THE BOOK--Playing The Percentages In Baseball

Thanks for linking to the post, Tango.

It’s a great study.  I especially like the break down right-vs-left, pitch selection, etc.

Still, there’s part of me who thinks that given a large enough sample of pitchers, some of them will be below average even for an extended period and some will be above average for an extended period just by luck.  After all, flip 1000 coins 1000 times, and you’ll get 5 coins with over 60% head and 5 coins with over 60% tails.  Has the entire population of pitchers been looked at?  Are we sure we’re not just looking a random bell curve?

Geri: are you talking about this?

http://www.tangotiger.net/dipsbands.html

Josh, in the graph showing Cain’s babip vs the league by type of pitch, was that Cain’s babip on changeups vs the league’s babip on changeups, etc?

Geri, you are correct that the whole population is what matters. However, if you accept the binomial model for randomness, the results for pitcher BABIP are spread out wider than what we’d expect from chance alone. I estimate somewhere around 1000 balls in play is about where you’d take the midpoint between a pitcher’s observed BABIP and the league average. That’s using road data only. (Is that close to what you get, Tango?)

I believe I’ve shown it was 3700 BIP.  Though I seem to remember also getting something close to 2000 BIP in another study.

1000 is very very low to get r=.50.  I’d question any number under 1500.

Derek Carty looked at this several weeks back… I linked to it here.  He had the number as well, so check the archives…

Brian,

against left handed batters, Cain breaks down like this:

CH: 0.236
CU: 0.256
FF: 0.265
FT: 0.265
SL: 0.340

for the sample of pitches (thrown by right handers) against left handed batters:

CH: 0.283
CU: 0.306
FF: 0.288
FT: 0.309
SL: 0.318