THE BOOK--Playing The Percentages In Baseball

Tuesday, July 12, 2011

Felix v Weaver and/or GB v FB

Bill James is having a discussion on groundball and flyball pitchers, etc.  Lots of talk, and this is just one clip from it:

...What is critical, then, is not whether a pitcher gives up fly balls or ground balls, but whether he gives up Line Drives--and both ground ball pitchers and fly ball pitchers give up pretty much the same percentage of Line Drives.  That’s all that really matters.

Of course, if I were pitching to Dustin Pedroia in a Dunkin’ Doughnuts commercial, I would give up 95% line drives, so Pedroia would hit .700 or .750 against me--but this is only because I am not a major league pitcher.  If you give up too many line drives, you’re not going to be in the majors, so you’re irrelevant to the issue.

Also, I shouldn’t blame statisticians for the belief that ground ball rates are important; it actually didn’t originate with us.  What really happened is more like this:  that analysts proclaimed that getting ground balls was terribly, terribly important for a pitcher, and then staticiams realized that they could measure how many ground balls each pitcher gets.

What you have, then, is statisticians in possession of information that analysts have long proclaimed to represent a tremendously important distinction among pitchers.  What happens then depends on whether you have a conservative statistical analyst or a radical one.  If you’re a conservative statistical analyst, like Elias, what happens then is that you proclaim that you have this tremendously important statistical information.  If you’re a radical statistical analyst, like me, what happens then is that you say, “Well, I have this data that you guys think is important, but actually, it doesn’t seem to have very much value.” But it is much easier to say that you have this really important data, rather than that you have some more data but it appears to be meaningless, so. . .that’s what most people prefer to believe, or did prefer to believe, before Voros established very clearly that the other camp had a better argument.

My response:

This might be helpful to your readers.  This is the run values, by event:
+1.4 FB (only HR)
+0.3 walks, line drives
+0.05 FB (including HR)
-0.1 GB, FB (sans HR)
-0.3 so, pop ups

So, depending how you include or exclude your air balls (with pops? with liners? with HR?), you’ll get a different answer as to whether there’s any value in separating grounders from the other events.

A good comparison is Felix (groundball pitcher) and Weaver (flyball pitcher).  Both have a similar K-BB differential per PA.  But Weaver has lots more HR [ed note: they’ve both given up exactly 106 HR, but Weaver has faced 1000 fewer batters].  But his BABIP is much lower.  Overall, we see that Weaver has given up in his career 3.52 runs per 9 IP (we call this RA9).  Felix’s RA9 is 3.62.  (If you go by ERA, Felix has the advantage the other way: 3.19 to 3.32.)

All to say is that just sticking with K minus (BB-IBB+HB) per (PA-IBB), and we have Felix at +.14, and Weaver at +.14.  All the other extra information we have from Felix being a GB pitcher and Weaver being a FB pitcher cancels out. 

Which is basically your main point I think.