from Amazon
Friday, November 09, 2007
Personal Platoon Splits
Nate discusses platoon splits.
Secondly, these platoon splits will allow us to adjust for changes to player context. What I mean by that is, if a left-handed hitter moves from a division (like the NL Central) without very much left-handed pitching, and to a division (like the AL Central) with plenty of good left-handed pitching, we’ll able able to account for that in his projection.
Sounds both reasonable and unreasonable. So, let’s go to the data and see how many PA LHH faced:
Team RHP LHP %LHP
Hou 1537 207 12%
Stl 1757 319 15%
Chc 1789 312 15%
Mil 1810 413 19%
Pit 1603 389 20%
Cin 1820 616 25%
TOT 10316 2256 18%
Team RHP LHP %LHP
Det 1824 219 11%
Min 2458 562 19%
Cle 2200 617 22%
Chw 2099 622 23%
Kcr 2100 649 24%
TOT 10681 2669 20%
All data from baseball-reference.com. All transcription errors mine.
His extreme example shows an aggregate of 18% in one division to 20% in another division. So, I think his example is unreasonable. I get what Nate is saying, but I wouldn’t look at it in terms of “divisional”. It’s very team-centric. It is of course a great idea to determine how many PA you will get against LHP and RHP, both at the team level, and at the individual level (platoons, etc). So, his position is very reasonable.
The other thing is he says:
How much weight should we give to a player’s previous platoon split history, as opposed to simply applying league average platoon splits?
Andy Dolphin did work on this in The Book, and Andy did give the equation as to how much to regress toward a player’s platoon splits. I don’t have The Book handy (I really ought to keep the pdf version on my flash drive). I think Andy’s regression equation was 2000/(2000+lhPA). Don’t quote me though. So, if you have 1000 LHP PA, you regress 67% toward the mean, which matches Nate’s data. But, Nate’s data pretty much flattens out, which doesn’t seem right. Can someone pull out The Book for me?