FREE hit counter and Internet traffic statistics from

Thursday, May 24, 2007

Deep Data Dive

As promised yesterday my column on Baseball Prospectus this morning dives deeper into the PITCHf/x data tracked by the 2007 version of Gameday (a new update was released on May 10th and is much more performant).

In this article I take a look at the velocity and location data that includes over 40,000 pitches and discover that given a one-inch margin of error the system agrees with umpires to the tune of 90%. Not bad and very similar to the QuesTec results published by Robert Adair in an article titled "Cameras and Computers, or Umpires?" that was published in Volume 32 of SABR's The Baseball Research Journal.


David Gassko said...

GREAT article, Dan.

anthony said...

Outstanding article, especially where you analyzed the pitch speeds to show the effect of the heavy air at parks like Petco.

Guy said...

Excellent piece Dan. I look forward to more of your analysis of this data.

A thought on the issue of declining velocity by inning: You have two kinds of selection bias in the later innings. One, as you note, is that pitchers who are pitching well that day will pitch longer. Two is that good pitchers in general, who tend to be faster, will pitch longer. Given the dropoff in pitches thrown starting in inning 3 (and then accelerating), I suspect the actual drop in velocity by inning 6 is more than 1 mph, and further deterioration might be quite rapid from that point on for many pitchers (were they allowed to continue). So I'm not sure we can conclude yet that "pitchers are probably not taken out of games primarily because of their loss in velocity."

You should be able to correct for the second kind of selection bias by measuring each pitcher against his own 1st inning velocity. It would also be interesting to see a pitcher's average speed for his final 10 or 12 pitches before being pulled, as a percentage of his own 1st-inning velocity, and broken down by # of IP he lasted. That should tell us if pitchers are starting to lose velocity when manager's pull them, and if so, at what inning this starts to happen.