THE BOOK--Playing The Percentages In Baseball

Saturday, March 01, 2008

More Pitch f/x Data

Joe P. Sheehan Sheehan of the Baseball Analysts looks at the (linear weights) run values of pitches in various locations.  He only tackles a few situations.

I am not really sure what to make of the data/charts.  In the first chart, there sure isn’t any evidence that balls lower in the zone (or out of the zone) are better than up in the zone.  Aren’t pitchers supposed to keep everything down, more or less?

I wish these guys when using linear weights would tell us several things:  One, exactly how they weight all the possible events.  Two, the average lwts of all the pitches in their (the pitch f/x) data base.  Joe Walsh says that the average pitch among all the pitches in the pitch f/x database is negative because he thinks that for some reason the parks that have the pitch f/x system tend to be pitchers parks and the pitchers in the database tend to be better pitchers (I am not sure why that would be).  I told him that I thought that his values for the events were too negative.  Three, I really wish when they give us the linear weights values of, say, a fastball in various situations, such as these location charts, that they would tell us the lwts value of the average fastball (in their database I guess) so we have some reference as to whether a -.03 per pitch is a good pitch or a bad pitch, compared to an average pitch of that type.  According to Walsh, the average lwts value of the various pitches (fastball, curve, etc.) is quite different.

Again, I don’t know what to make of these charts and numbers that Joe presents.  So much depends on what is going on at each count, how often a particular pitch is thrown, etc.  For example, just telling us the value of a pitcher’s fastball, curveball, or whatever, tells us very little about the quality of that pitch unless we know how often that pitcher throws that pitch at the various counts.  For example, if I have a great fastball but I throw it all the time, it won’t have a great or even good lwts value (unless I am Mo Rivera I guess).  If I have just an average fastball but I throw it less often especially in fastball counts, it might have a great lwts value.  Similarly, if I have a curveball that I throw in fastball counts, such as when I am behind, if I can throw strikes with it, it is going to have great value even if it is not such a good curveball.  So giving us run values of pitches for individual pitchers means virtually nothing unless we know the context.  But that rant was mostly for Joe (Walsh).  Not that I am criticizing his work.  I am not.  I am just pointing out the complexity and difficulty of analyzing this kind of data.

Anyway, back to Joe and this article, if you look at the 3-0 count chart, Joes says this about pitches in the strike zone:

Somewhat counter intuitively, even though hitters “know” a strike is coming, pitches thrown in the strike zone in 3&0 counts still favor pitchers. This just speaks to how hard hitting actually is.

He is referring to the fact that pitches in the strike zone on a 3-0 count favor the pitcher.  Sorry, Joe, it is not because hitting is hard and “despite the fact that hitters know a strike is probably on its way.” It is because hitters rarely swing at any 3-0 pitch, so that one in the zone is almost always a called strike and if Joe is figuring the lwts the way I think he is, and the way Walsh does (in this case, the lwts value of a called strike is the difference between a 3-0 and a 3-1 count, or around -.078), a 3-0 pitch in the zone SHOULD be quite negative (since it is almost always a called strike)!