Note: This section is now in read-only mode.
Please use our new community site for future posts.

Back To Board

A Rough Analysis In Statistical Accuracy

Posted By Eric

I've been a participant in a PB league for three somewhat-successful years now and while I have noticed that some of my players under- or overperformed I've never really taken the time to analyze just how much the differences were. Below is a table comparing the players on my team's real-life stats (mlb) with their simmed stats (pb). I only used stats where a player had 100 sim ab or bf. Also note that this is a real-time league. I have nothing but intuition for proof that a real-time league should be a bit less accurate than a standard league because of the fluctuating base stats but I'm going to make that assumption.

Player vsR(mlb) vsR(pb) Diff vsL(mlb) vsL(pb) Diff
Jorge Posada 944 824 -120(13%) 940 887 -057(6%)
Mike Sweeney 905 903 -002(0%) 1023 886 -137(13%)
Paul Konerko 856 937 +081(9%) 794 658 -136(17%)
Travis Lee 745 711 -034(5%)
Jeff Kent 1032 1079 +047(4%) 983 1054 +071(7%)
Julio Lugo 832 848 +015(2%)
Adrian Beltre 837 767 -070(8%) 827 863 +036(4%)
Jose Valentin 861 885 +024(3%)
Placido Polanco 718 840 +122(15%) 892 811 -081(9%)
Bernie Williams 994 915 -079(8%) 875 816 -059(7%)
Todd Hollandsworth 815 859 +044(5%)
Raul Mondesi 834 752 -082(10%) 927 822 -105(11%)

Average % diff 6.8% 9.3%

Against righties it looks like my players were fairly close to expected performance. An average of 6.8% difference seems reasonable to me. Polanco's +15% is most likely attributed to his hot start and my heavy usage of him during that time. Posada's -13% is less easy to explain. A guess might be that there is a slight performance hit when a player plays while in a gold-Ready status. Posada, Williams and Mondesi seemed to be my biggest under-performers and they all played a portion of the season in gold-Ready. Just a thought - haven't really researched that theory.

Against lefties the difference in performance is larger but still within the 10% margin of error mentioned above. Hmmm vsR AB = 4166, vsL = 1604. vsR %Diff = 6.8%, vsL = 9.3%. Sample-size in action? :o)

Pitcher vsR(mlb) vsR(pb) Diff vsL(mlb) vsL(pb) Diff
Scott Elarton 757 794 +037(5%) 832 843 +011(1%)
Matt Clement 677 812 +135(17%) 806 767 -039(5%)
Jeff Suppan 866 870 +004(0%) 819 795 -024(3%)
Dustin Hermanson 715 760 +045(6%) 965 1207 +242(20%)
Carl Pavano 522 636 +114(18%) 915 1102 +187(17%)
Tony Armas Jr 547 587 +040(7%) 873 783 -090(10%)
Alex Fernandez 838 912 +074(8%) 741 708 -033(4%)
Danny Graves 662 688 +026(4%) 719 824 +105(13%)
Jeff Tam 526 441 -085(16%) 914 1089 +175(16%)
Vicente Padilla 627 854 +227(27%) 941 774 -167(18%)
Danny Patterson 770 869 +099(11%) 768 766 -002(0%)
Lance Painter 724 734 +010(1%) 840 912 +072(8%)

Average % diff 10.0% 9.6%

Yeah, yeah - not exactly the Braves rotation here but we'll be good...eventually...if we can stay healthy. The % of difference is much higher for my pitching than my hitting and mostly for the worst. Rampant platooning and a possible league tendency toward offense is the most likely explanation for this. Alot of my pitchers had substantial platoon differences and opposing managers tended to take advantage of this.