suchasrowand/orcolumndominance,thatassistcoordination.Eveninthis
17
highly constrained environment, differences in the cardinalmeasuresof payoffs yield far greater
variabilityandinequalitywhenkislargethanwhenk=1.Themeaningofthischangeisthatasthe
varietyofoutcomesgrows,theworthofcoordinationorcollaborationgrowsaswell.
5.2 Whythe2×2CaseIsSoImportant
Therearemanyreasonswhy2×2gamesarecrucialbothtothestudyofgametheoryspecifically,and
behavioralsciencemoregenerally.Theseinclude:
1. They offer a highly useful starting point for illustrating and contrasting many problems
andparadoxesinstrategicanalysis.
2. Theyarewidelyusedbyintroductoryinstructorsofgametheoryinthebehavioralsciences.(Is
thispedagogicalusejustified?Wewouldarguethatitis,withappropriatequalifications.)
3. They greatly facilitate analogy generation and storytelling in connecting specific real‐
worldproblemstoabstractmodels.Hence,theyprovidevaluableexercisesintyingthephysical
worldtomathematics.
4. They offer minimal repeated‐game models for the dynamics of learning, signaling, and
othercomplexhumanbehaviors.
5. Throughahandfulofspecialcases(e.g.,thePrisoner’sDilemma,StagHunt,andBattleofthe
Sexes),theysuccessfullyillustratefundamentalproblemsinstrategyandsociety.
Items(1),(2),and(4)arehighlyrelatedforbothteachingandresearch,especiallyifwebelievethat
dyadicrelationsareofconsiderableimportanceindescribinghumanandotheranimalbehavior.Itis
thusforgoodreasonthatelementarytextbooksinthebehavioralsciencesaboundwith2×2‐game
examples.
6 Forn>3,aBasicChangeintheParadigm
Onthewhole,wewouldarguethatthestudyof2×2gamesisimportantbecausesomuchofhuman
activityiswellmodeledbyinteractionsbetweentwoindividualsorbetweenoneindividualandan
18
institution,withrelativelyfewchoicesforeachpartyintheshortterm.Thisperspective,however,
canblindusagainsttheenormouscomplexitiesthatarisewhenincreasingmatrixsize.Forexample,
inthecaseofa3×3matrix,thenumberofdistinctstrategiccasesrisesto(9!)2/(3!)2≈
3.6578×109;andinthecaseof3playerswith2strategieseach,wehavea2×2×2matrixwith(8!)3
/(2!)3≈8.1935×1012strategicallydifferentpossibilities.
Inthecaseof3×3games,thesimpleexampleofTable13issufficienttodestroythehopesof
those interested in developing plausible dynamic strategic solutions. One natural candidate for
simpledynamicsisoptimalresponse;thatis,playersconsiderwheretheyhavebeeninaprevious
playofthegame,andusethatasthebasisfortheircurrentoptimization.However,abriefglanceat
Table13 shows that if the rowand columnplayersbeginwith strategies (sC,sR) = (1,1), then the
column player will move to sC = 3, and a 4‐cycle will emerge that never converges to the joint
maximumPSNEat(sC,sR)=(2,2).(Notethatifthepayoffat(sC,sR)=(2,2)were(1,1)insteadof(9,9),
thatparticular outcomewould still be aPSNE.)Quint, Shubik, andYan (1995)demonstrated the
extensive potential for cycling in a large class of n × n games involving both sequential and
simultaneousmoves.
Table13:ASimple3×3MatrixGame
sC=1 sC=2 sC=3 sR=1 4,1 0,0 1,4
sR=2 0,0 9,9 0,0
sR=3 1,4 0,0 4,1
6.1 SmoothandRoughGames
Inany2×2game,itispossibletocompute4firstdifferencesbetweencellpayoffs.Inthe3×3case,
onecancompute12firstdifferencesand4seconddifferences,andthepayoffsetsstillhaveonlyfew
hillsandvalleys.However, formatricesof4×4andabove, thepotential roughnessof thepayoff
structureincreasesrapidly.Althoughthemathematicalstructureisclearlydefined,theanalysisof
applied decision problems becomes extremely difficult unless some appropriate smoothing
mechanism is imposed on the payoff surfaces. In corporate, military, and political planning, the
adjustment,evaluation,refinement,anddiscardingofmanyfeaturesofastrategicprocesstendto
19
narrowthefinalchoicestoasetthatincludescertainlylessthanten,andoftennotmorethantwoor
three,alternatives.
Thesesomewhatterseremarkswillbeenlargedinfurtherwork.
6.2 Coordination,Opposition,andNoncooperativeEquilibria
As we increase the number of strategies, or players, or both, the relative numbers of games of
coordinationandoppositionbecomevanishinglysmall.
Ifweconsiderm×nmatrixgameswithm,n≥3,therelativenumberofPSNEsdrops,andMSNEs
proliferate.Thereisaliteratureillustratingthis,whichincludeslimitingformulasfortheprobability
ofencounteringaPSNEinalarge,randommatrixgame.(SeeGoldberg,Goldman,andNewman,1968,
Dresher,1970,andI.Powers,1990.)
6.3 LifeIsaSetofMeasureZero
Aguidingprincipleforexploringtheenormousuniverseofmatrixgamesistoselectappropriate
limitingprocessestoobtainrobustsetsofgamesaddressingimportantquestionsofinterest.Wedo
thathereintakingthelimitofallcardinal2×2gameswithaminimalgridsize.
Ingeneral,manyproblemsofinterestrequiresuchspecificationthattheycanberegardedassets
ofmeasurezerowithrespecttofarlargerabstractcategoriestowhichtheybelong.Animportant
example(towhichwewillreturninasubsequentpaper)isthepresenceofties.Inmanycontextsof
humanactivity,tiesinperceivedvaluationarepresent.Furthermore,whenoneindividualhasfiner
perceptionsthananother,theincreasedprecisionalmostalwaysworkstohisorheradvantage.In
thepresentessay,wehaveruledouttiesforsimplicityandmanageability.Ifwehadpermittedthem,
thenumberofcanonicalordinalgameswouldhaveincreasedfrom144to726(NEEDREFERENCE
HERE).
20
7 ConcludingRemarks
Theprincipalpurposeofthispaperwastofindareasonablynaturalwaytoconsiderallcardinal2×2
gameswithinagivenfinitegrid.Theessentiallycombinatoricaspectsoftheinvestigationcalledfor
simulationtoevaluatestructuralaspectsthataredifficulttoseeusingonlyanalyticmethods.Our
study of this structure provided a sufficiently rich background for examining in detail both
noncooperative equilibria and entropy‐player solutions. In a subsequent paper,wewill consider
othersolutionconceptsthatenableonetoinvestigatetheinfluenceofstructureonbehaviorwith
variousintents.
Theamountof“fat left inthesystem”dependsonthesolutionused. Inmanycases,areferee,
government,orotheroutsideagencycouldbeusedtoguidethesystemtoasuperioroutcome,while
consuming fewer resources than it adds. The gap between the current solution and the joint
maximumisthemaximumworthofthecoordinator.Giventhebasicanalysisofthepresentpaper,we
nowareinapositiontoconsiderdevelopingimprovedmeasuresofefficiencyandsymmetryforany
outcomeinamatrixgame.Weplantodiscussthistopicaswellinafuturepaper.
Inshort,thisfirstessaywasaimedatprovidingasimpleideaoftheworthofgovernment,witha
quickandcrudeestimate.Asecondessaywillbedevotedto themanyproblemsofstructureand
behaviorarisingfromnoncooperativeequilibriain2×2games.Finally,athirdessaywilladdressthe
developmentofmoresophisticatedefficiencymeasuresbasedupontheanalysesofthepriorwork.
21
References
[1] Abbott,E.A.,1952[1884],Flatland:ARomanceofManyDimensions(6thed.),NewYork:Dover.
[2] Dresher, M., 1970, “Probability of a Pure Equilibrium Point in n‐Person Games”, Journal of
CombinatorialTheory,8,134‐145.
[3] Goldberg,K.,Goldman,A. J.,andNewman,M.,1968,“TheProbabilityofanEquilibriumPoint”,
JournalofResearchofNationalBureauofStandardsU.S.A.,72,B,93‐101.
[4] Kilgour,D.M,andN.M.Frazer.1988.ATaxonomyofAllOrdinal2x2GamesTheoryanddecision
24:99‐117
[5] Powers,I.Y.,1990,“LimitingDistributionsoftheNumberofPureStrategyNashEquilibriainn‐
PersonGames”,InternationalJournalofGameTheory,19,3,277‐286.
[6] Powers,I.Y.andShubik,M.,1991,“TheNamesoftheGames”,InternationalSecurityandArms
ControlDiscussionPaper,No.7,YaleUniversity.
[7] Quint,T.,Shubik,M.,andYan,D.,1996,“DumbBugsandBrightNoncooperativePlayers:
Games,Context,andBehavior”,inUnderstandingStrategicInteraction:EssaysinHonorof
ReinhardSelten,(W.Albers,etal.,eds.),Berlin:Springer‐Verlag,pp.185‐197.
22
Appendix1
FIGURE 1 FIGURE 2
3, 4 2 ,34,2 , 132 , 43 3, 4 1,,4 43,24, 2
,44,31,13,21,11,1 3 2,412,3 ,2,12ºªºª ººªª ºª»« »« »« »« »« ¼¼ ¼¬ ¼ ¼
,4 2,34 4,4 3,,4 4, 4 2,323,2 4,1 1,1 ,3 1,12,3 3,221,1 3,2 1
ºªª ººªºª »« »« »« »«¼ ¼¬ ¼ ¼2P
3
42P
4
3
1P1
22
1 1P2 3 41 1 2 3 4
19,3,93,122 33,80,105,76,125
23
FIGURE 3 FIGURE 4
20
1 1 2 3 4 3 4
24
1 2 3 4 1 2 3 4
101 35,53,21,103
FIGURE 5 FIGURE 6
1 1 2 3 4 3 4
25
FIGURE 7 FIGURE 8
21
ª4,4 1,2ºª4,4 3,1ºª4,4 2,3º ª3,4 4,1ºª3,4 1,2ºª3,4 4,1ºª3,4 2,3ºª2,3 4,1º«¬3,1 2,3»«¼¬1,22,3»«¼¬3,11,2¼» «¬2,31,2»«¼¬2,34,1»«¼¬1,22,3»«¼¬1,2
4,1»«¼¬1,2 3,4»¼
1 1 2 3 4 3 4
26
3
2P4 4
3
2P
22
1P12 3 4
1P11 1 2 3 4
140,29,106 30,128,16,71,110
1 1 2 3 4 3 4
27
FIGURE 9 FIGURE 10
FIGURE 11 FIGURE 12
22
ª4,4 3,2ºª4,4 3,2ºª4,4 2,1ºª4,4 2,1º ª3,4 4,2ºª3,4 2,1ºª3,4 4,2ºª3,4 2,1º«¬1,3 2,1»«¼¬2,11,3»«¼¬3,21,3¼¬»«1,3 3,2»¼ ¬«1,3 2,1»«¼¬1,34,2»«¼¬2,11,3»«¼¬4,2
1,3»¼
1 1 2 3 4 3 4
28
FIGURE 13 FIGURE 14
3
2P
4 4
3
2P
22
1P12 3 4
1P1
1 1 2 3 4
28,85,113,132 86,63,96,44
1 1 2 3 4 3 4
29
FIGURE 15 FIGURE 16
23
1 1 2 3 4 3 4
30
M M
FIGURE 17 FIGURE 18
M M ª3,2 2,1º ª3,3 4,2ºª3,3 4,2ºª3,3 2,1º «¬4,31,4»¼ «¬1,42,1»«¼¬2,11,4»«¼¬4,21,4»¼
22 13,1, ,43 24 ,4, 4,2413 ,2 ,2 1,3 1,3 1,34,
ºª ººªª »« »« »«134,2,4 , ,4, 14412 2,, 32 2,4
2, 131 4,3, 2 ,3 4,1 3,2 1,3 1,3ºª ºª ºªºª »« »«« »»«
3
¬ ¼¼¼ ¼ ¼¬ ¼ ¼2P
4 4
3
2P
22
1P12 3 4
1P1
1 1 2 3 4
43,64,112 114,45,6,88
1 1 2
143 89,24,142
FIGURE 19 FIGURE 20
24
M M
ª2,2 3,1º ª3,3 4,1ºª3,3 4,1ºª2,2 4,1ºª2,2 4,1º «¬4,3 1,4»¼ «¬1,42,2»«¼¬2,21,4»«¼¬3,31,4»«¼¬1,4 3,3»¼
1 2 3 4 1 2 3 4
57 12,99,138,1
FIGURE 21 FIGURE 22
2P 2P
3
4 4
3
1P 1P
22
3
2P
4 4
3
2P
22
1P1 1P1
32
M M ª2,3 4,2ºª2,3 3,1º ª2,3 4,1ºª3,2 4,1ºª2,3 4,1º «¬1,43,1»«¼¬4,21,4»¼ «¬1,43,2»«¼¬2,31,4¼¬»«3,21,4»¼
FIGURE 23 FIGURE 24
25
Appendix2WhatisaSolutiontoaMatrixGame? Appendix2 MartinShubik
Game# Payoff
Matrix
Shape
Joint
Max PSNEs Symmetric
NashPayoff
Dom.
Pareto
Optima Transpose Row Col.
1 (1,4) (3,3)
22 6 1 Sym 2 2 2 3 NA (2,2) (4,1)
2 (1,2) (3,1)
11 7 1
2 4 2 2 115 (2,4) (4,3)
3 (1,1) (3,2)
3 8 1 Sym 4 4 2 1 NA (2,3) (4,4)
4 (1,4) (3,3)
20 6 1
4 2 1 3 108 (4,2) (2,1)
2P 2P
3
4 4
3
1P1 1P
22
12 3 41 1 2 3 4
118,144 120,100,136
5 (1,3) (3,4)
16 7 1
3 4 1 2 117 (4,1) (2,2)
6 (1,3) (3,2)
18 6 0 2.5 2.5 0 3 134 (4,1) (2,4)
7 (1,2) (3,3)
7 8 2 Sym
4 4
0 1 NA (4,4) (2,1) 3 3
8 (1,2) (3,1)
9 8 1
4 4 1 1 113 (4,4) (2,3)
9 (1,1) (3,2)
5 7 1
3 2 1 2 105 (4,3) (2,4)
10 (1,1) (3,4)
2 7 2 Sym
3 4
0 2 NA (4,3) (2,2) 4 3
11 (1,4) (2,3)
24 5 1
3 2 2 4 120 (3,2) (4,1)
12 (1,4) (2,2)
22 6 1 Sym 3 3 2 3 NA (3,3) (4,1)
13 (1,4) (2,1)
19 7 1
4 3 1 2 128 (3,2) (4,3)
14 (1,3) (2,4)
17 6 1
4 2 2 2 112 (3,1) (4,2)
15 (1,3) (2,2)
15 8 1
4 4 1 1 131 (3,1) (4,4)
16 (1,2) (2,3)
10 7 1
3 4 1 2 137 (3,4) (4,1)
34
17 (1,2) (2,4)
11 7 1
4 3 2 2 86 (3,1) (4,3)
18 (1,2) (2,1)
8 7 1
3 4 2 2 109 (3,4) (4,3)
19 (1,1) (2,3)
3 8 1 Sym 4 4 2 1 NA (3,2) (4,4)
20 (1,1) (2,2)
1 8 1
4 4 2 1 102 (3,3) (4,4)
21 (1,1) (2,4)
6 6 1
29
3 3 1 3 123 (3,3) (4,2)
22 (1,4) (2,3)
23 6 1
4 2 2 3 114 (4,2) (3,1)
Game#
Payoff
Matrix
Shape
Joint
Max PSNEs Symmetric
NashPayoff
Dom.
Pareto
Optima Transpose Row Col.
23 (1,4) (2,2)
21 7 1
4 3 2 2 87 (4,3) (3,1)
24 (1,4) (2,1)
20 6 1
3 3 1 3 127 (4,2) (3,3)
25 (1,3) (2,4)
18 6 1
3 2 2 3 118 (4,1) (3,2)
26 (1,3) (2,2)
15 8 1 Sym 4 4 2 1 NA (4,4) (3,1)
27 (1,3) (2,2)
16 7 1
3 4 1 2 135 (4,1) (3,4)
28 (1,3) (2,1)
13 8 1
4 4 2 1 83 (4,4) (3,2)
29 (1,2) (2,3)
9 8 1
4 4 1 1 132 (4,4) (3,1)
30 (1,2) (2,3)
10 7 1
3 4 2 2 98 (4,1) (3,4)
31 (1,2) (2,4)
12 6 1
3 3 2 3 89 (4,1) (3,3)
32 (1,2) (2,1)
7 8 1
4 4 2 1 107 (4,4) (3,3)
33 (1,1) (2,3)
3 7 1
3 4 2 2 81 (4,2) (3,4)
34 (1,1) (2,2)
2 7 1 3 4 2 2 104 (4,3) (3,4)
35 (1,1) (2,4)
6 6 1 Sym 3 3 2 3 NA (4,2) (3,3)
36 (1,1) (2,4)
5 7 1
4 3 1 2 125 (4,3) (3,2)
37 (1,4) (4,3)
21 7 1
2 2 1 2 116 (2,2) (3,1)
38 (1,4) (4,2)
23 6 1
2 3 1 3 88 (2,3) (3,1)
39 (1,4) (4,1)
22 6 0
2.5 2.5 0 3 138 (2,2) (3,3)
40 (1,4) (4,1)
24 5 1 2 3 1 4 100 (2,3) (3,2)
41 (1,3) (4,4)
15 8 2 Sym
4 4
0 1 NA (2,2) (3,1) 2 2
42 (1,3) (4,4)
13 8 1
4 4 1 1 106 (2,1) (3,2)
43 (1,3) (4,2)
17 6 1
2 4 1 2 97 (2,4) (3,1)
44 (1,3) (4,2)
14 7 0
2.5 2.5 0 2 126 (2,1) (3,4)
30
Game#
Payoff
Matrix
Shape
Joint
Max PSNEs Symmetric
NashPayoff
Dom.
Pareto
Optima Transpose Row Col.
45 (1,3) (4,1)
18 6 1
2 4 1 3 91 (2,4) (3,2)
46 (1,2) (4,3)
11 7 2 4 3
0 2 133 (2,4) (3,1) 2 4
47 (1,2) (4,3)
8 7 1
4 3 1 2 94 (2,1) (3,4)
48 (1,2) (4,4)
7 8 1
4 4 1 1 82 (2,1) (3,3)
49 (1,2) (4,1)
12 6 1
2 4 1 3 121 (2,4) (3,3)
50 (1,2) (4,1)
10 7 0
2.5 2.5 0 2 143 (2,3) (3,4)
51 (1,1) (4,3)
2 7 1
4 3 1 2 79 (2,2) (3,4)
52 (1,1) (4,2)
4 7 1
4 2 1 2 101 (2,3) (3,4)
53 (1,1) (4,2)
6 6 2 Sym
4 2
0 3 NA (2,4) (3,3) 2 4
54 (1,1) (4,4)
1 8 1
4 4 1 1 92 (2,2) (3,3)
55 (1,1) (4,4)
3 8 2 4 4
0 1 122 (2,3) (3,2) 2 3
56 (1,4) (4,3)
19 7 1
3 2 1 2 110 (3,2) (2,1)
57 (1,4) (4,3)
21 7 0
2.5 2.5 0 2 141 (3,1) (2,2)
58 (1,4) (4,2)
20 6 1
3 3 1 3 84 (3,3) (2,1)
59 (1,4) (4,1)
24 5 0
2.5 2.5 0 4 136 (3,2) (2,3)
60 (1,4) (4,1)
22 6 1
3 3 1 3 99 (3,3) (2,2)
61 (1,3) (4,4)
13 8 2 4 4
0 1 140 (3,2) (2,1) 3 2
62 (1,3) (4,4)
15 8 1
4 4 1 1 111 (3,1) (2,2)
63 (1,3) (4,2)
14 7 1
3 4 1 2 95 (3,4) (2,1)
64 (1,3) (4,2)
17 6 0
2.5 2.5 0 2 130 (3,1) (2,4)
65 (1,3) (4,1)
18 6 0
2.5 2.5 0 3 144 (3,2) (2,4)
66 (1,3) (4,1)
16 7 1
3 4 1 2 90 (3,4) (2,2)
31
Game#
Payoff
Matrix
Shape
Joint
Max PSNEs Symmetric
NashPayoff
Dom.
Pareto
Optima Transpose Row Col.
67 (1,2) (4,3)
8 7 2 4 3
0 2 129 (3,4) (2,1) 3 4
68 (1,2) (4,3)
11 7 1
4 3 1 2 96 (3,1) (2,4)
69 (1,2) (4,4)
7 8 2 Sym
4 4
0 1 NA (3,3) (2,1) 3 3
70 (1,2) (4,4)
9 8 1
4 4 1 1 85 (3,1) (2,3)
71 (1,2) (4,1)
10 7 1
3 4 1 2 119 (3,4) (2,3)
72 (1,2) (4,1)
12 6 0
2.5 2.5 0 3 142 (3,3) (2,4)
73 (1,1) (4,3)
5 7 1
4 3 1 2 80 (3,2) (2,4)
74 (1,1) (4,3)
2 7 2 Sym
4 3
0 2 NA (3,4) (2,2) 3 4
75 (1,1) (4,2)
6 6 1
4 2 1 3 103 (3,3) (2,4)
76 (1,1) (4,2)
4 7 2 3 4
0 2 139 (3,4) (2,3) 4 2
77 (1,1) (4,4)
3 8 1
4 4 1 1 93 (3,2) (2,3)
78 (1,1) (4,4)
1 8 2 4 4
0 1 124 (3,3) (2,2) 3 3
79 (1,1) (2,2)
2 7 1
3 4 1 2 51 (3,4) (4,3)
80 (1,1) (2,3)
4 7 1
3 4 1 2 73 (3,4) (4,2)
81 (1,1) (2,4)
5 7 1
4 3 2 2 33 (3,2) (4,3)
82 (1,2) (2,1)
7 8 1
4 4 1 1 48 (3,3) (4,4)
83 (1,2) (2,3)
9 8 1 4 4 2 1 28 (3,1) (4,4)
84 (1,2) (2,4)
12 6 1
3 3 1 3 58 (3,3) (4,1)
85 (1,3) (2,1)
13 8 1
4 4 1 1 70 (3,2) (4,4)
86 (1,3) (2,1)
14 7 1
3 4 2 2 17 (3,4) (4,2)
87 (1,3) (2,2)
16 7 1
3 4 2 2 23 (3,4) (4,1)
88 (1,3) (2,4)
18 6 1
3 2 1 3 38 (3,2) (4,1)
32
Game#
Payoff
Matrix
Shape
Joint
Max PSNEs Symmetric
NashPayoff
Dom.
Pareto
Optima Transpose Row Col.
89 (1,4) (2,1)
20 6 1
3 3 2 3 31 (3,3) (4,2)
90 (1,4) (2,2)
21 7 1
4 3 1 2 66 (3,1) (4,3)
91 (1,4) (2,3)
23 6 1
4 2 1 3 45 (3,1) (4,2)
92 (1,1) (2,2)
1 8 1
4 4 1 1 54 (4,4) (3,3)
93 (1,1) (2,3)
3 8 1
4 4 1 1 77 (4,4) (3,2)
94 (1,2) (2,1)
6 7 1
3 4 1 2 47 (4,3) (3,4)
95 (1,2) (2,4)
11 7 1
4 3 1 2 63 (4,3) (3,1)
96 (1,3) (2,1)
14 7 1
3 4 1 2 68 (4,2) (3,4)
97 (1,3) (2,4)
17 6 1 4 2 1 2 43 (4,2) (3,1)
98 (1,4) (2,1)
19 7 1
4 3 2 2 30 (4,3) (3,2)
99 (1,4) (2,2)
22 6 1
3 3 1 3 60 (4,1) (3,3)
100 (1,4) (2,3)
24 5 1
3 2 1 4 40 (4,1) (3,2)
101 (1,1) (3,2)
5 7 1
2 4 1 2 52 (2,4) (4,3)
102 (1,1) (3,3)
1 8 1
4 4 2 1 20 (2,2) (4,4)
103 (1,1) (3,3)
6 6 1
2 4 1 3 75 (2,4) (4,2)
104 (1,1) (3,4)
2 7 1
4 3 2 2 34 (2,2) (4,3)
105 (1,1) (3,4)
4 7 1
2 3 1 2 9 (2,3) (4,2)
106 (1,2) (3,1)
9 8 1
4 4 1 1 42 (2,3) (4,4)
107 (1,2) (3,3)
7 8 1
4 4 2 1 32 (2,1) (4,4)
108 (1,2) (3,3)
12 6 1
2 4 1 3 4 (2,4) (4,1)
109 (1,2) (3,4)
8 7 1
4 3 2 2 18 (2,1) (4,3)
110 (1,2) (3,4)
10 7 1
2 3 1 2 56 (2,3) (4,1)
33
Game#
Payoff
Matrix
Shape
Joint
Max PSNEs Symmetric
NashPayoff
Dom.
Pareto
Optima Transpose Row Col.
111 (1,3) (3,1)
15 8 1
4 4 1 1 62 (2,2) (4,4)
112 (1,3) (3,1)
17 6 1
2 4 2 2 14 (2,4) (4,2)
113 (1,3) (3,2)
13 8 1
4 4 1 1 8 (2,1) (4,4)
114 (1,3) (3,2)
18 6 1
2 4 2 3 22 (2,4) (4,1)
115 (1,3) (3,4)
14 7 1
4 2 2 2 2 (2,1) (4,2)
116 (1,3) (3,4)
16 7 1
2 2 1 2 37 (2,2) (4,1)
117 (1,4) (3,1)
21 7 1
4 3 1 2 5 (2,2) (4,3)
118 (1,4) (3,1)
23 6 1
2 3 2 3 25 (2,3) (4,2)
119 (1,4) (3,2)
19 7 1
4 3 1 2 71 (2,1) (4,3)
120 (1,4) (3,2)
24 5 1
2 3 2 4 11 (2,3) (4,1)
121 (1,4) (3,3)
20 6 1
4 2 1 3 49 (2,1) (4,2)
122 (1,1) (3,2)
3 8 2 4 4
0 1 55 (4,4) (2,3) 3 2
123 (1,1) (3,3)
6 6 1
3 3 1 3 21 (4,2) (2,4)
124 (1,1) (3,3)
1 8 2 4 4
0 1 78 (4,4) (2,2) 3 3
125 (1,1) (3,4)
4 7 1
3 4 1 2 36 (4,2) (2,3)
126 (1,2) (3,1)
11 7 0
2.5 2.5 0 2 44 (4,3) (2,4)
127 (1,2) (3,3)
12 6 1 3 3 1 3 24 (4,1) (2,4)
128 (1,2) (3,4)
10 7 1
3 4 1 2 13 (4,1) (2,3)
129 (1,2) (3,4)
8 7 2 4 3
0 2 67 (4,3) (2,1) 3 4
130 (1,3) (3,1)
17 6 0
2.5 2.5 0 2 64 (4,2) (2,4)
131 (1,3) (3,1)
15 8 1
4 4 1 1 15 (4,4) (2,2)
132 (1,3) (3,2)
13 8 1
4 4 1 1 29 (4,4) (2,1)
34
45
Game#
Payoff
Matrix
Shape
Joint
Max PSNEs Symmetric
NashPayoff
Dom.
Pareto
Optima Transpose Row Col.
133 (1,3) (3,4)
14 7 2 3 4
0 2 46 (4,2) (2,1) 4 2
134 (1,4) (3,1)
23 6 0
2.5 2.5 0 3 6 (4,2) (2,3)
135 (1,4) (3,1)
21 7 1
4 3 1 2 27 (4,3) (2,2)
136 (1,4) (3,2)
24 5 0
2.5 2.5 0 4 59 (4,1) (2,3)
137 (1,4) (3,2)
19 7 1
4 3 1 2 16 (4,3) (2,1)
138 (1,4) (3,3)
22 6 0
2.5 2.5 0 3 39 (4,1) (2,2)
139 (1,1) (4,3)
5 7 2 4 3
0 2 76 (2,4) (3,2) 2 4
140 (1,2) (4,4)
9 8 2 4 4
0 1 61 (2,3) (3,1) 2 3
141 (1,3) (4,1)
16 7 0
2.5 2.5 0 2 57 (2,2) (3,4)
142 (1,4) (4,2)
20 6 0
2.5 2.5 0 3 72 (2,1) (3,3)
143 (1,4) (4,3)
19 7 0
2.5 2.5 0 2 50 (2,1) (3,2)
144 (1,4) (4,2)
23 6 0
2.5 2.5 0 3 65 (3,1) (2,3)
Game#correspondstothenumberingsystemestablishedinthecompanionpaper
PayoffMatrixgivesthenormalformofeachgamewithpayoffslistedas(rowpayoff,columnpayoff)
Shapecorrespondstotheshapeofthepayoffset'sconvexhullasshowninAppendix1
JointMaxgivesthehighestpossiblecombinedpayoffforthetwoplayers
Symmetricismarked"Sym"ifthegameissymmetric,otherwiseitisleftblank
NashPayoffliststhepayoffsofthenoncooperativeequilibrium
iftherearetwoequilibriawithdifferentpayoffsums,theonewiththehighestsumislistedfirst
Dom.specifiesthenumberofrowandcolumnstrategiesthatarestrictlydominated
ParetoOptimagivesthenumberofpayoffpairsthatareParetooptimal
Transposeliststhegamenumbercorrespondingtothetransposeofthegameshown
Appendix3
#include <iostream>#include <fstream>#include <stdio.h>#include <time.h>#include <stdlib.h>
#include <iomanip>
using namespace
std;
const int maxnum=1<<11;
const double di=1<<9;
const int loop=300000;
class cmatr
47
{ public: int aul,bul,aur,bur,adl,bdl,adr,bdr; public: int
d1,d2,d3,d4,d5,d6,d7,d8; public: int nash; public: double
na1,na2,na3,na4; public: void setnum(int a1,int a2,int a3,int a4,int
a5,int a6,int a7,int a8){
aul=a1;bul=a2;aur=a3;bur=a4;
adl=a5;bdl=a6;adr=a7;bdr=a8;
d1=a1;d2=a2;d3=a3;d4=a4;
d5=a5;d6=a6;d7=a7;d8=a8;}public: void print(){
cout<<"cm"<<endl<<"aul "<<aul<<" bul "<<bul<<" aur "<<aur<<" bur
"<<bur<<endl; cout<<"adl "<<adl<<" bdl "<<bdl<<" adr "<<adr<<" bdr
"<<bdr<<endl; cout<<d1<<' '<<d2<<' '<<d3<<' '<<d4<<endl<<d5<<' '<<d6<<'
'<<d7<<' '<<d8<<endl;}public: bool eql(cmatr a){
for (int i=1;i<=4;i++){
if ((aul==a.aul)&&(bul==a.bul)&&(aur==a.aur)&&(bur==a.bur)&&(adl==a.adl)&&(bdl==a.bdl)&&(adr==a.adr)&&(bdr==a.bdr)) return true;
a.setnum(a.adr,a.bdr,a.aul,a.bul,a.aur,a.bur,a.adl,a.bdl)
; }return false;
}public: void rep(){
aul=1;aur=1;bul=1;bur=1;adl=1;adr=1;bdl=1;bdr=1;36
if (d1>d3) aul++; if (d1>d5) aul++; if (d1>d7) aul++;if (d3>d1) aur++; if (d3>d5) aur++; if (d3>d7) aur++;
if (d5>d1) adl++; if (d5>d3) adl++; if (d5>d7) adl++;
if (d7>d1) adr++; if (d7>d3) adr++; if (d7>d5) adr++;
if (d2>d4) bul++; if (d2>d6) bul++; if (d2>d8) bul++;
if (d4>d2) bur++; if (d4>d6) bur++; if (d4>d8)
bur++; if (d6>d2) bdl++; if (d6>d4) bdl++; if (d6>d8)
bdl++; if (d8>d2) bdr++; if (d8>d4) bdr++; if (d8>d6)
bdr++;}
};class matr{ public: int ul,ur,dl,dr;
public: int
d1,d2,d3,d4; public:
void randnum(){
ul=rand()%maxnum+1;d1=ul;
ur=rand()%maxnum+1;d2=ur;
dl=rand()%maxnum+1;d3=dl;
dr=rand()%maxnum+1;d4=dr;}public: bool eql(){
if ((ul==ur)||(ul==dl)||(ul==dr)) return
true; if ((ur==dl)||(ur==dr)) return true; if
(dl==dr) return true; return false;}public: void rep(){
int a1=1,a2=1,a3=1,a4=1; if (ul>ur) a1++; if
(ul>dl) a1++; if (ul>dr) a1++; if (ur>ul) a2++; if
(ur>dl) a2++; if (ur>dr) a2++; if (dl>ul) a3++;
if (dl>ur) a3++; if (dl>dr) a3++; if (dr>ul)
a4++; if (dr>ur) a4++; if (dr>dl) a4++;
ul=a1;ur=a2;dl=a3;dr=a4;}public: cmatr combine(matr b){
cmatr c;c.setnum(ul,b.ul,ur,b.ur,dl,b.dl,dr,b.dr);
return c;}public: void print(){
49
cout<<ul<<' '<<ur<<' '<<dl<<' '<<dr<<endl;}
};double a[500000],b[500000],a1[500000],b1[500000]; double
oa[150][4000],ob[150][4000],oa1[150][4000],ob1[150][4000]; int
soa[150],sob[150],soa1[150],sob1[150]; int ty[500000]; char
nu[4]=""; string s="out"; string s2="solution"; string s1; int m=0;int main(){ srand((int)time(0)); int aul,bul,aur,bur,adl,bdl,adr,bdr; int i,j,k,t,tp; int res[150]; double
mr[150],mc[150],vr[150],vc[150],sr[150],sc[150],ms[150],vs[150],ss[150]; double
mr1[150],mc1[150],vr1[150],vc1[150],sr1[150],sc1[150],ms1[150],vs1[150],ss1[150]; double
mmr,mmc,vvr,vvc,ssr,ssc,mms,vvs,sss; double
mmr1,mmc1,vvr1,vvc1,ssr1,ssc1,mms1,vvs1,sss1; int rc[150],rc1[150]; cmatr samp[150];
cmatr c,d; cmatr *p; matr pa,pb;
FILE * fp=NULL;
fp=fopen("www.txt","r");
for (i=1;i<=144;i++){
fscanf(fp,"%d%d%d%d%d%d%d%d",&aul,&bul,&aur,&bur,&adl,&bdl,&adr,&bdr
); samp[i].setnum(aul,bul,aur,bur,adl,bdl,adr,bdr); res[i]=0;}for (i=1;i<=144;i++){
fscanf(fp,"%d",&samp[i].nash); if
((samp[i].nash==1)||(samp[i].nash==0))
fscanf(fp,"%lf%lf",&samp[i].na1,&samp[i].na2);if (samp[i].nash==2)
fscanf(fp,"%lf%lf%lf%lf",&samp[i].na1,&samp[i].na2,&samp[i].na3,&samp[i].na4);} fclose(fp);
fp=NULL; for
(i=1;i<=loop;i++){
pa.randnum(); while (pa.eql())
pa.randnum(); pb.randnum(); while
(pb.eql()) pb.randnum();
c=pa.combine(pb); c.rep();
a1[i]=(double)
(c.d1+c.d3+c.d5+c.d7)/4;
b1[i]=(double)
(c.d2+c.d4+c.d6+c.d8)/4; for
(k=1;k<=144;k++){
if (samp[k].eql(c))
50
{ res[k]++; ty[i]=k; a[i]=0;
b[i]=0; if
(samp[k].nash==0){
a[i]=c.d1+c.d3+c.d5+c.d7;
a[i]=a[i]/4;
b[i]=c.d2+c.d4+c.d6+c.d8;
b[i]=b[i]/4;}else if (samp[k].nash==1){
if ((c.aul==samp[k].na1)&&(c.bul==samp[k].na2)){ a[i]=c.d1;b[i]=c.d2;};if ((c.aur==samp[k].na1)&&(c.bur==samp[k].na2)){ a[i]=c.d3;b[i]=c.d4;};if ((c.adl==samp[k].na1)&&(c.bdl==samp[k].na2)){ a[i]=c.d5;b[i]=c.d6;};if ((c.adr==samp[k].na1)&&(c.bdr==samp[k].na2)){ a[i]=c.d7;b[i]=c.d8;};
}else if (samp[k].nash==2){
if ((c.aul==samp[k].na1)&&(c.bul==samp[k].na2)){ a[i]=c.d1;b[i]=c.d2;};if ((c.aur==samp[k].na1)&&(c.bur==samp[k].na2)){ a[i]=c.d3;b[i]=c.d4;};if ((c.adl==samp[k].na1)&&(c.bdl==samp[k].na2)){ a[i]=c.d5;b[i]=c.d6;};if ((c.adr==samp[k].na1)&&(c.bdr==samp[k].na2)){ a[i]=c.d7;b[i]=c.d8;};if ((c.aul==samp[k].na3)&&(c.bul==samp[k].na4)){ a[i]+=c.d1;b[i]+=c.d2;};if ((c.aur==samp[k].na3)&&(c.bur==samp[k].na4)){ a[i]+=c.d3;b[i]+=c.d4;};if ((c.adl==samp[k].na3)&&(c.bdl==samp[k].na4)){ a[i]+=c.d5;b[i]+=c.d6;
51
};if ((c.adr==samp[k].na3)&&(c.bdr==samp[k].na4)){ a[i]+=c.d7;b[i]+=c.d8;};a[i]+=(double) (c.d1+c.d3+c.d5+c.d7)/4;
b[i]+=(double)
(c.d2+c.d4+c.d6+c.d8)/4;
a[i]=(double)a[i]/3; b[i]=(double)b[i]/3;};break;
}}
}for (i=1;i<=144;i++){ mr[i]=0;mc[i]=0;vr[i]=0;vc[i]=0;sr[i]=0;sc[i]=0; rc[i]=0;
mr1[i]=0;mc1[i]=0;vr1[i]=0;vc1[i]=0;sr1[i]=0;sc1[i]=0
; rc1[i]=0; soa[i]=0;sob[i]=0; soa1[i]=0;sob1[i]=0;}mmr=0;mmc=0;vvr=0;vvc=0;ssr=0;ssc=0;mms=0;vvs=0;sss=0;
mmr1=0;mmc1=0;vvr1=0;vvc1=0;ssr1=0;ssc1=0;mms1=0;vvs1=0;sss1=0;
for (i=1;i<=loop;i++){ a[i]=a[i]/di;
b[i]=b[i]/di;
a1[i]=a1[i]/di;
b1[i]=b1[i]/di
;}for (i=1;i<=loop;i++){ tp=ty[i]; rc[tp]++; rc1[tp]++;
sr[tp]+=a[i];sr1[tp]+=a1[i];
sc[tp]+=b[i];sc1[tp]+=b1[i];
ss[tp]+=a[i]+b[i];ss1[tp]+=a1[i]+b1[i];
ssr+=a[i];ssc+=b[i];sss+=a[i]+b[i];
ssr1+=a1[i];ssc1+=b1[i];sss1+=a1[i]+b1[i]
; soa[tp]++; oa[tp][soa[tp]]=a[i];
sob[tp]++; ob[tp][sob[tp]]=b[i];
soa1[tp]++; oa1[tp][soa1[tp]]=a1[i];
sob1[tp]++; ob1[tp][sob1[tp]]=b1[i];}for (i=1;i<=144;i++){ mr[i]=sr[i]/rc[i];mr1[i]=sr1[i]/rc1[i];
mc[i]=sc[i]/rc[i];mc1[i]=sc1[i]/rc1[i];
ms[i]=ss[i]/rc[i];ms1[i]=ss1[i]/rc1[i];
52
sr[i]=0;sc[i]=0;ss[i]=0;sr1[i]=0;sc1[i]=0;ss1[i]=0
;}mmr=ssr/loop;mmc=ssc/loop;mms=sss/loop;
mmr1=ssr1/loop;mmc1=ssc1/loop;mms1=sss1/loop
; ssr=0;ssc=0;sss=0; ssr1=0;ssc1=0;sss1=0; for
(i=1;i<=loop;i++){ tp=ty[i]; sr[tp]+=(a[i]‐mr[tp])*(a[i]‐mr[tp]);sr1[tp]+=(a1[i]‐mr1[tp])*(a1[i]‐
mr1[tp]); sc[tp]+=(b[i]‐mc[tp])*(b[i]‐mc[tp]);sc1[tp]+=(b1[i]‐
mc1[tp])*(b1[i]‐mc1[tp]);
ss[tp]+=(a[i]+b[i]‐ms[tp])*(a[i]+b[i]‐ms[tp]);ss1[tp]+=(a1[i]+b1[i]‐ms1[tp])*(a1[i]+b1[i]‐ms1[tp]);
ssr+=(a[i]‐mmr)*(a[i]‐mmr);ssr1+=(a1[i]‐mmr1)*(a1[i]‐mmr1); ssc+=(b[i]‐mmc)*(b[i]‐
mmc);ssc1+=(b1[i]‐mmc1)*(b1[i]‐mmc1); sss+=(a[i]+b[i]‐mms)*(a[i]+b[i]‐
mms);sss1+=(a1[i]+b1[i]‐mms1)*(a1[i]+b1[i]‐mms1);}vvr=ssr/loop;vvc=ssc/loop;vvs=sss/loop;
vvr1=ssr1/loop;vvc1=ssc1/loop;vvs1=sss1/loop;
for (i=1;i<=144;i++){ vr[i]=sr[i]/rc[i];vr1[i]=sr1[i]/rc1[i];
vc[i]=sc[i]/rc[i];vc1[i]=sc1[i]/rc1[i]
;
vs[i]=ss[i]/rc[i];vs1[i]=ss1[i]/rc1[i];} cout.setf(ios::fixed);
fp=fopen("out.txt","w")
;fprintf(fp,"r mean %.3lf c mean %.3lf sum mean %.3lf r var %.3lf c var %.3lf sum var
%.3lf\n",mmr,mmc,mms,vvr,vvc,vvs);fprintf(fp,"r mean %.3lf c mean %.3lf sum mean %.3lf r var %.3lf c var %.3lf sum var
%.3lf\n",mmr1,mmc1,mms1,vvr1,vvc1,vvs1); for (i=1;i<=144;i++){
//cout<<setprecision(3)<<"BOX "<<i<<":"<<"r mean:"<<mr[i]<<" c mean"<<mc[i]<<"
sum mean:"<<ms[i]<<" r var"<<vr[i]<<" c var"<<vc[i]<<" sum var:"<<vs[i]<<endl; fprintf(fp,"BOX
%d:r mean %.3lf c mean %.3lf sum mean %.3lf r var %.3lf c var %.3lfsum var %.3lf\n",i,mr[i],mc[i],ms[i],vr[i],vc[i],vs[i]);
fprintf(fp,"BOX %d:r mean %.3lf c mean %.3lf sum mean %.3lf r var %.3lf c var %.3lfsum var %.3lf\n",i,mr1[i],mc1[i],ms1[i],vr1[i],vc1[i],vs1[i]);
} fclose(fp); for
(i=1;i<=144;i++){
m++;itoa(m,nu,10); s1=s+nu+".txt";
fp=fopen(s1.c_str(),"w"); fprintf(fp,"BOX
53
%d\n",i); for (k=1;k<=soa[i];k++)
fprintf(fp,"%.3lf %.3lf\n",oa[i][k],ob[i][k]);fclose(fp);
}m=0;for (i=1;i<=144;i++){
m++;itoa(m,nu,10); s1=s2+nu+".txt";
fp=fopen(s1.c_str(),"w"); fprintf(fp,"BOX %d\n",i);
for (k=1;k<=soa1[i];k++) fprintf(fp,"%.3lf
%.3lf\n",oa1[i][k],ob1[i][k]);fclose(fp);
}return 0;
}
Game# Type Rowmean Columnmean Summean Rowvar Columnvar Sumvar
1NE
Entropy1.5791.989
1.5771.995
3.1563.984
0.6580.350
0.6180.315
1.2970.684
2NE
Entropy1.6091.996
3.2021.994
4.8113.990
0.6230.324
0.4280.324
1.0620.662
3NE
Entropy3.1971.994
3.2282.011
6.4254.005
0.4290.330
0.4030.325
0.8200.667
4NE
Entropy3.1981.992
1.5851.996
4.7843.987
0.4150.319
0.6290.329
1.0440.647
5NE
Entropy2.3771.986
3.1842.002
5.5613.987
0.6390.321
0.4370.341
1.0390.654
6NE
Entropy1.9951.994
2.0011.999
3.9963.993
0.3360.336
0.3470.347
0.6690.669
7NE
Entropy2.5181.983
2.5361.993
5.0533.976
0.3700.327
0.3730.318
0.7410.645
8NE
Entropy3.2172.023
3.2242.011
6.4414.034
0.4220.333
0.4050.332
0.8480.678
9NE
Entropy2.3911.998
1.6032.011
3.9944.009
0.6580.347
0.6130.314
1.2440.641
10NE
Entropy2.5431.999
2.5382.001
5.0814.001
0.3770.327
0.3660.320
0.7320.630
11NE
Entropy2.3982.004
1.6052.000
4.0034.004
0.6310.335
0.6440.328
1.2600.666
12NE
Entropy2.4092.004
2.3991.994
4.8083.999
0.6470.340
0.6390.338
1.2930.673
13NE
Entropy3.1922.007
2.4102.012
5.6014.019
0.4150.324
0.6340.336
1.0510.669
14NE
Entropy3.2001.998
1.6061.995
4.8073.993
0.4260.329
0.6480.328
1.0580.668
15NE
Entropy3.1922.010
3.2011.986
6.3933.996
0.4300.342
0.4200.319
0.8660.666
16NE
Entropy2.3901.989
3.2342.014
5.6244.003
0.6550.335
0.4140.331
1.0340.664
17NE
Entropy3.2062.014
2.4202.015
5.6254.029
0.4350.341
0.6070.320
1.0480.660
18NE
Entropy2.3781.986
3.2172.013
5.5953.998
0.6410.341
0.4240.334
1.0360.663
Appendix4
Table14:Theoutputfork=9forthe144games
19NE
Entropy3.1701.970
3.1991.999
6.3693.969
0.4400.325
0.4320.327
0.8680.643
20NE
Entropy3.2092.005
3.1962.008
6.4054.013
0.4350.332
0.4160.312
0.8410.638
21NE
Entropy2.4042.004
2.392 2.004 44
4.7964.008
0.6200.337
0.6410.341
1.3030.695
22NE
Entropy3.1991.990
1.5991.995
4.7983.985
0.4190.326
0.6450.342
1.0500.652
23NE
Entropy3.2362.025
2.3891.985
5.6254.010
0.4340.329
0.6490.328
1.0900.653
Game# Type Rowmean Columnmean Summean Rowvar Columnvar Sumvar
26NE
Entropy3.1921.995
3.1841.989
6.3753.984
0.4430.335
0.4300.340
0.8480.675
27NE
Entropy2.3951.988
3.2042.000
5.5993.988
0.6480.329
0.4430.356
1.1010.673
28NE
Entropy3.2081.999
3.2232.017
6.4314.016
0.4250.324
0.4240.343
0.8400.692
29NE
Entropy3.2152.010
3.2032.002
6.4174.012
0.4180.331
0.4200.328
0.8360.634
30NE
Entropy2.3791.987
3.2072.006
5.5863.993
0.6440.337
0.4070.333
1.0550.666
31NE
Entropy2.4122.010
2.3761.988
4.7883.997
0.6430.331
0.6300.326
1.3020.674
32NE
Entropy3.1921.984
3.1981.991
6.3903.975
0.4200.319
0.4170.332
0.8220.637
33NE
Entropy2.3861.991
3.2061.995
5.5923.987
0.6290.326
0.4200.303
1.1120.636
34NE
Entropy2.4132.015
3.2092.010
5.6214.024
0.6280.333
0.4050.321
1.0420.630
35NE
Entropy2.4021.989
2.4302.011
4.8324.000
0.6180.320
0.6480.324
1.2830.642
36NE
Entropy3.1931.993
2.4182.003
5.6113.996
0.4390.351
0.6440.331
1.0850.701
37NE
Entropy1.5931.997
1.6162.017
3.2094.015
0.6200.326
0.6460.323
1.2620.661
38NE
Entropy1.5801.988
2.4062.004
3.9863.992
0.6230.318
0.6390.325
1.2180.619
39NE
Entropy1.9991.997
2.0202.019
4.0194.016
0.3260.326
0.3340.334
0.6630.663
40NE
Entropy1.6042.000
2.4122.006
4.0164.005
0.6630.347
0.6470.336
1.3090.672
41NE
Entropy2.2782.008
2.2681.998
4.5464.005
0.3630.346
0.3320.315
0.6950.655
42NE
Entropy3.2152.011
3.2131.988
6.4284.000
0.4130.326
0.4140.319
0.8530.668
43NE
Entropy1.6202.005
3.2041.994
4.8243.999
0.6120.334
0.4290.333
1.0630.670
44NE
Entropy2.0292.028
2.0172.016
4.0464.044
0.3270.327
0.3310.331
0.6450.645
45NE
Entropy1.5871.995
3.2182.013
4.8054.008
0.6340.328
0.4060.314
1.0450.645
46NE
Entropy2.2692.003
2.5562.023
4.8254.025
0.3490.331
0.3720.339
0.7150.669
47NE
Entropy3.2021.998
2.4161.991
5.6183.989
0.4450.339
0.6050.313
1.0400.624
48NE
Entropy3.1961.983
3.193 2.007 45
6.3893.989
0.4170.327
0.4180.325
0.8090.618
49NE
Entropy1.5921.992
3.1931.979
4.7853.971
0.6380.338
0.4400.331
1.0700.663
50NE
Entropy1.9891.987
2.0042.002
3.9923.990
0.3320.332
0.3350.335
0.6720.672
Game# Type Rowmean Columnmean Summean Rowvar Columnvar Sumvar
51NE
Entropy3.2072.000
2.3731.991
5.5793.991
0.4060.317
0.6260.328
1.0820.657
52NE
Entropy3.2012.011
1.5801.995
4.7814.005
0.4190.333
0.6380.332
1.0900.674
53NE
Entropy2.2782.009
2.2501.982
4.5283.991
0.3400.325
0.3440.323
0.6630.635
54NE
Entropy3.2022.007
3.1772.006
6.3804.013
0.4370.331
0.4390.339
0.8890.664
55NE
Entropy2.2531.985
2.5211.994
4.7743.980
0.3650.345
0.3970.347
0.7390.670
56NE
Entropy2.4332.024
1.6002.002
4.0344.026
0.6230.342
0.6330.325
1.3080.696
57NE
Entropy1.9891.988
2.0092.008
3.9993.996
0.3630.363
0.3300.330
0.6810.681
58NE
Entropy2.4112.002
2.4312.024
4.8414.026
0.6200.310
0.6550.346
1.3150.668
59NE
Entropy2.0082.006
2.0152.014
4.0234.020
0.3130.313
0.3290.329
0.6560.656
60NE
Entropy2.3781.992
2.4101.995
4.7883.988
0.6390.328
0.6610.344
1.2770.677
61NE
Entropy2.5271.991
2.2792.010
4.8064.001
0.3690.330
0.3630.341
0.7240.659
62NE
Entropy3.1871.981
3.2101.992
6.3973.973
0.4320.324
0.4140.323
0.8320.646
63NE
Entropy2.3591.972
3.2102.006
5.5693.978
0.6330.328
0.4050.330
1.0280.660
64NE
Entropy2.0222.020
2.0072.005
4.0284.025
0.3400.340
0.3440.344
0.6690.669
65NE
Entropy1.9941.992
1.9861.985
3.9803.977
0.3150.315
0.3540.354
0.6430.643
66NE
Entropy2.4152.016
3.1781.982
5.5943.998
0.6440.340
0.4350.338
1.0730.692
67NE
Entropy2.5572.015
2.5291.999
5.0864.014
0.3710.331
0.3790.340
0.7630.685
68NE
Entropy3.1931.994
2.4202.008
5.6144.001
0.4340.335
0.6040.316
1.0570.655
69NE
Entropy2.5221.984
2.5622.021
5.0844.005
0.3790.329
0.3560.317
0.7340.657
70NE
Entropy3.1951.989
3.2142.000
6.4093.989
0.4350.331
0.4250.324
0.8850.658
71NE
Entropy2.4272.018
3.2042.000
5.6314.018
0.6330.336
0.4000.313
1.0790.656
72NE
Entropy2.0072.006
1.9971.996
4.0044.002
0.3400.340
0.3230.323
0.6380.639
73NE
Entropy3.1961.996
2.382 1.989 46
5.5783.984
0.4240.330
0.6590.342
1.0570.670
74NE
Entropy2.5362.003
2.5502.015
5.0864.017
0.3730.329
0.3610.328
0.7310.666
75NE
Entropy3.1891.983
1.6182.007
4.8073.990
0.4390.337
0.6250.316
1.0360.640
Game# Type Rowmean Columnmean Summean Rowvar Columnvar Sumvar
76NE
Entropy2.5322.001
2.2772.005
4.8094.006
0.3800.340
0.3400.321
0.7230.658
77NE
Entropy3.2122.002
3.1971.995
6.4093.996
0.4290.332
0.4260.332
0.8690.657
78NE
Entropy2.5041.968
2.5372.001
5.0413.969
0.3790.334
0.3780.332
0.7560.669
79NE
Entropy2.4272.011
3.1982.003
5.6254.014
0.6370.325
0.4130.319
1.0540.649
80NE
Entropy2.4122.025
3.2011.999
5.6134.024
0.6330.328
0.4220.333
1.0500.651
81NE
Entropy3.1961.986
2.4082.006
5.6053.992
0.4330.332
0.6370.337
1.0740.676
82NE
Entropy3.1902.009
3.1781.990
6.3683.999
0.4330.345
0.4540.344
0.8770.684
83NE
Entropy3.2132.021
3.2051.992
6.4194.013
0.4490.349
0.4260.326
0.8670.678
84NE
Entropy2.3982.001
2.3831.988
4.7803.988
0.6330.329
0.6420.339
1.3070.686
85NE
Entropy3.1881.994
3.1982.000
6.3863.994
0.4300.326
0.4330.327
0.8850.653
86NE
Entropy2.4372.022
3.1801.974
5.6173.997
0.6490.336
0.4500.337
1.0570.664
87NE
Entropy2.3992.003
3.2022.002
5.6024.005
0.6100.327
0.4240.333
1.0150.646
88NE
Entropy2.3871.994
1.6212.012
4.0084.006
0.6320.324
0.6360.333
1.2570.661
89NE
Entropy2.4172.011
2.4202.007
4.8374.019
0.6300.325
0.6490.340
1.3350.688
90NE
Entropy3.1941.988
2.3882.000
5.5823.988
0.4430.335
0.6330.338
1.1200.696
91NE
Entropy3.2112.003
1.6242.020
4.8354.023
0.4240.321
0.6440.333
1.0920.666
92NE
Entropy3.1781.979
3.1941.990
6.3723.969
0.4600.339
0.4190.322
0.9050.664
93NE
Entropy3.2112.000
3.2172.028
6.4284.028
0.4000.318
0.4270.335
0.8320.660
94NE
Entropy2.4001.993
3.1982.004
5.5983.997
0.6470.335
0.4250.341
1.0770.683
95NE
Entropy3.1861.980
2.4172.010
5.6023.991
0.4220.329
0.6510.342
1.0580.654
96NE
Entropy2.3931.985
3.1681.970
5.5613.956
0.6240.330
0.4430.341
1.0670.646
97NE
Entropy3.2092.007
1.5811.980
4.7903.988
0.4160.328
0.6220.322
1.0520.656
98NE
Entropy3.1932.002
2.404 2.007 47
5.5974.009
0.4220.336
0.6120.324
1.0240.643
99NE
Entropy2.3911.994
2.4232.018
4.8144.012
0.6510.337
0.6230.328
1.2600.660
100NE
Entropy2.3721.984
1.6021.993
3.9743.977
0.6630.344
0.6510.337
1.3080.677
Game# Type Rowmean Columnmean Summean Rowvar Columnvar Sumvar
101NE
Entropy1.5941.997
3.2061.995
4.8003.992
0.6330.331
0.4220.330
1.0380.676
102NE
Entropy3.2092.025
3.2012.011
6.4114.036
0.4510.350
0.4370.344
0.8960.685
103NE
Entropy1.6032.002
3.1761.998
4.7794.000
0.6580.346
0.4190.323
1.0840.661
104NE
Entropy3.2062.010
2.3721.989
5.5783.999
0.4250.327
0.6600.345
1.1160.677
105NE
Entropy1.5681.974
2.4012.022
3.9693.997
0.6170.320
0.6280.329
1.2850.660
106NE
Entropy3.2142.015
3.2152.009
6.4304.024
0.4220.338
0.4310.337
0.8530.681
107NE
Entropy3.1861.990
3.2282.019
6.4144.009
0.4310.329
0.4260.335
0.8600.696
108NE
Entropy1.6182.013
3.2061.993
4.8244.006
0.6360.338
0.4110.316
1.0310.665
109NE
Entropy3.2052.003
2.4052.012
5.6104.016
0.4270.332
0.6510.339
1.0530.649
110NE
Entropy1.5711.989
2.4031.991
3.9743.980
0.6710.351
0.6450.327
1.2680.652
111NE
Entropy3.2262.019
3.2051.987
6.4314.006
0.4100.325
0.4300.333
0.8340.656
112NE
Entropy1.6122.005
3.2152.021
4.8274.026
0.6330.333
0.4200.333
0.9950.637
113NE
Entropy3.1921.990
3.2061.994
6.3983.984
0.4360.335
0.4230.335
0.8270.664
114NE
Entropy1.6111.996
3.2092.006
4.8204.002
0.6410.349
0.4110.334
1.0400.677
115NE
Entropy3.2092.001
1.6302.027
4.8394.028
0.4200.325
0.6230.332
1.0410.654
116NE
Entropy1.6001.998
1.5971.991
3.1973.989
0.6510.336
0.6370.340
1.2520.664
117NE
Entropy3.2121.999
2.3882.000
5.6003.998
0.4000.316
0.6520.342
1.0890.676
118NE
Entropy1.6282.017
2.4092.002
4.0374.018
0.6390.344
0.6400.338
1.2780.700
119NE
Entropy3.1811.975
2.3891.996
5.5703.971
0.4520.344
0.6490.345
1.1150.688
120NE
Entropy1.6172.004
2.4052.007
4.0224.012
0.6730.341
0.6220.325
1.3010.664
121NE
Entropy3.1951.993
1.5531.968
4.7483.961
0.4340.323
0.5990.327
1.0780.662
122NE
Entropy2.5522.014
2.2541.990
4.8064.004
0.3820.344
0.3500.331
0.7190.658
123NE
Entropy2.3591.969
2.395 1.996 48
4.7543.965
0.6310.325
0.6490.336
1.2910.645
124NE
Entropy2.5322.002
2.5392.009
5.0714.011
0.3690.332
0.3780.335
0.7540.680
125NE
Entropy2.3841.987
3.1952.007
5.5793.994
0.6240.320
0.4220.335
1.0320.634
Game# Type Rowmean Columnmean Summean Rowvar Columnvar Sumvar
126NE
Entropy2.0032.002
2.0001.999
4.0034.000
0.3390.339
0.3350.335
0.6880.688
127NE
Entropy2.4082.011
2.4262.016
4.8344.027
0.6540.342
0.6410.332
1.2840.666
128NE
Entropy2.3831.970
3.2022.008
5.5853.978
0.6720.342
0.4100.330
1.0320.656
129NE
Entropy2.5372.001
2.5512.022
5.0884.023
0.3820.345
0.3750.335
0.7740.695
130NE
Entropy2.0022.000
2.0202.019
4.0224.019
0.3330.333
0.3370.337
0.6720.672
131NE
Entropy3.2182.017
3.2051.991
6.4234.008
0.4140.335
0.4260.325
0.8170.671
132NE
Entropy3.1891.999
3.2001.997
6.3893.996
0.4350.338
0.4280.332
0.8700.679
133NE
Entropy2.5452.007
2.2872.020
4.8324.027
0.3540.322
0.3480.327
0.7130.658
134NE
Entropy2.0212.020
1.9961.995
4.0184.015
0.3330.333
0.3280.328
0.6480.649
135NE
Entropy3.2252.014
2.3961.997
5.6214.011
0.4190.327
0.6420.337
1.0340.634
136NE
Entropy1.9991.998
1.9811.979
3.9803.977
0.3460.346
0.3210.321
0.6790.679
137NE
Entropy3.2122.002
2.3841.994
5.5963.995
0.4160.320
0.6450.343
1.0700.657
138NE
Entropy2.0042.002
1.9991.997
4.0023.999
0.3400.340
0.3340.334
0.7030.703
139NE
Entropy2.2712.003
2.5431.998
4.8144.001
0.3500.327
0.3760.333
0.7370.667
140NE
Entropy2.2852.017
2.5211.998
4.8054.015
0.3450.331
0.3720.330
0.7100.651
141NE
Entropy1.9771.975
1.9911.990
3.9683.965
0.3200.320
0.3440.344
0.6740.674
142NE
Entropy2.0072.005
2.0022.001
4.0094.006
0.3480.348
0.3230.323
0.6790.679
143NE
Entropy2.0112.009
2.0092.008
4.0204.017
0.3280.328
0.3270.327
0.6450.645
144NE
Entropy2.0142.013
2.0172.016
4.0314.028
0.3190.319
0.3290.329
0.6410.642
49