Lure Of PitchFX

We came home this evening and caught the final bit of the Giants game on the radio. They were playing the Phillies, and Brad Penny, who was pitching for the Giants for the very first time, was throwing a gem, shutting the Phillies out through 8 innings, the Giants going on to win 4-0. I was following the early part of the game using MLB Gameday, which has very nice visualizations of the trajectory of each pitch thrown. (“Whoa. Penny’s pitching well”, I thought).

A company called Sportsvision has been capturing this pitch-by-pitch data for a little over two years now; they have a set of high-speed cameras installed at every major league ballpark to track the flight of each pitch and compute the trajectory parameters. This system, which they call PitchFX, generates the data for the Gameday visualizations and for similar graphics shown on TV.

It turns out this PitchFX data is freely available from the MLB Gameday site (in tedious XML format), and at something like 500,000 pitches thrown each season in the major leagues, that is a lot of data. By now there are a number of good websites featuring this data — for instance, here is a set of graphs showing Brad Penny’s performance this evening.

If they gave PhDs for sabermetrics there’s room for hundreds of dissertations using the PitchFX data; it’s fun to think of what analyses one could do with the data. Once they get their image processing algorithms worked out, the folks at Sportsvision are aiming for a complete digital record of every game event, including where the ball gets hit and where the fielders move to on each play. Then we’d have thousands of baseball dissertation possibilities.