THE BOOK--Playing The Percentages In Baseball

Thanks, Tango.  I’ll put that on the laundry list of ways to chop of the data!  Another suggestion was to group pitchers by strike percentage, which I also plan to do.  (Other ways to group might include swing percentage induced, called strike rate, etc.)

Yes, those are all good too.

Think of the way baseball works.  You start at a 0-0 count, and you have 12 possible counts, at which you can end at any one of them.  The less percentage of contacted pitches, the deeper you go in the count.  If the rate of called balls to non-contacted pitches is high, you’ll end up with more balls per PA across the board.  So, a guy like RJ, who has a low contact rate and a low called balls rate means that he goes deeper in the count, and he goes deeper into the 2-strike counts than the 3-ball counts.  And the other thing is the number of foul strikes.

What I did was create a basic Markov to model this.  It’s rather fascinating, and I should write it up at some point now that I have far more data to test it against.  (I did not do sequencing, which may be an issue.)

In any case, except for guys who are incredibly smart and break the Markov (Greg Maddux notably), it’s a pretty straightforward process to determine the number of pitches for each event for most pitchers.  For a guy like Maddux, he’d have a very specific Markov chain that you’d have to model.  (Obviously, each pitcher would have his own Markov chain, but, most pitchers can be handled with the same transition chain, with just the inputs changed.  Maddux has different frequency/inputs and different transition rates, which makes him unique.)

I’ll be interested to see the results for strike%.  More generally, strike% may prove to be a useful stat for understanding pitchers’ effectiveness.  It’s certainly not as simple as high strike% = good pitcher: Roger Clemens, for example, is just average (63%), and some below-average pitchers have been successful (notably Zito, Zambrano, and Glavine).  However, if you rank pitchers by strike% at Bball-Ref (I looked at starters with 500+ IP), it’s clear that a high strike% is generally associated with success:  the top 20 pitchers have ERA+ of 117, vs. 94 for bottom 20 (unweighted averages).  The top 15 include Schilling, Oswalt, Halladay, Pedro, RJ, Maddux, Mussina and Santana.  And eyeballing the list, it also looks like a high strike% is associated with being able to go deeper into games, and perhaps longer careers.

As a coda:

http://www.hardballtimes.com/main/article/in-search-of-efficient-pitchers/

It’s a small effect and it doesn’t look like it’s controlled by the pitcher anyhow.