March 10, 2008
Tipsters
Chris Mazzone notes that Mike Pelfrey tipped his pitches last year and writes:
...it always amazes me when stories like this come out and it seems like the team the pitcher plays for is the last to find out...
Recognizing that a pitcher is tipping pitches is a pattern recognition problem. The batters the pitcher faces are (subconsciously) trying to develop a recognizer that will tell them what Pelfrey is throwing. They may not even be aware they are developing such a recognizer, or what clues they are storing and discarding. Suddenly, it dawns on a batter (or a bench coach) that the thing they are picking up is a wide glove, and then everyone knows the sign.
The home team, especially the catcher, has no need for such a pattern recognizer, because they know which pitch is coming. If teams want to detect tips early, they should assign someone behind home plate who doesn't know the signs and pitch selection of the subject to predict each pitch before it's thrown. If the source is getting a high number right, then a team can investigate what might be tipping off the scout.
It's the sort of problem a computer might be very good at solving. Given photographic examples of a pitcher, and the type of pitch that was thrown as training data, differences in the model should indicate where a pitcher is tipping his pitches. This seems like a good problem for my former colleagues to solve.
The Baseball Musings pledge drive continues through March. Please consider making a donation.
Posted by David Pinto at
12:47 PM
|
Pitchers
|
TrackBack (0)