[FREE PDF sample] Combinatorial game theory: a special collection in honor of elwyn berlekamp, john

Page 1


CombinatorialGameTheory:ASpecialCollectionin HonorofElwynBerlekamp,JohnH.Conwayand RichardK.GuyRichardJ.Nowakowski

https://ebookmass.com/product/combinatorial-game-theory-aspecial-collection-in-honor-of-elwyn-berlekamp-john-hconway-and-richard-k-guy-richard-j-nowakowski/

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

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

Eric Hobsbawm: A Life in History Richard J. Evans

https://ebookmass.com/product/eric-hobsbawm-a-life-in-history-richardj-evans/

ebookmass.com

A History of Modern Africa 3rd Edition Richard J. Reid

https://ebookmass.com/product/a-history-of-modern-africa-3rd-editionrichard-j-reid/

ebookmass.com

Through The Looking Glass: John Cage and Avant-Garde Film

Richard H. Brown Jr

https://ebookmass.com/product/through-the-looking-glass-john-cage-andavant-garde-film-richard-h-brown-jr/

ebookmass.com

Why Privacy Matters Neil Richards

https://ebookmass.com/product/why-privacy-matters-neil-richards/

ebookmass.com

Amp It Up: Leading for Hypergrowth by Raising Expectations, Increasing Urgency, and Elevating Intensity

https://ebookmass.com/product/amp-it-up-leading-for-hypergrowth-byraising-expectations-increasing-urgency-and-elevating-intensity-frankslootman/

ebookmass.com

5 Steps to a 5: AP Physics C 2021 Greg Jacobs

https://ebookmass.com/product/5-steps-to-a-5-ap-physics-c-2021-gregjacobs/

ebookmass.com

Programando em C C++ "A Biblia" 1st Edition Lars Klander

https://ebookmass.com/product/programando-em-c-c-a-biblia-1st-editionlars-klander/

ebookmass.com

Soldiering in India 1764-1787: extracts form journals and letters left by Lt. Colonel Allen Macpherson and Lt. Colonel John Macpherson of the East India Compan y■s Service William Charles Macpherson

https://ebookmass.com/product/soldiering-in-india-1764-1787-extractsform-journals-and-letters-left-by-lt-colonel-allen-macpherson-and-ltcolonel-john-macpherson-of-the-east-india-compan-y%ca%bcs-servicewilliam-charles-macpher/ ebookmass.com

Burlar al Diablo Napoleón Hill

https://ebookmass.com/product/burlar-al-diablo-napoleon-hill/

ebookmass.com

The Third Revolution: XI Jinping and the New Chinese State Elizabeth Economy

https://ebookmass.com/product/the-third-revolution-xi-jinping-and-thenew-chinese-state-elizabeth-economy/

ebookmass.com

RichardJ.Nowakowski,BruceM.Landman,FlorianLuca,MelvynB.Nathanson, JaroslavNešetřil,andAaronRobertson(Eds.)

CombinatorialGameTheory

DeGruyterProceedingsin Mathematics

Combinatorial GameTheory

ASpecialCollectioninHonorofElwynBerlekamp,John H.ConwayandRichardK.Guy

RichardJ.Nowakowski,BruceM.Landman,FlorianLuca, MelvynB.Nathanson,JaroslavNešetřil,andAaron Robertson

MathematicsSubjectClassification2010

05A,05C55,05C65,05D,11A,11B,11D,11K,11N,11P,11Y,91A46

Editors

RichardJ.Nowakowski DalhousieUniversity Dept.ofMathematics&Statistics ChaseBuilding HalifaxNSB3H3J5 Canada r.nowakowski@dal.ca

BruceM.Landman UniversityofGeorgia DepartmentofMathematics Athens,GA30602 USA

Bruce.Landman@uga.edu

FlorianLuca UniversityofWitwatersrand SchoolofMathematics 1JanSmutsAvenue Johannesburg2000 RepublicofSouthAfrica Florian.Luca@wits.ac.za

MelvynB.Nathanson LehmanCollege(CUNY) DepartmentofMathematics 250BedfordParkBoulevardWest BronxNY10468 USA

Melvyn.Nathanson@lehman.cuny.edu

JaroslavNešetřil CharlesUniversity ComputerScienceInstitute(IUUK) Malostranskenam.25 11800Praha CzechRepublic Nesetril@iuuk.mff.cuni.cz

AaronRobertson ColgateUniversity DepartmentofMathematics 219McGregoryHall HamiltonNY13346 USA arobertson@colgate.edu

ISBN978-3-11-075534-3 e-ISBN(PDF)978-3-11-075541-1 e-ISBN(EPUB)978-3-11-075549-7

LibraryofCongressControlNumber:2022934372

BibliographicinformationpublishedbytheDeutscheNationalbibliothek TheDeutscheNationalbibliothekliststhispublicationintheDeutscheNationalbibliografie; detailedbibliographicdataareavailableontheInternetathttp://dnb.dnb.de.

©2022WalterdeGruyterGmbH,Berlin/Boston Typesetting:VTeXUAB,Lithuania Printingandbinding:CPIbooksGmbH,Leck

www.degruyter.com

Preface

Whatis1 + 1 + 1?

JohnH.Conway,1973

Individually,eachofElwynR.Berlekamp,JohnH.Conway,andRichardK.Guyhave receivedmuch,rightlydeserved,praise.Eachmadelastingcontributionstomanyareasofmathematics.Thisvolumeisdedicatedtotheirworkincombinatorialgame theory.Itisduetotheireffortsthatcombinatorialgametheoryexistsasasubject.

BriefHistoryofhow WinningWays cametobe

Boutonfirstanalyzednim[67],littlerealizinghowcentralnimwastobe.Inthenext twodecades,otherresearcherscontributedtheanalysisofafewother,specificgames. ThechesschampionEmanuelLaskercameclosetoacompletetheoryofimpartial games.Itwasinthe1930sthatGrundy[68]andSprague[72]gaveacompleteanalysis,nowknownastheSprague–Grundytheory.Despitebeinganeleganttheoryand easytoapply,thesubjectlanguishedbecausetherewasnocleardirectioninwhich todevelopthetheory.Inthelate1940s,RichardK.Guyrediscoveredthetheoryand definedtheoctalgames.In1956,GuyandC.A.B.SmithpublishedTheG-valuesofvariousgames [42].Thisgavetheworldaninfinitenumberofimpartialgamesandledto manyinteresting,easytostate,andyetstillunsolvedconjectures.

Theanalysisofpartizangameslookedoutofreach.TheFields’medalistJohnMilnor[70]in1953publishedSumsofpositionalgames.Thisonlycoveredgamesinwhich playersgainedwhentheyplayedandwasnoteasytoapply.In1960,JohnConway metMichaelGuy,Richard’sson.Throughthisfriendship,JohnmetRichardandasked aboutpartizangames.Thisturnedouttobearecurringthemeintheirworkinthe nexttwodecades.Alsoin1960,ElwynBerlekampgotropedintoplaying3 × 3dots&-boxesgameagainstacomputer.Helost,butknowingabouttheSprague–Grundy theory,heanalyzedthegame.(Recently,Elwynclaimedthathehadneverlostagame since.)ElwynmetRichardatthe1967ChapelHillconferenceandsuggestedthatthey writeabook.Richardagreed,gotJohnandElwyntogetherin1969,andworkbegan. Theanalysisofeachnonimpartialgamewaswellthoughtoutbutadhoc.John,with histraininginsettheory,startedtoseeastructureemergingwhengamesweredecomposedintocomponents.Hegavethenamesof1and1/2totwoabstractgamesand wasdelighted(giggledlikeababywasthephraseheused)whenhediscoveredthat, asgames,1/2 + 1/2 = 1.Hewrote OnNumbersandGames [28]inaweek.Thiscaused somefrictionamongthethree,but,eventually,workrestartedonWinningWays[3,4].

R.Austin,S.Devitt,D.Duffus,andmyself,asgraduatestudentsatCalgary, scouredtheearlypage-proofs.Wesuggestednumerousjokesandpuns.Fortunately, theauthorsrejectedallofthem.

https://doi.org/10.1515/9783110755411-201

Oneotherpersondeservestobementioned,LouiseGuy,Richard’swife.Agraciousladymadeeveryvisitortotheirhousefeelwelcome.Somepeoplehaveasked whythecombinatorialgameplayers,LeftandRight,arefemaleandmale,respectively. Theoriginalreasonshavebeenforgotten,butafterWinningWaysappeared,itbecame amarkofrespecttorememberthemas Louiseand Richard.

WhyElwyn,John,andRichardareimportant

Manybooksarewritten,enjoyalittlesuccess,andthenareforgottenbyallbuta few. OnNumbersandGames butespecially WinningWays [3,4]arestillpopulartoday.Thispopularityisduetothepersonalitiesandtheirapproachtomathematics. Allweregreatambassadorsformathematics,writingexplanatoryarticlesandgiving manypubliclectures.Morethanthat,theyunderstoodthatmathematicsneedsahumantouch.Thesedays,itiseasytogetacomputertoplayagamewell,buthowdoyou getapersontoplaywell?Thiswasoneoftheiraims. WinningWays is800+ pagesof puns,humor,easy-to-remembersayings,andverses.Theseprovidegreatandmemorableinsightsintothegamesandtheirstructures,andthebookisstillarichsourceof materialforresearchers.MathscinetreportsthatWinningWaysiscitedbyover300articles,GoogleScholarreportsover3000citations.Yet,anyreaderwillbehardpressed tofindasinglemathematicalproofinthebook.Elwyn,John,andRichardwroteitto entertain,drawinareader,andgivethemanintuitivefeelingforthegames.

Afterthepublicationof WinningWays,eventhoughallwerewellknownfortheir researchoutsideofcombinatorialgametheory,theyremainedactiveinthesubject. Eachwasinterestedinmanypartsofthesubject,but,veryloosely,theirmaininterests were:

–ElwynBerlekampconsideredtheproblemofhowtodefineandquantifythenotionofthe“urgency”ofamove.Hemadegreatstrideswithhisconceptofan enrichedenvironment[11,24,25].Hewasalsofascinatedbygo[7,8,9,10,12,11] anddots-&-boxes[13,18,23].

–JohnConwayremainedinterestedinpushingthetheoryofsurrealnumbers,particularlyinfinitegames[30,37,41],gamesfromgroupsandcodes[32,39],and misèregames[35].

–RichardK.Guyretainedaninterestinsubtractionandoctalgames,writingabook forinquisitiveyoungsters[52].Hecontinuedtopresentthetheoryasitwas[54,57, 58,59,61]andalsosummarizedtheimportantproblems[56,60,62,64].

Standingontheirshoulders

Mostofthepapersinthisvolumecanbetraceddirectlybackto WinningWays and OnNumbersandGames,ortothecontinuinginterestsofthethree.Severalthough,

illustratehowfarthesubjecthasdeveloped.Ageneralapproachofimpartialmisère gameswasonlystartedbyPlambeck[71].A.Siegel(astudentofBerlekamp),amajor figuredevelopingthistheory,pushesthisfurtherinChapter20.Thetheoryofpartizanmisèregameswasonlystartedin2007[69].Whilstplayinginthecontextofall misèregames,Chapter10analyzesaspecificgame.Chapter16containsimportantresultsforanalyzingmisèredead-endinggames.In WinningWays,dots-&-boxesand top-entailsdonotfitintothetheory,eachinaseparateway.Theyareonlypartially analyzedandthatviaadhocmethods.Chapter17findsanormalplayextensionthat coversbothtypesofgames.(Theauthorsthinkthiswouldhaveintriguedthembutare notsureiftheywouldhavefullyapproved.)

Chapters1,5–9,12,15,18,and19eitherdirectlyextendthetheoryorconsidera relatedgametoonesgiveninWinningWays.AsisevidencedbyRichardK.Guy’searly contributions,itisalsoimportanttohavenewsourcesofgames.Thesearepresented inChapters2,3,11,13,and14.

SerendipitygaveChapter4.ThispaperisthefoundationofChapters1and5.It givesasimple,effect-for-humans,testforwhengamesarenumbers.Theauthorsare surethatElwyn,John,andRichardwouldhavestarteditwitharhymingcoupletthat everyonewouldthenremember.

Elwyn,John,andRichardgavefreelyoftheirtime.Manypeoplewillrememberthe coffee-timeandeveningsattheMSRIandBIRSWorkshops.Eachwouldbeatalarge tablefullyoccupiedbyanyonewhowishedtobethere,discussingandsometimes solvingproblems.Studentswereespeciallywelcome.Allcombinatorialgamesworkshopsnowfollowthisinclusivemodel.Alargenumberofpapersoriginateatthese workshops,haveseveralcoauthors,andincludestudents.Theysharedtheirtimeoutsideofconferencesandworkshops.Manystudentswillrememberthoseoffhandmoments,withoneormoreofthem,thatoftenstretchedtohours.Iwasasecond-year undergraduatestudentwhenonmeetingJohn,heimmediatelyaskedmewhatwas 1 + 1 + 1?EvenafterIanswered“3”,hestilltookthetimetoexplaintheintricaciesof 3-playergames.(Thequestionisstillunanswered.)

Theirwit,wisdom,andwillingnesstoplayprovidedpeoplewithpleasure.They willbesorelymissed,buttheirlegacyliveson.

BooksandPapersinCombinatorialGameTheory

andRichardK.Guy,plussixotherseminalpapers

[1]ElwynR.Berlekamp.SomerecentresultsonthecombinatorialgamecalledWelter’sNim.In Proc.6thAnn.PrincetonConf.InformationScienceandSystems,pages203–204,1972.

[2]ElwynR.Berlekamp.Thehackenbushnumbersystemforcompressionofnumericaldata. Inform.andControl,26:134–140,1974.

[3]ElwynR.Berlekamp,JohnH.Conway,andRichardK.Guy. WinningWaysforYourMathematical Plays.Vol.1.AcademicPressInc.[HarcourtBraceJovanovichPublishers],London,1982.Games ingeneral.

[4]ElwynR.Berlekamp,JohnH.Conway,andRichardK.Guy. WinningWaysforYourMathematical Plays.Vol.2.AcademicPressInc.[HarcourtBraceJovanovichPublishers],London,1982. Gamesinparticular.

[5]ElwynR.Berlekamp.Blockbustinganddomineering. J.Combin.Theory(Ser.A),49:67–116, 1988.Anearlierversion,entitled Introductiontoblockbustinganddomineering,appearedin: TheLighterSideofMathematics,Proc.E.StrensMemorialConf.onRecr.Math.anditsHistory, Calgary,1986,SpectrumSeries(R.K.GuyandR.E.Woodrow,eds.),Math.Assoc.ofAmerica, Washington,DC,1994,pp.137–148.

[6]ElwynR.Berlekamp.Two-person,perfect-informationgames.In TheLegacyofJohnvon Neumann (HempsteadNY,1988), Proc.Sympos.PureMath.,volume50,pages275–287.Amer. Math.Soc.,Providence,RI,1990.

[7]ElwynR.Berlekamp.IntroductoryoverviewofmathematicalGoendgames.In Proceedingsof SymposiainAppliedMathematics, CombinatorialGames,volume43,pages73–100.American MathematicalSociety,1991.

[8]ElwynR.Berlekamp.Introductoryoverviewofmathematicalgoendgames.InR.K.Guy,editor, Proc.Symp.Appl.Math., CombinatorialGames,volume43,pages73–100.Amer.Math.Soc., Providence,RI,1991.

[9]ElwynR.BerlekampandDavidWolfe. MathematicalGo:ChillingGetstheLastPoint.AKPeters, Ltd.,Wellesley,Massachusetts,1994.

[10]ElwynR.BerlekampandDavidWolfe. MathematicalGoEndgames:Nightmaresforthe ProfessionalGoPlayer.IshiPressInternational,SanJose,London,Tokyo,1994.

[11]ElwynR.BerlekampandY.Kim.Whereisthe“thousand-dollarko?”.InR.J.Nowakowski, editor, GamesofNoChance,Proc.MSRIWorkshoponCombinatorialGames,July,1994, Berkeley,CA, MSRIPubl.,volume29,pages203–226.CambridgeUniversityPress,Cambridge, 1996.

[12]ElwynR.Berlekamp.Theeconomist’sviewofcombinatorialgames.InRichardJ.Nowakowski, editor, GamesofNoChance, MSRI,volume29,pages365–405.CambridgeUniv.Press, Cambridge,1996.

[13]ElwynR.Berlekamp. TheDotsandBoxesGame:SophisticatedChild’sPlay.AKPeters,Ltd., Natick,MA,2000.

[14]ElwynR.Berlekamp.Sumsof N × 2amazons.InF.T.BrussandL.M.LeCam,editors, Lecture Notes–MonographSeries,volume35,pages1–34.InstituteofMathematicalStatistics, Beechwood,Ohio,2000.PapersinhonorofThomasS.Ferguson.

[15]ElwynR.Berlekamp,JohnH.Conway,andRichardK.Guy. WinningWaysforYourMathematical Plays.Vol.1,secondedition.AKPeters,Ltd.,2001.

[16]ElwynR.Berlekamp.The4g4g4g4g4problemsandsolutions.InRichardJ.Nowakowski, editor, MoreGamesofNoChance, MSRIPublications,volume42,pages231–241.Cambridge UniversityPress,2002.

[17]ElwynR.Berlekamp.FourgamesforGardner.InD.WolfeandT.Rodgers,editors, Puzzler’s Tribute:AFeastfortheMind,pages383–386.AKPeters,Ltd.,Natick,MA,2002.Honoring MartinGardner.

[18]ElwynR.BerlekampandK.Scott.Forcingyouropponenttostayincontrolofaloony dots-and-boxesendgame.InRichardJ.Nowakowski,editor, MoreGamesofNoChance, MSRI Publications,volume42,pages317–330.CambridgeUniversityPress,2002.

[19]ElwynR.Berlekamp.Idempotentsamongpartisangames.InRichardJ.Nowakowski,editor, MoreGamesofNoChance, MSRIPublications,volume42,pages3–23.CambridgeUniversity Press,2002.

[20]ElwynR.Berlekamp,JohnH.Conway,andRichardK.Guy. WinningWaysforYourMathematical Plays.Vol.2,secondedition.AKPeters,Ltd.,2003.

[21]ElwynR.Berlekamp,JohnH.Conway,andRichardK.Guy. WinningWaysforYourMathematical Plays.Vol.3,secondedition.AKPeters,Ltd.,2003.

[22]ElwynR.Berlekamp,JohnH.Conway,andRichardK.Guy. WinningWaysforYourMathematical Plays.Vol.4,secondedition.AKPeters,Ltd.,2004.

[23]ElwynR.Berlekamp.Yellow-brownhackenbush.InMichaelH.AlbertandRichardJ. Nowakowski,editors, GamesofNoChance3, MSRI,volume56,pages413–418.Cambridge Univ.Press,2009.

[24]ElwynR.BerlekampandRichardM.Low.Entrepreneurialchess. Internat.J.GameTheory, 47(2):379–415,2018.

[25]ElwynR.Berlekamp.Temperaturesofgamesandcoupons.InUrbanLarsson,editor, Gamesof NoChance5, MathematicalSciencesResearchInstitutePublications,volume70,pages21–33. CambridgeUniversityPress,2019.

[26]JohnH.Conway.Allnumbersgreatandsmall.Res.PaperNo.149,Univ.ofCalgaryMath.Dept., 1972.

[27]JohnH.ConwayandH.S.M.Coxeter.Triangulatedpolygonsandfriezepatterns. Math.Gaz., 57:87–94;175–183,1973.

[28]JohnH.Conway. OnNumbersandGames.AcademicPress,1976.

[29]John H.Conway.Allgamesbrightandbeautiful. Amer.Math.Monthly,84(6):417–434,1977.

[30]JohnH.Conway.Loopygames. Ann.DiscreteMath.,3:55–74,1978.Advancesingraphtheory (CambridgeCombinatorialConf.,TrinityCollege,Cambridge,1977).

[31]JohnH.Conway.Agamutofgametheories. Math.Mag.,51(1):5–12,1978.

[32]JohnH.ConwayandN.J.A.Sloane.Lexicographiccodes:error-correctingcodesfromgame theory. IEEETrans.Inform.Theory,32(3):337–348,1986.

[33]JohnH.Conway.Morewaysofcombininggames.InR.K.Guy,editor, Proc.Symp.Appl.Math., CombinatorialGames,volume43,pages57–71.Amer.Math.Soc.,Providence,RI,1991.

[34]JohnH.Conway.Numbersandgames.InR.K.Guy,editor, Proc.Symp.Appl.Math., CombinatorialGames,volume43,pages23–34.Amer.Math.Soc.,Providence,RI,1991.

[35]W.L.SibertandJ.H.Conway.MathematicalKayles. Internat.J.GameTheory,20(3):237–246, 1992.

[36]JohnH.Conway.Onnumbersandgames.In SummerCourse1993:TheRealNumbers(Dutch), CWISyllabi,volume35,pages101–124.Math.CentrumCentrumWisk.Inform.,Amsterdam, 1993.

[37]JohnH.Conway.Thesurrealsandthereals.Realnumbers,generalizationsofthereals, andtheoriesofcontinua.In SyntheseLib.,volume242,pages93–103.KluwerAcad.Publ., Dordrecht,1994.

[38]JohnH.Conway.Theangelproblem.InR.J.Nowakowski,editor, GamesofNoChance,Proc. MSRIWorkshoponCombinatorialGames,July,1994,Berkeley,CA, MSRIPubl.,volume29, pages3–12.CambridgeUniversityPress,Cambridge,1996.

[39]JohnH.Conway. m13.In SurveysinCombinatorics, LondonMath.Soc.,LectureNoteSer., volume241,pages1–11.CambridgeUniv.Press,Cambridge,1997.

[40]JohnH.Conway. OnNumbersandGames,2ndedition.AKPeters,Ltd.,2001.Firstedition publishedin1976byAcademicPress.

[41]JohnH.Conway.Moreinfinitegames.InRichardJ.Nowakowski,editor, MoreGamesofNo Chance, MSRIPublications,volume42,pages31–36.CambridgeUniversityPress,2002.

[42]RichardK.GuyandCedricA.B.Smith.The G-valuesofvariousgames. Proc.Camb.Phil.Soc., 52:514–526,1956.

[43]RichardK.Guy.TwentyquestionsconcerningConway’ssylvercoinage. Amer.Math.Monthly, 83:634–637,1976.

[44]RichardK.Guy.Gamesaregraphs,indeedtheyaretrees.In Proc.2ndCarib.Conf.Combin.and Comput.,pages6–18.LetchworthPress,Barbados,1977.

[45]RichardK.Guy.Partisanandimpartialcombinatorialgames.In Combinatorics(Proc.Fifth HungarianColloq.,Keszthely,1976),Vol.I, Colloq.Math.Soc.JánosBolyai,volume18,pages 437–461.North-Holland,Amsterdam,1978.

[46]RichardK.Guy.Partizanandimpartialcombinatorialgames. Colloq.Math.Soc.JánosBolyai, 18:437–461,1978.Proc.5thHungar.Conf.Combin.Vol.I(A.HajnalandV.T.Sós,eds.), Keszthely,Hungary,1976,North-Holland.

[47]RichardK.Guy.Partizangames.In ColloquesInternationauxC.N.R.No.260—Problèmes CombinatoiresetThéoriedesGraphes,pages199–205,1979.

[48]RichardK.Guy.Anyonefortwopins?InD.A.Klarner,editor, TheMathematicalGardner,pages 2–15.WadsworthInternat.,Belmont,CA,1981.

[49]RichardK.Guy.Graphsandgames.InL.W.BeinekeandR.J.Wilson,editors, SelectedTopicsin GraphTheory,volume2,pages269–295.AcademicPress,London,1983.

[50]RichardK.Guy.Johnisbell’sgameofbeanstalkandJohnConway’sgameofbeans-don’t-talk. Math.Mag.,59:259–269,1986.

[51]RichardK.Guy. FairGame,COMAPMath.ExplorationSeries.Arlington,MA,1989.

[52]RichardK.Guy. FairGame:Howtoplayimpartialcombinatorialgames.COMAP,Inc.,60Lowell Street,Arlington,MA02174,1989.

[53]RichardK.Guy,editor. CombinatorialGames, ProceedingsofSymposiainApplied Mathematics,volume43.AmericanMathematicalSociety,Providence,RI,1991.Lecturenotes preparedfortheAmericanMathematicalSocietyShortCourseheldinColumbus,Ohio,August 6–7,1990,AMSShortCourseLectureNotes.

[54]RichardK.Guy.Impartialgames.In CombinatorialGames (Columbus,OH,1990), Proc. Sympos.Appl.Math.,volume43,pages35–55.Amer.Math.Soc.,Providence,RI,1991.

[55]RichardK.Guy.Mathematicsfromfun&funfrommathematics;aninformalautobiographical historyofcombinatorialgames.InJ.H.EwingandF.W.Gehring,editors, PaulHalmos: Celebrating50YearsofMathematics,pages287–295.SpringerVerlag,NewYork,NY,1991.

[56]RichardK.Guy.Unsolvedproblemsincombinatorialgames.In AmericanMathematicalSociety ProceedingsoftheSymposiumonAppliedMathematics,volume43,1991.Checkmyhomepage foracopy,http://www.gustavus.edu/~wolfe.

[57]RichardK.Guy.WhatisaGame?In CombinatorialGames,ProceedingsofSymposiainApplied Mathematics,volume43,1991.

[58]RichardK.Guy.Combinatorialgames.InR.L.Graham,M.Grötschel,andL.Lovász,editors, HandbookofCombinatorics,volumeII,pages2117–2162.North-Holland,Amsterdam,1995.

[59]RichardK.Guy.Impartialgames.InR.J.Nowakowski,editor, GamesofNoChance,Proc. MSRIWorkshoponCombinatorialGames,July,1994,Berkeley,CA,MSRIPubl.,volume29, pages61–78.CambridgeUniversityPress,Cambridge,1996.Earlierversionin: Combinatorial Games,Proc.Symp.Appl.Math.(R.K.Guy,ed.),Vol.43,Amer.Math.Soc.,Providence,RI, 1991,pp.35–55.

[60]RichardK.Guy.Unsolvedproblemsincombinatorialgames.InR.J.Nowakowski,editor, Games ofNoChance, MSRIPubl.,volume29,pages475–491.CambridgeUniversityPress,1996.

[61]RichardK.Guy.WhatisaGame?InRichardNowakowski,editor, GamesofNoChance, MSRI Publ.,volume29,pages43–60.CambridgeUniversityPress,1996.

[62]IanCaines,CarrieGates,RichardK.Guy,andRichardJ.Nowakowski.Unsolvedproblems: periodsintakingandsplittinggames. Amer.Math.Monthly,106:359–361,1999.

[63]RichardK.Guy.AviezriFraenkelandcombinatorialgames. Elect.J.Combin,8:#I2,2001.

[64]RichardK.GuyandRichardJ.Nowakowski.Unsolvedproblemsincombinatorialgames.In RichardJ.Nowakowski,editor, MoreGamesofNoChance, MSRIPublications,volume42, pages457–473.CambridgeUniversityPress,2002.

[65]RichardK.GuyandRichardJ.Nowakowski.Unsolvedproblemsincombinatorialgametheory. InM.H.AlbertandR.J.Nowakowski,editors, GamesofNoChance3, MSRI,pages465–489. CambridgeUniv.Press,2009.

[66]AlexFinkandRichardK.Guy.Thenumber-padgame CollegeMath.J.,38:260–264,2007.

[67]CharlesL.Bouton.Nim,agamewithacompletemathematicaltheory. AnnalsofMathematics, 3(2):35–39,1902.

[68]PatrickM.Grundy.Mathematicsandgames. Eureka,2:6–8,1939.

[69]G.A.MesdalandPaulOttaway.SimplificationofPartizanGamesinmisèreplay.INTEGERS, 7:#G06,2007.

[70]JohnMilnor.Sumsofpositionalgames.In:H.W.KuhnandA.W.Tucker,eds. Contributionsto theTheoryofGames,Vol.2, Ann.ofMath.Stud.,volume28,pages291–301.Princeton,1953.

[71]T.E.Plambeck.Tamingthewildinimpartialcombinatorialgames.INTEGERS,5:#G05,2005.

[72]RolandPSprague.ÜbermathematischeKampfspiele. TôhokuMath.J.,41:438–444,1935–36.

Contents

Preface V

AnthonyBonato,MelissaA.Huggan,andRichardJ.Nowakowski

Thegameofflippingcoins | 1

KyleBurke,MatthewFerland,MichaelFisher,ValentinGledel,andCraig Tennenhouse

Thegameofblockingpebbles 17

KyleBurke,MatthewFerland,andShang-HuaTeng

TransverseWave:animpartialcolor-propagationgameinspiredbysocialinfluence andQuantumNim 39

AldaCarvalho,MelissaA.Huggan,RichardJ.Nowakowski,andCarlosPereirados

Santos

Anoteonnumbers | 67

AldaCarvalho,MelissaA.Huggan,RichardJ.Nowakowski,andCarlosPereirados Santos

Ordinalsums,clockwisehackenbush,anddominoshave | 77

AlexanderClowandStephenFinbow

Advancesinfindingidealplayonposetgames | 99

ErikD.DemaineandYevheniiDiomidov

Strings-and-CoinsandNimstringarePSPACE-complete | 109

EricDuchêne,MarcHeinrich,RichardNowakowski,andAlineParreau

Partizansubtractiongames 121

MatthieuDufour,SilviaHeubach,andAnhVo CircularNimgamesCN(7, 4) | 139

AaronDwyer,RebeccaMilley,andMichaelWillette Misèredomineeringon2 × n boards 157

ZacharyGatesandRobertKelvey

Relatorgamesongroups | 171

L.R.Haff

PlayingBynum’sgamecautiously | 201

MelissaA.HugganandCraigTennenhouse Geneticallymodifiedgames 229

DouglasE.IannucciandUrbanLarsson Gamevaluesofarithmeticfunctions | 245

YukiIrie

Abase-pSprague–Grundy-typetheoremfor p-calmsubtractiongames:Welter’s gameandrepresentationsofgeneralizedsymmetricgroups 281

UrbanLarsson,RebeccaMilley,RichardNowakowski,GabrielRenault,andCarlos Santos Recursivecomparisontestsfordicotanddead-endinggamesundermisère play | 309

UrbanLarsson,RichardJ.Nowakowski,andCarlosP.Santos Impartialgameswithentailingmoves | 323

JamesB.Martin

ExtendedSprague–Grundytheoryforlocallyfinitegames,andapplicationsto randomgame-trees 343

RyoheiMiyaderaandYushiNakaya

Grundynumbersofimpartialthree-dimensionalchocolate-bargames | 367

AaronN.Siegel

Onthestructureofmisèreimpartialgames 389

Thegameofflippingcoins

Abstract:Weconsiderflippingcoins,apartizanversionoftheimpartialgameturningturtles,playedonlinesofcoins.Weshowthatthevaluesofthisgamearenumbers,andthesearefoundbyfirstapplyingareduction,thendecomposingtheposition intoaniteratedordinalsum.Thisisunusualsincemovesinthemiddleofthelinedo noteliminatetherestoftheline.Moreover,if G isdecomposedintolines H and K, then G = (H : KR).Thisisincontrasttohackenbushstrings,where G = (H : K).

1Introduction

In WinningWays,Volume3[3],Berlekamp,Conway,andGuyintroducedturning turtlesandconsideredmanyvariants.Eachgameinvolvesafiniterowofturtles, eitheronfeetorbacks,andamoveistoturnoneturtleoverontoitsback,withthe optionofflippinganumberofotherturtles,totheleft,eachtotheoppositeofitscurrentstate(feetorback).Thenumberdependsontherulesofthespecificgame.The authorsmovedtoplayingwithcoinsasplayingwithturtlesiscruel.

ThesegamescanbesolvedusingtheSprague–Grundytheoryforimpartialgames [2],butthestructureandstrategiesofsomevariantsareinteresting.Thestrategyfor moebius(flipuptofivecoins)playedwith18coins,involves Möbiustransformations; formogul(flipuptosevencoins)on24coins,itinvolvesthe miracleoctadgenerator developedbyR.CurtisinhisworkontheMathieugroup M24 andtheLeechlattice [6,7];ternups[3](flipthreeequallyspacedcoins)requiresternaryexpansions;and turningcorners[3],atwo-dimensionalversionwherethecornersofarectangleare flipped,needsnim-multiplication.

Weconsiderasimplepartizanversionofturningturtles,alsoplayedwith coins.Wegiveacompletesolutionandshowthatitinvolvesordinalsums.Thisis somewhatsurprisingsincemovesinthemiddleofthelinedonoteliminatemovesat theend.Comparethiswithhackenbushstrings[2]anddominoshave[5].

Acknowledgement: AnthonyBonatowassupportedbyanNSERCDiscoverygrant.MelissaA.HugganwassupportedbyanNSERCPostdoctoralFellowship.TheauthoralsothankstheDepartmentof MathematicsatRyersonUniversity,whichhostedtheauthorwhiletheresearchtookplace.RichardJ. NowakowskiwassupportedbyanNSERCDiscoverygrant.

AnthonyBonato, DepartmentofMathematics,RyersonUniversity,Toronto,Ontario,Canada,e-mail: abonato@ryerson.ca

MelissaA.Huggan, DepartmentofMathematicsandComputerScience,MountAllisonUniversity, Sackville,NewBrunswick,Canada,e-mail:mhuggan@mta.ca

RichardJ.Nowakowski, DepartmentofMathematicsandStatistics,DalhousieUniversity,Halifax, NovaScotia,Canada,e-mail:r.nowakowski@dal.ca

https://doi.org/10.1515/9783110755411-001

Wewilldenoteheadsby0andtailsby1.Ourpartizanversionwillbeplayedwith alineofcoins,representedbya0–1sequence d1d2 ... dn,where di ∈ {0, 1}.Withthis position,weassociatethebinarynumber ∑n i=1 di2i 1.Leftmovesbychoosingsomepair ofcoins di, dj, i < j,where di = dj = 1,andflipsthemoversothatbothcoinsare0s. Rightalsochoosesapair dk, dℓ, k < ℓ,with dk = 0and dℓ = 1,andflipsthemover.If j isthegreatestindexsuchthat dj = 1,then dk, k > j,willbedeleted.Forexample,

1011 = {0001, 001, 1 | 1101, 111}.

Thegameeventuallyendssincetheassociatedbinarynumberdecreaseswithevery move.Wecallthisgameflippingcoins.

Anotherwaytomodelflippingcoinsistoconsidertokensonastripoflocations.Leftcanremoveapairoftokens,andRightisabletomoveatokentoanopen spacetoitsleft.Weusethecoinflippingmodelforthisgametobeconsistentwiththe literature.

ThegameisbiasedtoLeft.Ifthereareanonzeroevennumberof1sinaposition, thenLeftalwayshasamove;thatis,shewillwin.Leftalsowinsanynontrivialpositionstartingwith1.However,therearepositionsthatRightwins.Thetwo-partmethod tofindtheoutcomesandvaluesoftheremainingpositionscanbeappliedtoallpositions.First,applyamodificationtotheposition(unlessitisall1s),whichreduces thenumberofconsecutive1stoatmostthree.Afterthisreduction,buildaniterated ordinalsum,bysuccessivelydeletingeverythingafterthethirdlast1,thisdeletedpositiondeterminesthevalueofthenexttermintheordinalsum.Asaconsequence, theoriginalpositionisaRightwinifthepositionremainingattheendisoftheform 0 01,andthevalueisgivenbytheordinalsum.

ThenecessarybackgroundfornumbersisinSection2.Section3containsresults aboutoutcomesandalsoincludesourmainresults.First,weshowthatthevaluesare numbersinTheorem3.2.Next,analgorithmtofindthevalueofapositionispresented, andTheorem3.3statesthatthevaluegivenbythealgorithmiscorrect.

TheactualanalysisisinSection4.Itstartsbyidentifyingthebestmovesforboth playersinTheorem4.2.ThisleadsdirectlytothecoreresultLemma4.5,whichshows thatthevalueofapositionisanordinalsum.Theordinalsumdecompositionof G isfoundasfollows.Let GL bethepositionaftertheLeftmovethatremovestherightmost1s.LetH bethestringG \GL;thatis,thesubstringeliminatedbyLeft’smove.Let HR betheresultofRight’sbestmovein H.Nowwehavethat G = GL : HR.Incontrast, theordinalsumsforhackenbushstringsanddominoshave[5]involvethevalueof H not HR .

TheproofofTheorem3.3isgiveninSection4.1.Thefinalsectionincludesabrief discussionofopenproblems.

Finally,weposeaquestionforthereader,whichweanswerattheendofSection4.1:Whowins0101011111 + 1101100111 + 0110110110111andhow?

2Numbers

Allthevaluesinthispaperarenumbers,andthissectioncontainsallthenecessary backgroundtomakethepaperself-contained.Forfurtherdetails,consult[1,8].Positionsarewrittenintermsoftheiroptions;thatis, G = {Gℒ | Gℛ}.

Definition2.1 ([1,2,8]).Let G beanumberwhoseoptionsarenumbers,andlet GL , GR betheLeftandRightoptionsofthecanonicalformof G.

1.Ifthereisaninteger k, GL < k < GR,orifeither GL or GR doesnotexist,then G is theinteger,say n,closesttozerothatsatisfies GL < n < GR .

2.Ifboth GL and GR existandthepreviouscasedoesnotapply,then G = p 2q ,where q istheleastpositiveintegersuchthatthereisanoddinteger p satisfying GL < p 2q < GR .

Thepropertiesofnumbersrequiredforthispaperarecontainedinthenextthree theorems.

Theorem2.2 ([1,2,8]). LetGbeanumberwhoseoptionsarenumbers,andletGL and GR betheLeftandRightoptionsofthecanonicalformofG.IfG′ andG′′ areanyLeft andRightoptions,respectively,then

Theorem2.2showsthatifweknowthatthestringofinequalitiesholds,thenwe needtoonlyconsidertheuniquebestmoveforbothplayersinanumber.

Weincludethefollowingexamplestofurtherillustratetheseideas:

(a)0 = {|} = { 9 |} = { 1 2 | 7 4};

(b) 2 = {| 1} = { 5 2 | 31 16};

(c)1 = {0 |} = {0 | 100};

(d) 1 2 = {0 | 1} = {3 8 | 17 32}.

Forgames G and H,toshowthat G ⩾ H,weneedtoshowthat G H ⩾ 0,meaning thatweneedtoshowthat G H isaLeftwinmovingsecond.Formoreinformation, seeSections5.1,5.8,and6.3of[1].

Let G and H begames.The ordinalsum of G,the base,and H,the exponent,is

Intuitively,playingin G eliminates H,butplayingin H doesnotaffect G.Foreaseof reading,ifanordinalsumisaterminanexpression,thenweencloseitinbrackets. Notethat x : 0 = x = 0 : x sinceneitherplayerhasamovein0.Wedemonstrate howtocalculatethevaluesofotherpositionswiththefollowingexamples:

(a)1 : 1 = {1 |} = 2;

G′ ⩽ GL < G < GR ⩽ G′′

(b)1 :

=

Notethatinallcases,whenbaseandexponentarenumbers,theplayersprefertoplay intheexponent.Intheremainderofthispaper,alltheexponentswillbepositive. OneofthemostimportantresultsaboutordinalsumswasfirstreportedinWinning Ways.

Theorem2.3 (ColonPrinciple[2]). IfK ⩾ K′,thenG : K ⩾ G : K′ . TheColonPrinciplehelpsproveinequalitiesthatwillbeusefulinthispaper.

Theorem2.4. LetGandHbenumbersallofwhoseoptionsarealsonumbers,andlet H ⩾ 0.

1. IfH = 0,thenG : H = G.IfH > 0,then (G : H) > G. 2. GL < (G : HL) < (G : H) < (G : HR) < GR .

Proof. Foritem(1),theresultfollowsimmediatelybyTheorem2.3. Foritem(2),if H ⩾ 0andalltheoptionsof G and H arenumbers,then GL < G = (G : 0) ⩽ (G : HL) < (G : H) < (G : HR).Thesecond,third,andfourthinequalitieshold since H isanumberandthus0 ⩽ HL < H < HR andbyapplyingtheColonPrinciple. Tocompletetheproof,weneedtoshowthat (G : HR) < GR.Todoso,wecheckthat GR (G : HR) > 0;inwords,wecheckthatLeftcanalwayswin.Leftmovingfirstcan moveinthesecondsummandto GR GR = 0andwin.Rightmovingfirsthasseveral options:

1.Movingto GR GL > 0,since G anditsoptionsarenumbers.HenceLeftwins.

2.Movingto GR (G : HRL) > 0byinduction.

3.Movingto GRR G : HR,butLeftcanrespondto GRR GR > 0since G andits optionsarenumbers.

Inallcases,Leftwinsmovingsecond.Theresultfollows.

Toprovethatallthepositionsarenumbers,weuseresultsfrom[4].Asetofpositionsfromarulesetiscalleda hereditarilyclosedsetofpositionsofaruleset ifitis closedundertakingoptions.Thisgamesatisfiesrulesetpropertiesintroducedin[4]. Inparticular,thepropertiesarecalledtheF1property andtheF2property,whichboth highlightthenotionof First-move-disadvantage innumbersandaredefinedformally asfollows.

Definition2.5 ([4]).Let S beahereditarilyclosedruleset.Givenaposition G ∈ S,the pair(GL , GR) ∈ Gℒ ×Gℛ satisfiestheF1propertyifthereisGRL ∈ GRℒ suchthatGRL ⩾ GL orthereis GLR ∈ GLℛ suchthat GLR ⩽ GR .

Definition2.6 ([4]).Let S beahereditarilyclosedruleset.Givenaposition G ∈ S,the pair (GL , GR) ∈ Gℒ × Gℛ satisfiesthe F2property ifthereare GLR ∈ GLℛ and GRL ∈ GRℒ suchthat GRL ⩾ GLR .

Asprovenin[4],ifgivenanyposition G ∈ S,allpairs (GL , GR) ∈ Gℒ × Gℛ satisfy oneoftheseproperties,thenthevaluesofallpositionsarenumbers.Furthermore, satisfyingtheF2propertyimpliessatisfyingtheF1property,anditwasshownthatall positions G ∈ S arenumbersifandonlyifforany G ∈ S,allpairs (GL , GR) ∈ Gℒ × Gℛ satisfytheF1property.Combiningtheseresultsgivesthefollowingtheorem.

Theorem2.7 ([4]). LetSbeahereditarilyclosedruleset.AllpositionsG ∈ Sarenumbers ifandonlyifforanypositionG ∈ S,allpairs (GL , GR) ∈ Gℒ × Gℛ satisfyeithertheF1or theF2property.

3Mainresults

Beforeconsideringthevaluesandassociatedstrategies,weconsidertheoutcomes, thatis,wepartiallyanswerthequestion“Whowinsthegame?”Thefullanswerrequiresananalogousanalysistofindingthevalues.

Theorem3.1. LetG = d1d2 dn.Ifd1d2 dn containsanevennumberof1s,orifd1 = 1 andthereareleasttwo 1s,thenLeftwinsG.

Proof. ARightmovedoesnotdecreasethenumberof1sintheposition.Thus,ifin G, Lefthasamove,thenshestillhasamoveafteranyRightmovein G.Consequently, regardlessofd1,ifthereareanevennumberof1sinG,thenitwillbeLeftwhoreduces thegametoall0s.Similarly,ifd1 = 1andthereareanoddnumberof1s,thenLeftwill eventuallyreduce G toapositionwithasingle1,thatis,to d1 = 1and di = 0for i > 1. Inthiscase,Righthasnomoveandloses.

Theremainingcase,d1 = 0andanoddnumberof1s,ismoreinvolved.Theanalysisofthiscaseisthesubjectoftheremainderofthepaper.Wefirstprovethefollowing: Theorem3.2. All flippingcoins positionsarenumbers.

Proof. LetGbeaflippingcoinsposition.Ifonlyoneplayerhasamove,thenthegame isaninteger.Otherwise,letLbetheLeftmovetochange (di, dj) from (1, 1) to (0, 0).Let RbetheRightmovetochange (dk, dℓ) from (0, 1) to (1, 0).Nootherdigitsarechanged. Ifallfourindicesaredistinct,thenboth L and R canbeplayedineitherorder.Inthis case, GLR = GRL.ThustheF2propertyholds.Ifthereareonlythreedistinctindices, thentwoofthebitsareones.IfLeftmovesfirst,then di = dj = dk = 0.IfRightmoves first,thentherearestilltwoonesremainingafterhismove.AfterLeftmoves,wehave di = dj = dk = 0,andhence GL = GRL.TheF1propertyholds.

Therearenomorecasessincetheremustbeatleastthreedistinctindices.Since everypositionsatisfieseithertheF1ortheF2property,byTheorem2.7itfollowsthat everypositionisanumber.

Givenaposition G,thefollowingalgorithmreturnsavalue.

Algorithm Let G beaflippingcoinsposition.Let G0 = G.

1.Set i = 0.

2.Reductions:Let α and β bebinarystrings,andeithercanbeempty.

(a)If G0 = α013+jβ, j ⩾ 1,thenset G0 = α101jβ.

(b)If G0 = α013β with β containinganevennumberof1s,thenset G0 = α10β.

(c)Repeatuntilneithercaseapplies;thengotoStep3.

3.If Gi is0r1, r ⩾ 0,or1a0pi10qi1, a ⩾ 0,and pi + qi ⩾ 0,thengotoStep5.

Otherwise, Gi = α01a0pi10qi1, pi + qi ⩾ 1, a > 0,andsome α.Set Qi = 0pi10q

GotoStep4.

4.Set i = i + 1.GotoStep3.

5.IfGi = 0r1,thensetvi =−r.IfGi = 1a0pi10qi1,thenset

6.For j from i 1downto0,set vj = vj+1 : 1 2

+qj 1 .

.GotoStep6.

7.Returnthenumber v0.

Thealgorithmimplicitlyreturnstwodifferentresults:

1.ForStep3,thesubstrings Q0, Q1,..., Qi 1, Gi partitionthereducedversionof G; 2.Thevalue v0.

First,weillustratethealgorithmwiththefollowingexample.Considertheposition G = 10011110110110111011110011.Wehighlightateachstepwhichreductionisbeing appliedtotheunderlineddigits;2(a)isdenotedby †,whereas2(b)isdenotedby ‡. Thealgorithmgivesthat

10011110110110111011110011 = 10011110110110111011110011(†) = 100111101101101111010011(†) = 1001111011011101010011(‡) = 10011110111001010011(‡) = 100111110001010011(†) = 1010110001010011.

Step3partitionsthelastexpressioninto101(011)(000101)(0011) sothattheordinal sumisgivenby v0 = ((1 2 : 1 2) : 1 64) : 1 8 = 10257 16348.

Nowlet H = 01001110110111011101.Thereductionsgivethat 01001110110111011101 = 01001110110111011101 = 010011101110011101 = 0100111100011101 = 01010100011101.

Thelastexpressionpartitionsinto01(0101)(00011)(101) sothat

0 = (( 1 : 1 4) : 1 32) : 1 =− 893 1024

Thenexttheoremisthemainresultofthepaper.

Theorem3.3 (Valuetheorem). LetGbea flippingcoins position.Ifv0 isthevalue obtainedbythealgorithmappliedtoG,thenG = v0.

Inthenextsection,wederiveseveralresultsthatwillbeusedtoproveTheorem3.3.TheproofofTheorem3.3willappearinSection4.1.

4Bestmovesandreductions

Theproofsinthissectionuseinductionontheoptions.Analternatebutequivalent approachistoregardthetechniquesasinductionontheassociatedbinarynumberof thepositions.Theproofsrequiredetailedexaminationofthepositions,andwewill usenotationsuitabletothecasebeingconsidered.Often,atypicalpositionwillbe writtenasacombinationofgenericstringsandthesubstringunderconsideration. Forexample,111011000110101mightbeparsedas (11101)(100011)(0101) andwritten as α100011β ormorecompactlyas α10312β.

WerequireseveralresultsbeforebeingabletoproveTheorem3.3.Webeginby provingasimplifyingreduction,followedbythebestmovesforeachplayer,andthen theremainingreductionsusedinthealgorithm.

AsanimmediateconsequenceofTheorems3.2and2.2,wehavethefollowing:

Corollary4.1. Letα,β,andγbearbitrarybinarystrings.Wethenhavethatα1β0γ > α0β1γ.Moreover,foranintegerr ⩾ 0,wehavethatβ10r1 > β.

Proof. RecallthatbyTheorem3.2allflippingcoinspositionsarenumbers.ThusTheorem2.2applies.

ARightoptionof α0β1γ is α1β0γ,andsowehavethat α1β0γ > α0β1γ.Similarly,a Leftoptionof β10r1is β,andsowehavethat β10r1 > β.

Next,weprovethebestmovesforeachplayer.Rightwantstoplaythezerofurthest totherightandthe1adjacenttoit.Leftwantstoplaythetwoonesfurthesttotheright.

Theorem4.2. LetGbea flippingcoins position,whereinG,randn,r n,arethe greatestindicessuchthatdr = dn = 1.Letsbethegreatestindexsuchthatds = 0.Left’s bestmoveistoplay (dr, dn),andRight’sbestmoveistoplay (ds, ds+1).

Proof. Weprovethistheorembyinductionontheoptions.Notethatweusetheequivalentbinaryrepresentationofthegameposition.Iftherearethreeorfewerbits,then, byexhaustiveanalysis,thetheoremistrue.

Let G be d1d2 ... dn.WebeginbyprovingLeft’sbestmoves.Let r and n bethetwo largestindices,wheredr = dn = 1,andthusdk = 0forr < k < n.Letiandj,i < j,betwo indiceswithdi = dj = 1.WeusethenotationG(di, dj, dr, dn) tohighlightthesalientbits. TheclaimedbestLeftmoveisfromG(1, 1, 1, 1) toG(1, 1, 0, 0).Thismustbecomparedto anyotherLeftmove,representedbymovingfrom G(1, 1, 1, 1) to G(0, 0, 1, 1).Thatis,we needtoshowthat G(1, 1, 0, 0) G(0, 0, 1, 1) ⩾ 0.

Forthemovestobedifferent,atleastthreeofi,j,r,naredistinct.Wefirstassume thatthefourindicesaredistinct.Inthiscase,wehavethat i < j < r < n.Byapplying Corollary4.1twicewehavethat

G

Wemayassumethen,withoutlossofgenerality,that j = r or j = n.If j = n,then i < r,sincetherearetwodistinctmoves.Nowconsider G(di, dr, dn) = G(1, 1, 1).By Corollary4.1wehavethatif j = r,then G(1, 0, 0) > G(0, 0, 1),andif j = n,then G(1, 0, 0) > G(0, 1, 0).

WenowproveRight’sbestmove.Therearemorecasestoconsider.Let s bethe largestindexsuchthat ds = 0andtherefore ds+1 = 1.Let i, j, i < j beindiceswith di = 0and dj = 1.Theclaimedbestmoveis ds, ds+1,andthismustbecomparedtothe arbitraryRightmove di, dj.Forthemovestobedifferent,theremustbeatleastthree distinctindices.

Theoriginalpositioniseither

G(di, dj, ds, ds+1) = G(0, 1, 0, 1), i < s,

or G(ds, ds+1, dj) = G(0, 1, 1), i = s, j > s + 1

WeneedtoshoweitherD = G(1, 0, 0, 1) G(0, 1, 1, 0) ⩾ 0orD = G(1, 1, 0) G(1, 0, 1) ⩾ 0, respectively.SupposeRightplaysinthefirstsummandof D.Notethat,byinduction, thebestmovesofLeftandRightareknown.

1.First,suppose j < s.ByinductionRight’sbestmoveinthefirstsummandof D is to D′ = G(1, 0, 1, 0) G(0, 1, 1, 0).Since i < j,itfollowsthat G(1, 0, 1, 0) isaRight optionof G(0, 1, 1, 0),andthus D′ ispositivebyCorollary4.1.

2.If j = s + 1,thenthereareonlythreedistinctindices.Theoriginalgameis G(di, ds, ds+1) = G(0, 0, 1) and D = G(1, 0, 0) G(0, 1, 0).Since G(1, 0, 0) isaRight optionof G(0, 1, 0),itfollowsthat D ispositivebyCorollary4.1.

3.Suppose j > s + 1.

If i < s,thentheoriginalgameisoftheform

and

TwoapplicationsofCorollary4.1(appliedtothehighlightedterms)give

If i = s,then

and

OneapplicationofCorollary4.1(relevanttermsagainhighlighted)gives

Thus D ⩾ 0.

Next,weconsiderRightmovinginthesecondsummandofD=G(1, 0, 0, 1) G(0, 1, 1, 0). Notethatbythechoicesofthesubscripts, dℓ = 1if n ⩾ ℓ ⩾ s + 1.

1.If n > s + 2,thenRight’sbestmoveinthesecondsummandistochange dn 1, dn from (1, 1) to (0, 0).Leftcopiesthismoveinthefirstsummand,andtheresulting differencegameisnonnegativebyinduction.

2.Suppose n = s + 2.

i.If j < s + 1,then G(di, dj, ds, ds+1, ds+2) = G(0, 1, 0, 1, 1) and D = G(1, 0, 0, 1, 1) G(0, 1, 1, 0, 1).Right’sbestmoveisto G(1, 0, 0, 1, 1) G(0, 1, 0, 0, 0).Leftmoves toG(1, 0, 0, 0, 0) G(0, 1, 0, 0, 0).ThisispositivebyCorollary4.1,andLeftwins. Forthenexttwosubcases,exactlytwo1swilloccupytwoofthefourindexed positions.SinceRightismovinginthesecondsummand,heischangingtwo 1stotwo0s.ThusLeft’sbestresponseforeachcaseistomoveinthefirst summand,bringingthegameto G(0, 0, 0, 0) G(0, 0, 0, 0) = 0,andshewins. Forthesecases,weonlylisttheoriginalposition.Thestrategyforbothcases isasjustdescribed.

ii.If j = s + 1,then G(di, ds, ds+1, ds+2) = G(0, 0, 1, 1) and D = G(1, 0, 0, 1)

G(0, 1, 0, 1).

iii.If j = s + 2,then G(di, ds, ds+1, ds+2) = G(0, 0, 1, 1) and D = G(1, 0, 1, 0)

G(0, 1, 0, 1).

3.Nowsuppose n = s + 1.

i.If j < s + 1,thenlet ℓ < s + 1bethelargestindexsuchthat dℓ = 1.

Ifj < ℓ,thenwehaveG(di, dj, dℓ, ds, ds+1)=G(0, 1, 1, 0, 1) andD=G(1, 0, 1, 0, 1)

G(0, 1, 1, 1, 0).Right’sbestmoveisto G(1, 0, 1, 0, 1) G(0, 1, 0, 0, 0).Leftmoves toG(1, 0, 0, 0, 0) G(0, 1, 0, 0, 0),whichispositivesinceG(1, 0, 0, 0, 0) isaRight optionof G(0, 1, 0, 0, 0).

If j = ℓ,then G(di, dj, ds, ds+1) = G(0, 1, 0, 1) and D = G(1, 0, 0, 1) G(0, 1, 1, 0). Right’sbestmoveisto G(1, 0, 0, 1) G(0, 0, 0, 0).Leftmovesto G(0, 0, 0, 0)

G(0, 0, 0, 0) = 0,andLeftwins.

ii.If j = s + 1,then G(di, ds, ds+1) = G(0,

).Thisis positivebyCorollary4.1.

Inallcases,Leftwins D movingsecond,provingtheresult.

SupposeinapositionthatthebitsofthebestRightmovearedifferentfromthose ofthebestLeftmove.Thenextlemmaessentiallysaysthatthepositionsbeforeandafteronemovebyeachplayerareequal.Itisphrasedinawaythatisusefulforreducing thelengthoftheposition.RecallthatanontrivialpositionlookslikeG = α

a0p10q1β, where a, p,and q arenonnegativeintegers,and α and β arearbitrarybinarystrings. Forthealgorithm,itsufficestoprovetheresultforβbeingempty.However,itisuseful, certainlyforahuman,toreducethelengthofthepositionasmuchaspossible.

Lemma4.3. Letαbeanarbitrarybinarystring.Ifa ⩾ 0,thenwehavethatα01111a = α101a .

Proof. Let H = α01111a α101a.Weneedtoshowthat H = 0.Tosimplifytheproof, insomecasesthesecondplayerwillplaysuboptimalmoves.Wehaveseveralcasesto consider.

1.If a ⩾ 2,thenplayingthesamemoveintheothersummandisagoodresponse. Aftertwosuchmoves,wehaveeither

or

2.If a = 1,then H = α01111 α101.Thecasesare:

i.Leftplaysinthefirstsummandto α011 α101;thenRightmovesto α101

α101 = 0.

ii.Rightplaysinthesecondsummandtoα01111 α;thenLeftmovestoα011 α. Since (α011)L = α,wehave α011 > α.

iii.Rightplaysinthefirstsummandtoα10111 α101;thenLeftrespondstoα101

α101 = 0.

iv.Leftplaysinthesecondsummandtoα01111 α11;thenRightmovestoα10111 α11 = α11 α11 = 0byinduction.

3.If a = 0,then H = α0111 α1.Thereareseveralcasestoconsider.

i.IfLeftorRightplaysinthefirstsummand,thentheresponseisinthefirst summandgiving α1 α1 = 0.

ii.IfLeftplaysinthesecondsummand,thensincethereisaLeftmove,wehave α = β01b , b ⩾ 0.If b > 0,thenwehavethat β01b0111 β01b1,andLeft movesto β01b013 β101b.HereRightrespondsto β101b 1013 β101b,which byinductionisequalto β101b 11 β101b = 0.If b = 0,thenwehavethat β01b0111 β01b1 = β00111 β01,andwewanttoshowthatRightcanwinmovingsecond.Leftplaysto β00111 β10,andRightcanrespondto β01110 β1, which,byinduction,isequalto β1 β1 = 0.

iii.Rightplaysinthesecondsummand.ThenforaRightmovetoexist,α = β10a , a ⩾ 0.Thus H = β10a0111 β10a1,andRightmovesto β10a0111 β.Left respondsbymovingto β00a011 β.Wethenhavethat (β00a011)L = β,and thus β00a011 > β.Hencewefindthat β00a011 β > 0.

Inallcasesthesecondplayerwins H therebyprovingtheresult.

Therearereductionsthatcanbeappliedtothemiddleoftheposition,butextra conditionsareneeded.

Lemma4.4. Letαandβbearbitrarybinarystringswhereeither (a) βstartswitha 1,or (b) βstartswith 0 andhasanevennumberof 1s.Wethenhavethat

0111β = α10β

Proof. Let H = α0111β α10β.Weneedtoshowthat H = 0.Wehaveseveralcasesto consider.

1.Ifβ isemptyorβ = 1a,thenH = 0byLemma4.3.Thereforewemayassumethatβ hasatleastone1andone0.

2.If β = 1γ1(β mustendina1),theninbothsummandsthebestmovesarepairsof bitsinβand β.Ifeachplayercopiestheopponent’smoveintheothersummand, thenthisleadsto

andthelatterexpressionisequalto0byinduction.

3.If β 1γ1,then β = 0γ1,and γ1hasatleasttwo1s.Thebestmovesarein β and β andarethebestresponsestoeachother.Wethenderivethat

Inallcases, H = 0,andthisconcludestheproof.

InLemma4.4theconditionsarenecessary.Anexampleis

3/8 = 011101 1001 = 1/4

Here β startswitha0andhasanoddnumberof1s.

Thesereductionlemmasareimportantinevaluatingaposition.Thereducedpositionswillendin011or01.Byconsideringtheexactendofthestring,specifically,if thereareatleasttwo0s(inonespecialcase,three0s),thenwecanfindanordinal sumdecomposition.Thedecompositionisdeterminedbywherethethirdrightmost1 issituated.

Thenextresultisthestartoftheordinalsumdecompositionofaposition.The exponentisthevalueoftheRightoptionofthesubstringbeingremoved.

Lemma4.5. Letαbeanarbitrarybinarystring.Ifa ⩾ 1 andpandqarenonnegative integerssuchthatp + q ⩾ 1,then

Proof. Weprovethat

Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.
[FREE PDF sample] Combinatorial game theory: a special collection in honor of elwyn berlekamp, john by Education Libraries - Issuu