THE BOOK--Playing The Percentages In Baseball

MGL, can you show us the top 10 and bottom 10, and, if they’re not there, Mariano Rivera and Barry Zito.

Really cool results, MGL.  I’ve been thinking how something like this might be done for a while, and I like your implementation.  This is pretty much what I was expecting, result-wise given Dave Allen’s numbers.  Sniff test good and all that.  Could you report t-stats?

What happens if you run an interaction term for fastball and slider?  I think you’ll like the reason why I’m suggesting it.  Pitchers with fastball-slider combinations presumably could be trying to mix their pitches in such a way that the run values are equal on any given pitch.  It could be that the fastball has a higher platoon split for fastball-slider pitchers, while not as much for fastball-changeup pitchers.

I’m guessing the results would be the same for SP and RP separately, but if you have the data, it might be worth checking or at least adding in a variable for

MGL,

Good stuff. 

The run value differentials by pitch type is from Max Marchi.

Tango, top and bottom 10 for what?  Matt, I can post the file on Google docs.

Here is the data file for the regression on Google docs.

Remember each line is one pitcher season.  All pitchers in any season from 06-09 is included if he had at least 100 PA versus RHB.  Starters and relievers or starter and relief appearances are not distinguished.

http://spreadsheets.google.com/ccc?key=0AgwSI9hFVCZIdHVwWlpPam13WEJaakN5c3ZIZW9pZFE&hl=en#gid=0

Here is Zito, 2009:

zitob001 L 610 189 0.492 0.186 0 0.182 0.14 0 1.2

What that is is 49.2% fastballs, 18.6 sliders, no cutters, 18.2% curve balls, 14% change ups, and no splitters.  His OPS platoon ratio in 2009 was 1.20, about normal for a lefty.

In 06-08, his platoon OPS ratio was .97, 1.01, and 1.38, for around a 1.14 average for the 4 years.  Even though he comes over the top and throws plenty of curves and changeups, he still throws 68% fastballs and sliders, which gives him a normal ratio.

Rivera, 2009

rivem002 R 126 130 0.07 0 0.93 0 0 0 0.89

7% fastballs, no sliders, 93% cutters and that’s it.  He has a .89 (reverse) split, for OPS.

2006-2008 his platoon ratios were .73, .92, .79.  According to Marchi, the cutter has a platoon differential (in run value) of .41, a little less than a slider.  That means two things for Rivera:  One, you should regress his sample platoon ratio towards one that is maybe 1.1 or 1.2. How much to regress of course depends on the sample size. If you are using his whole career or the least 5 or 10 years, you probably don’t regress a whole lot.  Two, if you watch him pitch, you can probably see that his reverse split is NOT a fluke because of the way he uses his cutter against lefties (by jamming them).  Most RHP do not use the cutter much against lefties and if they do, they use it like a slider, under the hands or back-dooring it (Rivera sometimes back-doors the cutter against lefties).

Here are the top 10 lefties (with at least 300 PAvRHB in one year) with the highest % of changeups for that year:

glavine
glavine
hamels
hamels
santana
hamels
james, c
rogers
santana
james, c

They throw the change 32.8% of the time.  They face 77% RHB. The average LHP faces 74% RHB, so opposing managers don’t seem to bat more LHB than usual against these guys - in fact, less.

Their collective platoon ratio for those years, OPS, is .998.

For righties:

harden
volqez
shields
maddux
shields
jurrjens
campillo
jurrjens
shields
silva

They throw the change 27.7% of the time.  They face 51% RHB. The average RHP faces 50% RHB, so opposing managers don’t seem to bat more RHB than usual against these guys.

The average OPS platoon ratio for them is 1.05.

Danks throws around 20% change ups and has had a reverse platoon ratio for 3 years (07-09) of around .97.

Marcum throws around 20% changeups and has had platoon splits of 102, 121, and 119 for 07-09.

I have no problem hitting lefties against Danks.  I probably would not hit too many righties against Marcum, unless they are switch hitters and they prefer that side.  Lefties throwing change ups seem to have more of a reverse split than righties.

Plus, it all depends on the batter.  If he has a small split himself, then you might want to same side him against a pitcher with a small, but not necessarily reverse split.  The net effect of platoon splits between a batter and pitcher is basically the odds ratio of both of their splits.

As I said, you don’t want to force all of your switch hitters to hit same side versus pitchers with small or reverse splits. You want to take into consideration their own splits - i.e. which is their best side.  And you definitely need to consider the fact that they rarely hit from the same side.  There HAS to be penalty for that.  Whether it is small or large, I have no idea.

I think that Maddon (or whoever thought of it) is on to something, but I also think that he is painting with too broad of a brush.