Thispaperisintendedasatechnicalexplanation,includingpossibleformulas,fordesigninga Thispaperisintendedasatechnicalexplanation,includingpossibleformulas,fordesigninga newdebatetabulationprogrambasedonamatrixoptimizationalgorithm.Ihavespentalotof newdebatetabulationprogrambasedonamatrixoptimizationalgorithm.Ihavespentalotof timeexperimentingw
timeexperimentingwithdifferentformulasaithdifferentformulasandhavearrivedattheondhavearrivedattheonesIthinkworkbnesIthinkworkbest.Iam,est.Iam, howe
however,notacomputerver,notacomputerprogrammprogrammer.Ifanyonewanter.Ifanyonewantstostodeveldevelopaopaprogprogrambasedrambasedonon thesethese ideas,pleasedosoand
ideas,pleasedosoandsharewiththecomsharewiththecommunity.munity.
Section1:HowaMatrixOptimizationWorks Section1:HowaMatrixOptimizationWorks
Thekeyideathatmustbeundersto
Thekeyideathatmustbeunderstoodisodisthatcurrethatcurrentalgoritntalgorithmsforpairinghmsforpairing,whetherdoneby,whetherdoneby handonnotecardsorbyacom
handonnotecardsorbyacomputerprogram,aresequentiaputerprogram,aresequentialnotsimultaneous.Forexamlnotsimultaneous.Forexample,inple,in onecurren
onecurrentversionoftversionof apower-maapower-matchintchingalgorithmgalgorithm,debaters,debaters arerankedareranked fromfrom toptobottom,toptobottom, thenthedebatersarematchedtogethertwo-by-two:firstplacedebatessecond,thirddebates thenthedebatersarematchedtogethertwo-by-two:firstplacedebatessecond,thirddebates fou
fourthrth,, andand soon.Ifsoon.If amatcamatchh isprohisprohibiibitedted bytourbytournamnamentent rulrules—es— forfor exaexampmple,le, thethe firfirstandstand secondplacedebatersarefromthesameschool—thenthenextdebaterispicked,andfirst secondplacedebatersarefromthesameschool—thenthenextdebaterispicked,andfirst placedebatesthird. placedebatesthird. Everycurrentalgorithmusessomekindofsequentialprocess.Thelimitationisthatonlyone Everycurrentalgorithmusessomekindofsequentialprocess.Thelimitationisthatonlyone var
variabiablele cancan bebe conconsidsidereered,d, eveevenn ifif thathatt oneone varvariabiablele isis oftoftenen comcomposposeded ofof sevseveraerall difdifferferentent mea
measursures.es. ForFor exaexampmple,le, aa debdebateater’sr’s plaplacece inin thethe toutournarnamenmentt isis aa varvariabiable,le, whwhichich migmightht bebe composedofthemeasureswin-lossrecord,speakerpoints,opponentwins,etc.Itisimpossible composedofthemeasureswin-lossrecord,speakerpoints,opponentwins,etc.Itisimpossible forasequentialalgorithmtobalancetwovariablesatonce,suchasgeographyandskilllevel. forasequentialalgorithmtobalancetwovariablesatonce,suchasgeographyandskilllevel. Al
Alll aa seseququenentitialal alalgogoririththmm cacann dodo isis ococcacasisiononalallyly ststririkeke aa popotetentntiaiall mamatctchh foforr viviololatatiningg ononee cri
criterterionion oror anoanothether,r, e.ge.g.. elimeliminainatintingg aa mamatchtch becbecauausese thethe twotwo debdebateatersrs areare frofromm thethe samsamee state.
state.
Asimultaneousalgorithmcanoptimizeseveralvariablesatonce.Thebasicoutlineofsuchan Asimultaneousalgorithmcanoptimizeseveralvariablesatonce.Thebasicoutlineofsuchan alg
algoriorithmthm issimpissimpletoleto desdescricribe:be: stestep1p1 istoisto assassignign apointapoint valvalue,ue, orscororscore,basee,basedondon asmanasmanyy varia
variablesasdesiredblesasdesired,to,to everyeverypossiblematcpossiblematch;step2h;step2istopicktheistopickthe pairipairingthathasthehighestngthathasthehighest averagepointvaluepermatch.Step2isawell-knownandsolvedproblemincomputerscience averagepointvaluepermatch.Step2isawell-knownandsolvedproblemincomputerscience (se
(seee HunHungargarianian algalgoriorithmthm).There).Thereforfore,e, thethe onlonlyy proprobleblemm toto solsolveve isis howhow toto assassignign aa scoscorere toto everymatch.
everymatch.
Step1generatesamatrix,listingdebatersbyrowandbycolumn.Inaneven,side-constrained Step1generatesamatrix,listingdebatersbyrowandbycolumn.Inaneven,side-constrained roun
round,theteamsd,theteamsdueAffirmadueAffirmativecoultivecouldbelistedineachrow,anddbelistedineachrow,andtheteamsdueNegatheteamsdueNegativearetiveare listedbycolumn.Inanoddround,theteamscanberandomlyassignedtoeitherhalf,orthe listedbycolumn.Inanoddround,theteamscanberandomlyassignedtoeitherhalf,orthe programcouldassignteamsfromthesameschooltothesameside,sincetheyareblockedand programcouldassignteamsfromthesameschooltothesameside,sincetheyareblockedand cannotdebateeachother.Usingvariablesstoredforeachteam,theprogramcanpopulatethe cannotdebateeachother.Usingvariablesstoredforeachteam,theprogramcanpopulatethe
matrixwithdifferentscoresforeachpossiblematch.Inthecasewheretheteamsareblocked, matrixwithdifferentscoresforeachpossiblematch.Inthecasewheretheteamsareblocked, thescoreshouldbeset
thescoreshouldbesetto0.Instep2,thepto0.Instep2,theprogrampicksourogrampicksouttheoptimalsetofmttheoptimalsetofmatches.atches.
Section2:ParticularFormulasforParticularUses Section2:ParticularFormulasforParticularUses Formula1:StrengthofSchedule Formula1:StrengthofSchedule Theproblemwithahigh-highorhigh-lowpairingmethodisthatthestrengthofschedulecan Theproblemwithahigh-highorhigh-lowpairingmethodisthatthestrengthofschedulecan stillvarywidely,evenwithinabracket.Asevidence,lookattheopponentwinscolumninthe stillvarywidely,evenwithinabracket.Asevidence,lookattheopponentwinscolumninthe final
finalresultsofaresultsofa tourtournamenament;withint;withinthesamebrackenthesamebracket,say4-2s,oppont,say4-2s,opponentwinscanvaryentwinscanvaryfromfrom thelowteenstothehightwenties.Theproblemisthatthemethodmatchesteamsbyonlyone thelowteenstothehightwenties.Theproblemisthatthemethodmatchesteamsbyonlyone variable–strength(record,points)–andcannotconsiderasecondvariable–schedulestrength variable–strength(record,points)–andcannotconsiderasecondvariable–schedulestrength –atthesamet –atthesametime.ime. Thisisnoproblemforamatrixoptimizationmethod.ForapotentialmatchbetweenteamsA Thisisnoproblemforamatrixoptimizationmethod.ForapotentialmatchbetweenteamsA andB,theprogr
andB,theprogramneedsamneedstouseonlythestrengthtouseonlythestrengthsofteamsAandB,softeamsAandB, s s
A
Aandand s s B B,andalsothe,andalsothe strengthsoftheprioropponentsofteamA, strengthsoftheprioropponentsofteamA, s s C C ,, s s D D,, s s E E ,...,...
{
{
}}
,andofteamB,,andofteamB, s s N N ,, s sOO,, s s P P ,...,...{
{
}}
.Then.Then theprogramcalculates theprogramcalculates! ! AAofthesetoftheset
{
{
s s B B,, s sC C ,, s s D D,, s s E E ,...,...}}
andand! !B
Bofthesetoftheset
{
{
s s A A,, s s N N ,, s sOO,, s s P P ,...,...}}
.Each.Eachpotentialmatchisscoredby potentialmatchisscoredby (1.1) (1.1) scor scoree sos sos == 2 2 1 1 ! ! A A + + 11 ! ! B B withahigherscorebeingbetter.
withahigherscorebeingbetter.Therefore,thismethodlooksforthTherefore,thismethodlooksfortheopponentforteamAthaeopponentforteamAthatt mo
mostst incincreareasesses thethe stastandandardrd devdeviatiationion ofof itsits oppopponeonentnt setset –– i.ei.e.,., thathatt momostst balbalanancesces outout itsits sche
schedule–andlikewiseforteamB.Usingtheharmodule–andlikewiseforteamB.Usingtheharmonicmeantocombinnicmeantocombinethesetwostandarethesetwostandardd deviationsensuresthatonlymatchesthatbalanceoutbothteams’schedulescanreceivehigh deviationsensuresthatonlymatchesthatbalanceoutbothteams’schedulescanreceivehigh scores. scores. Strengthcouldbemeasuredinanyway:winsalone;winsplusafractionforspeakerpoints;or Strengthcouldbemeasuredinanyway:winsalone;winsplusafractionforspeakerpoints;or pointsfortheteam’srankorderoutofthenteamsatthetournament,wherefirstplace=n pointsfortheteam’srankorderoutofthenteamsatthetournament,wherefirstplace=n points,second=n–1,…,andlast=n–n=0.(Mypreferredwaytomeasureateam’sstrength points,second=n–1,…,andlast=n–n=0.(Mypreferredwaytomeasureateam’sstrength isdiscussedinalatersec
VariantA VariantA
Themethodforstrength-of-schedulepairingsstatedaboveignoresbrackets.Indeed,becauseit Themethodforstrength-of-schedulepairingsstatedaboveignoresbrackets.Indeed,becauseit istryingtopusheveryteam
istryingtopusheveryteamtoanequallybalancedschedtoanequallybalancedschedule,themethodstatedaboule,themethodstatedabovewillturnvewillturn atournamentintoakindofpartialroundrobin.Eachteamwillfaceopponentsfromthetop, atournamentintoakindofpartialroundrobin.Eachteamwillfaceopponentsfromthetop, middle,andbottombrackets.
middle,andbottombrackets.
Tocorrectthis,anadditionalfactor, Tocorrectthis,anadditionalfactor,winswins
A
A
!
!
winswins B B ,needstobeconsidered.Toforcewithin- ,needstobeconsidered.Toforcewithin-bracketpairings,thescoringneedstobeadjustedsuchthatthebestone-bracketpull-upmatch bracketpairings,thescoringneedstobeadjustedsuchthatthebestone-bracketpull-upmatch hasalowerscorethantheworstwithin-bracketpairing.Inthisway,theprogramwillchoosea hasalowerscorethantheworstwithin-bracketpairing.Inthisway,theprogramwillchoosea pulpull-ul-upp mamatchtch onlonlyy ifitifit isnecesisnecessarsarybecauybecauseasea brabrackeckett hashas anuneveanunevennumbnnumberoferof teateams.ms. TheThe programwillthenchoosethebestpossiblepull-upmatch.Itisalsoworthconsideringmultiple programwillthenchoosethebestpossiblepull-upmatch.Itisalsoworthconsideringmultiple br
bracackekett pupullll-u-upsps asas wewell.ll. AlAlththououghgh ththesesee araree rararere,, sosomemetitimemess ththeyey araree nenececessssararyy atat smsmalalll to
toururnanamementntss wiwithth ononlyly aa fefeww papartrticicipipatatiningg scschohoolols.s. ItIt isis nonott didifffficiculultt toto hahaveve ththee prprogograramm cons
considerthesemultiderthesemultiplebrackiplebrackets,too.Thebesttwo-bets,too.Thebesttwo-brackracketpull-upetpull-upshouldhavealowerscoreshouldhavealowerscore thantheworstone-bracketpull-up,andsoon. thantheworstone-bracketpull-up,andsoon. Ifstrengthisturnedintoa0.01to1scale(forexample,bydividingteamrankpointsbyn),then Ifstrengthisturnedintoa0.01to1scale(forexample,bydividingteamrankpointsbyn),then thehigheststandarddeviationpossibleis0.5,andthereforethehighestpossiblescorefora thehigheststandarddeviationpossibleis0.5,andthereforethehighestpossiblescorefora matchis0.5.T
matchis0.5.Thescoringformulahescoringformulacouldbeadjustedtocouldbeadjustedto
(1.2)
(1.2) score score
sos
sos..bracketsbrackets ==
2 2 1 1 ! ! A A + + 11 ! ! B B +
+0.50.5
!!
roundsrounds"
"
winswinsA
A
"
"
winswins B B( (
))
Thiswouldshiftthescalingup.After5rounds,forexample,afive-bracketpull-upmatchwould Thiswouldshiftthescalingup.After5rounds,forexample,afive-bracketpull-upmatchwould bescored0.01to0.5,afour-bra
bescored0.01to0.5,afour-bracketpull-upwouldbescoredcketpull-upwouldbescored0.51to1,…,andawithin-brack0.51to1,…,andawithin-bracketet matchwouldbescored2.51to3.Withinbrackets,theprogramwouldstilltrytopushteamsto matchwouldbescored2.51to3.Withinbrackets,theprogramwouldstilltrytopushteamsto equallybalancedschedules.Inotherwords,a3-1teamwithaneasyschedulewouldbepaired equallybalancedschedules.Inotherwords,a3-1teamwithaneasyschedulewouldbepaired againstastrong3-1team,whilea3-1teamwithatoughschedulewouldbepairedagainsta againstastrong3-1team,whilea3-1teamwithatoughschedulewouldbepairedagainsta weak3-1team.Inthisway,astrength-of-schedulepairinghelpsspeedthesortingprocessby weak3-1team.Inthisway,astrength-of-schedulepairinghelpsspeedthesortingprocessby ensuringthatteamsneithercoasttogoodrecordsduetosignificantlyweakerschedulesthan ensuringthatteamsneithercoasttogoodrecordsduetosignificantlyweakerschedulesthan theirbracketnorha
Formula2:Geographyplusinitialrankings Formula2:Geographyplusinitialrankings
Oneconsiderationforsomesta
Oneconsiderationforsomestateandnationaltournamteandnationaltournamentsisgeographicmixing.entsisgeographicmixing.Byrule,someByrule,some tournamentsb
tournamentsblockteamsfromlockteamsfromthesamegeograpthesamegeographicdistrictfrommhicdistrictfrommeeting.(Theseblockseeting.(Theseblockscanbecanbe treatedjustaswithin-schoolblocksare:byassigningthesematchesazero.)Sometournaments treatedjustaswithin-schoolblocksare:byassigningthesematchesazero.)Sometournaments merel
merelywantywant toencouragtoencouragee geogrgeographiaphiccmixinmixing.Ing.In manmanyycasecases,s,statstateeandnationaandnationall tourntournamenamentsts migh
mightalsoliketotalsoliketo runapartialroundrunapartialround robinrobin,whereeveryteamdebat,whereeveryteamdebatessometop,middleessometop,middle,and,and bottombrack
bottombracketteams(perhapetteams(perhapsbasedonrankinsbasedonrankingsfromtheregularseagsfromtheregularseason).son).
Theformula(1.1)outlinedabovealreadyrunsapartialroundrobin.Themeasureofstrength Theformula(1.1)outlinedabovealreadyrunsapartialroundrobin.Themeasureofstrength could
could comecomefromregularseasonfromregularseason pointpointsandnotsandnot varyvaryduringthetournamduringthetournament.ent. AllthatneedstoAllthatneedsto beaddedis
beaddedisameasureofgeographameasureofgeographicdiversityicdiversity.The.Thekeyconsidkeyconsideratioerationisnis thedistanthedistancebetweencebetween teams.Ifgeographicdistrictsareused,teamswithinadistrictcouldbeassigneda“distance”of teams.Ifgeographicdistrictsareused,teamswithinadistrictcouldbeassigneda“distance”of zero;teamsinadjacentdistricts,adistanceof1;andteamsinnon-adjacentdistricts,adistance zero;teamsinadjacentdistricts,adistanceof1;andteamsinnon-adjacentdistricts,adistance of2.Alternatively,theprogramcoulduseeachschool’sZIPcode,findlongitudeandlatitude of2.Alternatively,theprogramcoulduseeachschool’sZIPcode,findlongitudeandlatitude data,andassigntheactualdistancebetweentwoteams.Nomatterhowdistanceismeasured, data,andassigntheactualdistancebetweentwoteams.Nomatterhowdistanceismeasured, themetho
themethodofdof assigassigningning ascoreforgeographascoreforgeographicdiversiticdiversityisyis thesame.thesame. ForapotentialForapotentialmatchofmatchof teamAandteamB,whenteamAhadtheprioropponentCandteamBhadtheprioropponent teamAandteamB,whenteamAhadtheprioropponentCandteamBhadtheprioropponent D,theprogramwouldlookup
D,theprogramwouldlookup d d
AB
AB,,d d AC AC ,,d d AB AB,,d d BC BC ,,d d BD BD
{
{
}}
,thedistancebetweeneachcombination,thedistancebetweeneachcombination ofteamslisted.Thofteamslisted.Theaveragedistancecreaeaveragedistancecreatedbythismatedbythismatchwouldbetchwouldbe
(2.1) (2.1) d d avg avg == 66 2 2 d d AB AB ++ 1 1 d d AC AC ++ 1 1 d d AD AD ++ 1 1 d d BC BC ++ 1 1 d d BD BD whereahigherscoreindicatemoregeographicspread.Theuseoftheharmonicmeanensures whereahigherscoreindicatemoregeographicspread.Theuseoftheharmonicmeanensures thatonlymatcheswherealldistancesarelongreceivehighscores.Thisfactorcouldbeusedin thatonlymatcheswherealldistancesarelongreceivehighscores.Thisfactorcouldbeusedin atournamentonitsownasawaytopairrounds,especiallyforrounds1and2,oritcouldbe atournamentonitsownasawaytopairrounds,especiallyforrounds1and2,oritcouldbe combinedwithafactorforstrengthofschedule. combinedwithafactorforstrengthofschedule. Formula3:Judges Formula3:Judges Themostexcitingapplicationofmatrixoptimizationisjudgeassignment.Ifthetournamenthas Themostexcitingapplicationofmatrixoptimizationisjudgeassignment.Ifthetournamenthas collectedeachteam’sjudgingpreferences,thesecanbeusedtoassigneachdebateamutually collectedeachteam’sjudgingpreferences,thesecanbeusedtoassigneachdebateamutually preferredjudge.Afterthedebatematcheshavepaired,theprogramwouldbuildamatrixof preferredjudge.Afterthedebatematcheshavepaired,theprogramwouldbuildamatrixof eachdebateversuseachjudge.IfteamAgavejudgejapreferenceof
preferenceof
preferenceof j j B B(usuallyona1=mostpreferredto6=strikescale),thenthescorefordebate(usuallyona1=mostpreferredto6=strikescale),thenthescorefordebate ABtobegivenjudgejis
ABtobegivenjudgejis
(3.1)
(3.1) judge judge score score== 11 2
2 j j A A ++22 j j B B
whereahigherscoreindic
whereahigherscoreindicatesamorepreferredjudatesamorepreferredjudge.Ifatournamge.Ifatournamenthassurplusjudgesenthassurplusjudges,then,then several“blank”roundswillneedtobeaddedtomakethematrixsquare.Theseblankrounds several“blank”roundswillneedtobeaddedtomakethematrixsquare.Theseblankrounds shouldbescoredby
shouldbescoredby
(3.2)
(3.2) judge judge score score== 11
3 3++rdsrds remain remain theinverseoftheroundsofcommitmentajudgehasremaining.Thiswillputthejudgeswith theinverseoftheroundsofcommitmentajudgehasremaining.Thiswillputthejudgeswith thefewest
thefewestroundsofcommitroundsofcommitmentmentremaininremainingintoagintoarounroundoff,anditdoff,andit willpushjudgwillpushjudgeswiththeeswiththe mostroundsofcommitmentintoactualdebates.Finally,judgeswhohavealreadyseenateam mostroundsofcommitmentintoactualdebates.Finally,judgeswhohavealreadyseenateam orareblockedfromseeingateamshouldreceivea0forthatround.Thisneedstobeupdated orareblockedfromseeingateamshouldreceivea0forthatround.Thisneedstobeupdated duringatournament. duringatournament. VariantA VariantA The
The aboaboveve metmethodhod conconsidsidereeredd thethe prepreferferencenceses ofof eveeveryry teateamm equequallally.y. IfIf thethe goagoall isis toto givgivee higherprioritytoteamsinbreakrounds,thentheprogramneedstoincludeamultiplierin(3.1) higherprioritytoteamsinbreakrounds,thentheprogramneedstoincludeamultiplierin(3.1) toinflate
toinflatethethe judgjudgescoresfortheseteamsescoresfortheseteams.Forexample.Forexample,thisschemecould,thisschemecouldwork:eitherteamwork:eitherteam hastwo
hastwolosses=4points,eithlosses=4points,eitherteamhaserteamhasonelossbutneithonelossbutneitherhastwo=3points,boterhastwo=3points,bothteamshteams areundefe
areundefeatedated== 2points,andneith2points,andneitherteamcanmakeerteamcanmakeeliminatieliminationroundsonrounds== 1point.The(3.1)1point.The(3.1) formulawouldbemodifiedto
formulawouldbemodifiedto
(3.3)
(3.3) judge judge score score== multiplier multiplier 2
2jj A A++22jj B B
VariantB VariantB
Ofcourse,itisalsopossibletoaccomp
Ofcourse,itisalsopossibletoaccomplishsimilargoalswithoutteamlishsimilargoalswithoutteam’sjudgepreferences.Ifthe’sjudgepreferences.Ifthe tournamentranksallthejudgesonanexperiencescale,say,10pointsforveryexperiencedto0 tournamentranksallthejudgesonanexperiencescale,say,10pointsforveryexperiencedto0 pointsforanovicejudg
(3.4)
(3.4) judge judge score score==multiplier multiplier
!!
expexperienceerienceItwouldmakenosensetoassignthesescoreswithonlytheexperiencevariable,sinceevery Itwouldmakenosensetoassignthesescoreswithonlytheexperiencevariable,sinceevery roundajudgecouldseewouldhavethesamescore. roundajudgecouldseewouldhavethesamescore. VariantC VariantC Perhapsthemostexcitingoptionisthatalldebatesinalldivisionsofalldebateeventscouldbe Perhapsthemostexcitingoptionisthatalldebatesinalldivisionsofalldebateeventscouldbe consideredsimultaneously.Theprogramwouldcreateamatrixwithalldebatematchesandall consideredsimultaneously.Theprogramwouldcreateamatrixwithalldebatematchesandall judges.The
judges.Thenewadditionalnewadditionalvariablewouldvariablewould bethebethe judge’sappropriatenessforjudge’sappropriatenessfor eachdivisioneachdivision forfor eachevent.Perhapsaninexperiencedjudgewouldbegivenascoreof10(mostappropriate) eachevent.Perhapsaninexperiencedjudgewouldbegivenascoreof10(mostappropriate) fornoviceLincolnDouglasdebatebutascoreof0(inappropriate)forvarsitypolicydebate.This fornoviceLincolnDouglasdebatebutascoreof0(inappropriate)forvarsitypolicydebate.This is,inessence,thesamethingasanexperiencescore,butgivenforeverydivisionandevent. is,inessence,thesamethingasanexperiencescore,butgivenforeverydivisionandevent. Theprogramwou
Theprogramwouldusejudgesinthemostapldusejudgesinthemostappropriatedivisionsandeventswpropriatedivisionsandeventswheretheycanstillheretheycanstill judgedebates,
judgedebates, butibutifftheyaretheyare blockedagainstblockedagainst allallthetheteamsinteamsin thatdivisionthatdivision ororevent,event,thenthethenthe programwoulds
programwouldseamlesslyslipthejudgeamlesslyslipthejudgesintothenextbestdesintothenextbestdivisionorevent.ivisionorevent.
Tomaketheformulacomplete,theprogramwouldalsoneedtoconsiderwhichdivisionand Tomaketheformulacomplete,theprogramwouldalsoneedtoconsiderwhichdivisionand eventhadthehighestpriorityforappropriatejudges.Giventhespeedandtechnicaldemands, eventhadthehighestpriorityforappropriatejudges.Giventhespeedandtechnicaldemands, thehighestpriorityevent
thehighestpriorityeventwouldmostlikelybwouldmostlikelybepolicydebate.Thepolicydebate.Theformulawouldtheneformulawouldthenbebe
(3.5)
(3.5) judge judge score score== prioritypriority
div
div..evev
!!
appropriatenessappropriatenessThiscouldbecombinedwiththemultiplierforbreakroundsorevenmutualpreferencescores, Thiscouldbecombinedwiththemultiplierforbreakroundsorevenmutualpreferencescores, aslongasthesewereusedinalldivisionsofallevents. aslongasthesewereusedinalldivisionsofallevents. VariantD VariantD Thesamemethodcanbeusedforassigningjudgingpanels.Thejudgeswouldbeassignedone Thesamemethodcanbeusedforassigningjudgingpanels.Thejudgeswouldbeassignedone byone.Ifthetournamentusesthree-judgepanels,theprogramwoulddividethejudgesinto byone.Ifthetournamentusesthree-judgepanels,theprogramwoulddividethejudgesinto thr
threepoolseepools anandd aa foufourthrth setset ofunusofunusedjudgeedjudges.Thepoolss.Thepools coucouldbeldbe divdivideidedindin ananyy numnumberber ofof ways
ways::geoggeographyraphy,experienc,experience,stylistice,stylistic prefepreferencerences,sex,s,sex,age,etc.Eachpoolwouldbeage,etc.Eachpoolwouldbe assiassignedgned usingthematrixoptimizationmethod.
usingthematrixoptimizationmethod.
Formula4:IEs Formula4:IEs
Itisalsopossibletousema
ItisalsopossibletousematrixoptimizationtosetutrixoptimizationtosetupIEpanelsusingmupIEpanelsusingmultiplevariables.TheIErsltiplevariables.TheIErs wouldneedtobeaddedtothepanelonebyone.
Section3:ABetterMeasureofTeamStrength Section3:ABetterMeasureofTeamStrength
Thebette
ThebettermeasurrmeasureofeofteamteamstrengthstrengthII havecomhavecomeupwithiseupwithis weighweightedwins.tedwins.IfIf teamteamAA defeadefeatsts teamsB,C,andD
teamsB,C,andD,teamA’sweightedw,teamA’sweightedwinswouldbe3+winsofB+inswouldbe3+winsofB+winsofC+winsofD.IwinsofC+winsofD.Inthisnthis way,teamAearnsadditionalwinsforthestrengthofitsdefeatedopponents.IfteamAlosesto way,teamAearnsadditionalwinsforthestrengthofitsdefeatedopponents.IfteamAlosesto teamsEandF,teamA’sweightedlosseswouldbe2+lossesofE+lossesofF.TeamAearnsits teamsEandF,teamA’sweightedlosseswouldbe2+lossesofE+lossesofF.TeamAearnsits additionallossesifitlosestoweakopponents.Theratio
additionallossesifitlosestoweakopponents.Theratio ww
.
.
winswins ww
.
.
winswins++ww.
.
losseslossesmeasuresateam’s measuresateam’s strengthona0to1scale. strengthona0to1scale. TheweightedwinsratioisNOTthesamethingasopponentwins.Opponentwinsmeasuresthe TheweightedwinsratioisNOTthesamethingasopponentwins.Opponentwinsmeasuresthe strengthofateam’sschedule;weightedwinsconsidersthestrengthofopponentstoadjustthe strengthofateam’sschedule;weightedwinsconsidersthestrengthofopponentstoadjustthe measureofateam’sstrength. measureofateam’sstrength.