I'm getting pretty data-driven with my softball team. Every week they have to attempt and steal five bases, and their times are being captured at 120 FPS by a video camera. I go back and get their times down to 1/100 sec accuracy, and log them.
Assuming their times are normalized (which they tend to be), I work out z-scores on a spreadsheet and look at their probability of success in stealing a base against a certain time (in softball, players can't leave the bag until the pitcher releases the ball, so I generally start it on pitcher release). If they have more than 75% probability of beating that P-2B time, they have a green light (there is an extra 0.2+ seconds for a tag to be applied if the throw isn't on the money).
If they have more than a 50% success probability, they can expect I might send them in a certain situation when called upon. Anybody else would only be in those other situations like hit-and-run, full counts (even less than two outs if a walk still moves them over), etc. The players kind of get into it, too. They keep trying to improve their average time and are working better off pitcher release because of it.
I do the same on the other side with catchers, and I can get their exchange times as well when on camera. It kind of makes for a fun endeavor, but it takes a hell of a long time to do.
Comment