BatterHandednessProjectHerbWilsonContentsIntroduction1DataUpload1Join - PDF document

1 BatterHandednessProject-HerbWilsonConten
BatterHandednessProject-HerbWilsonContentsIntroduction1DataUpload1JoinwithLahmandatabase1ChangeinProportionofRHPPAbyYear2MLB-widedierencesinBAagainstLHPvs.RHP2EquilibrationofBattingAverage4Individualvariationinsplits4LogisticregressionusingBattingAveragesplits.7LogisticregressionsusingweightedOn-baseAverage(wOBA)11SummaryofResults17IntroductionThisprojectisanexplorationofbatterperformanceagainstlike-handedandopposite-handedpitchers.Wehavelongknownthat,collectively,battershavehigherbattingaveragesagainstopposite-handedpitchers.Dierencesinperformanceagainstleft-handedversusright-han

2 dedpitcherswillbereferredtoassplits.Theg
dedpitcherswillbereferredtoassplits.Thegeneralityofsplitsfavoringopposite-handedpitchersmasksvariabilityinthemagnitudeofbattingsplitsamongbattersandvariabilityinsplitsforasingleplayeramongseasons.Inthiscontribution,Itesttheadequacyofsplitvaluesinpredictingbatterhandednessandthenexamineindividualvariabilitytoexploresomeofthenuancesoftherelationships.ThedatausedprimarilycomefromRetrosheeteventsdatawiththeLahmandatasetbeingusedforsomebiographicalinformationsuchasfullname.IusedtheRprogramminglanguageforallstatisticaltestingandforthecreationofthegraphics.Acopyofthecodeisavailableonrequestbyconta

3 ctingmeatwhwilson@colby.eduDataUploadThe
ctingmeatwhwilson@colby.eduDataUploadTheRetrosheeteventsdataaregivenbyyear.Therststepintheanalysisistouploaddataframesforeachyear,thenusetherbindfunctiontostitchdatasetstogethertomakeadataframeandusethefunctioncolnamestoaddcolumnnames.Forthisstudy,Icombinedthedatafrom1955until2017,yieldingadataframewithnearly10millionrows.JoinwithLahmandatabaseNextIusetheLahmanMasterdatabasetogetrstandlastnamesofplayersusingtheleft_joinfunctiononRetro_IDnumber.Ithenpastethelastnamesandrstnamestogethertogetasingleeldforplayername.Finally,IlteroutjoinedMastercolumnsIdonotneed.1 Change

4 inProportionofRHPPAbyYearIbeginbylooking
inProportionofRHPPAbyYearIbeginbylookingatthenumberofPlateAppearances(PA)againstleft-handedpitcher(LHP)versusright-handedpitchers(RHP)overthe63yearsinthisdataframe.Irstcreateacolumnwiththeyear.Then,Ilteroutanyeventsthatdonotpertaintothebatter(e.g.,stolenbase,passedballorwildpitchwhilebatting)tocreateaPAcolumn.Switch-hitterswereremovedfromtheanalysis. MLB-widedierencesinBAagainstLHPvs.RHPInowexaminethemagnitudeofbattersplitsagainstLHPandRHPforeachyearofthestudy.IconsiderLHPrstandthenRHP.HerearethemeanbattersplitsbyyearagainstLHP.2 HerearethemeanbattersplitsbyyearagainstRH

5 P.3 EquilibrationofBattingAverageToexami
P.3 EquilibrationofBattingAverageToexamineindividualvariabilityinsplits,IneedtoconsideronlyplayerswithaminimumnumberofABsagainstLHPandRHPinaseasontoavoidbiasedBattingAveragesbecauseoflowsamplesize.AsurveyofsomerandomlychosenplaysindicatesthatBA,BAagainstLHPandBAagainstRHPbegintostabilizeafter100AB.Therefore,Irequirethatallbattersmusthaveatleast100ABsagainstLHPand100ABsagainstRHPtobeincludedinananalysis.IndividualvariationinsplitsThenextthreehistogramsshowthedistributionofLHbattersplitsagainstLHPandRHPandRHbattersplitsagainstLHPandRHP.Irandomlychosethreeseasonstopresent.Aclearpatternisthatth

6 eproportionofLHbatterswithoppositesplits
eproportionofLHbatterswithoppositesplits(higherBAagainstLHP)islowerthantheproportionofRHbatterswithoppositesplits.##Joining,by="batter_RetroID"4 ##Joining,by="batter_RetroID"5 ##Joining,by="batter_RetroID"6 LogisticregressionusingBattingAveragesplits.##Joining,by="batter_RetroID"####Call:glm(formula=bat_code~lhp_rhp,family=binomial,data=full_splits)####Coefficients:##(Intercept)lhp_rhp##1.56128.355####DegreesofFreedom:152Total(i.e.Null);151Residual##NullDeviance:162.1##ResidualDeviance:126AIC:130####Call:##glm(formula=bat_code~lhp_rhp,family=binomial,data=full_splits)####DevianceResiduals:#

7 #Min1QMedian3QMax##-2.66040.12440.42480.
#Min1QMedian3QMax##-2.66040.12440.42480.63061.9995####Coefficients:##EstimateStd.Errorzvalue�Pr(|z|)##(Intercept)1.56130.25646.0901.13e-09***##lhp_rhp28.35555.78604.9019.55e-07***7 ##---##Signif.codes:0'***'0.001'**'0.01'*'0.05'.'0.1''1####(Dispersionparameterforbinomialfamilytakentobe1)####Nulldeviance:162.09on152degreesoffreedom##Residualdeviance:126.03on151degreesoffreedom##AIC:130.03####NumberofFisherScoringiterations:5 ##Joining,by="batter_RetroID"####Call:glm(formula=bat_code~lhp_rhp,family=binomial,data=full_splits)####Coefficients:##(Intercept)lhp_rhp##0.92817.282####Degreeso

8 fFreedom:155Total(i.e.Null);154Residual#
fFreedom:155Total(i.e.Null);154Residual##NullDeviance:192.6##ResidualDeviance:167.8AIC:171.8####Call:##glm(formula=bat_code~lhp_rhp,family=binomial,data=full_splits)##8 ##DevianceResiduals:##Min1QMedian3QMax##-2.2596-0.96710.59510.79151.5595####Coefficients:##EstimateStd.Errorzvalue�Pr(|z|)##(Intercept)0.92800.19584.7392.14e-06***##lhp_rhp17.28243.86434.4727.74e-06***##---##Signif.codes:0'***'0.001'**'0.01'*'0.05'.'0.1''1####(Dispersionparameterforbinomialfamilytakentobe1)####Nulldeviance:192.58on155degreesoffreedom##Residualdeviance:167.77on154degreesoffreedom##AIC:171.77####Numbero

9 fFisherScoringiterations:4 ##Joining,by=
fFisherScoringiterations:4 ##Joining,by="batter_RetroID"####Call:glm(formula=bat_code~lhp_rhp,family=binomial,data=full_splits)####Coefficients:##(Intercept)lhp_rhp##1.59640.488##9 ##DegreesofFreedom:113Total(i.e.Null);112Residual##NullDeviance:119.9##ResidualDeviance:80.07AIC:84.07####Call:##glm(formula=bat_code~lhp_rhp,family=binomial,data=full_splits)####DevianceResiduals:##Min1QMedian3QMax##-3.008750.076910.309800.542841.63588####Coefficients:##EstimateStd.Errorzvalue�Pr(|z|)##(Intercept)1.59580.32364.9318.18e-07***##lhp_rhp40.48798.56434.7282.27e-06***##---##Signif.codes:0'***'0

10 .001'**'0.01'*'0.05'.'0.1''1####(Dispers
.001'**'0.01'*'0.05'.'0.1''1####(Dispersionparameterforbinomialfamilytakentobe1)####Nulldeviance:119.932on113degreesoffreedom##Residualdeviance:80.072on112degreesoffreedom##AIC:84.072####NumberofFisherScoringiterations:6 10 LogisticregressionsusingweightedOn-baseAverage(wOBA)BattingAverageisarathercoarsemeasureofbattingperformance.Icalculatedweightedon-baseaveragesforallplayers.wOBAgivesamorenuancedassessmentofbattingperformancebyincorporatingwalksandpower.Iuselogisticregressiononthesameyearsanalyzedabove(1970,1999,1961)withbattingaveragesplitstocomparethestrengthofthemodels.LogisticRegress

11 ionforwOBAsplitsforthe1970MLBseason##Joi
ionforwOBAsplitsforthe1970MLBseason##Joining,by="batter_RetroID"####Call:glm(formula=bat_code~lhp_rhp,family=binomial,data=full_splits)####Coefficients:##(Intercept)lhp_rhp##1.56128.355####DegreesofFreedom:152Total(i.e.Null);151Residual##NullDeviance:162.1##ResidualDeviance:126AIC:130####Call:##glm(formula=bat_code~lhp_rhp,family=binomial,data=full_splits)####DevianceResiduals:##Min1QMedian3QMax##-2.66040.12440.42480.63061.9995####Coefficients:##EstimateStd.Errorzvalue�Pr(|z|)##(Intercept)1.56130.25646.0901.13e-09***##lhp_rhp28.35555.78604.9019.55e-07***##---##Signif.codes:0'***'0.00

12 1'**'0.01'*'0.05'.'0.1''1####(Dispersion
1'**'0.01'*'0.05'.'0.1''1####(Dispersionparameterforbinomialfamilytakentobe1)####Nulldeviance:162.09on152degreesoffreedom##Residualdeviance:126.03on151degreesoffreedom##AIC:130.03####NumberofFisherScoringiterations:511 LogisticRegressionforwOBAsplitsforthe1999MLBseason12 ##Joining,by="batter_RetroID"####Call:glm(formula=bat_code~lhp_rhp,family=binomial,data=full_splits)####Coefficients:##(Intercept)lhp_rhp##0.92817.282####DegreesofFreedom:155Total(i.e.Null);154Residual##NullDeviance:192.6##ResidualDeviance:167.8AIC:171.8####Call:##glm(formula=bat_code~lhp_rhp,family=binomial,data=full_split

13 s)####DevianceResiduals:##Min1QMedian3QM
s)####DevianceResiduals:##Min1QMedian3QMax##-2.2596-0.96710.59510.79151.5595####Coefficients:##EstimateStd.Errorzvalue�Pr(|z|)##(Intercept)0.92800.19584.7392.14e-06***##lhp_rhp17.28243.86434.4727.74e-06***##---##Signif.codes:0'***'0.001'**'0.01'*'0.05'.'0.1''1####(Dispersionparameterforbinomialfamilytakentobe1)####Nulldeviance:192.58on155degreesoffreedom##Residualdeviance:167.77on154degreesoffreedom##AIC:171.77####NumberofFisherScoringiterations:413 LogisticRegressionforwOBAsplitsforthe1961MLBseason14 ##Joining,by="batter_RetroID"####Call:glm(formula=bat_code~lhp_rhp,family=binomial,

14 data=full_splits)####Coefficients:##(Int
data=full_splits)####Coefficients:##(Intercept)lhp_rhp##1.59640.488####DegreesofFreedom:113Total(i.e.Null);112Residual##NullDeviance:119.9##ResidualDeviance:80.07AIC:84.07####Call:##glm(formula=bat_code~lhp_rhp,family=binomial,data=full_splits)####DevianceResiduals:##Min1QMedian3QMax##-3.008750.076910.309800.542841.63588####Coefficients:##EstimateStd.Errorzvalue�Pr(|z|)##(Intercept)1.59580.32364.9318.18e-07***##lhp_rhp40.48798.56434.7282.27e-06***##---##Signif.codes:0'***'0.001'**'0.01'*'0.05'.'0.1''1####(Dispersionparameterforbinomialfamilytakentobe1)####Nulldeviance:119.932on113deg

15 reesoffreedom##Residualdeviance:80.072on
reesoffreedom##Residualdeviance:80.072on112degreesoffreedom##AIC:84.072####NumberofFisherScoringiterations:615 16 SummaryofResults1)MLBbattersfaceright-handedpitchersmorethantwiceasoftenasleft-handedpitchers.2)Averagebattingsplitshavevariedaroundanequilibriumvalueovertheperiod1955-2017forleft-handedbatters.Weseeadecreaseinaveragesplitsoverthepost20yearsforright-handedbatters.Thereisenoughvariationtonecessitateconsiderationofindividualseasonsratherthancombiningallthedata.3)Attheindividuallevel,asmallerproportionofleft-handedbattersareoppositehitters(hittingLHPbetterthanRHP)comparedtothepropo

16 rtionofright-handedbatterswhoareopposite
rtionofright-handedbatterswhoareoppositehitters(hittingRHPbetterthanLHP).GreaterexperienceagainstRHPmayexplainthispatternforright-handedbatters.4)Logisticregression,atechniquetopredictbinaryoutcomes(inthiscase,left-handedorright-handedbatter),isappliedtoeachofthe63seasons,testingtheabilityofBAsplitstopredicthandednessofabatter.Ineverycase,thelogisticregressionwasstatisticallysignicant(p0.05).However,theregressionsshowedaconsistentasymmetry.Evensplits(BAversusLHP=BAversusRHP)predictsaright-handedbatter.Thisshiftintheregressioncurveisexplainedbythedearthofoppositeleft-handedbatters.5)Top

17 rovideamoregranularanalysis,Iperformedlo
rovideamoregranularanalysis,Iperformedlogisticregressionsoneachofthe63seasonsusingsplitsinweightedOn-baseAverage(wOBA).Again,eachofthe63regressionswasstatisticallysignicant.ComparisonoftheAICvalues(AkaikeInformationCriterion)betweentheBAandwOBAregressionsconsistentlyshowedwOBAprovidedabettert(lowerAICvalues).6)Sincebattersareexposedtoleft-handedpitcherslessoftenthanright-handedpitchers,right-handedbattersgetmorepracticeagainstsame-handedpitchingthanleft-handedbatters.Thisresultmayexplainthepaucityofleft-handedbatterswithoppositesplitsandthegreaterleft-rightsplitsforleft-handedbatt