March 9, 2010

Monday I introduced the idea of a Probabilistic Model of Steals. In that post, I looked at how the stage of the game, represented by the inning, changed the probability of attempting a steal, and the success rate on those attempts. Subsequent posts examined outs and the score difference. This posts looks at what happens when all three parameters are combined.

To review, I’m only concerned with the pure steal situation, only a runner on first. I define score difference from the point of view of the offensive player (Offensive score – defensive score). In looking at the data, a lead of seven runs in either direction seemed to be the point where teams really stopped running, so any lead of seven runs or greater is either placed in the 7 or -7 bin. I also put all extra innings in the 10 bin for that category, since each extra inning starts the same, with the score tied. The following table shows the probability of attempting a steal based all three parameters. Only combinations with at least 1500 steal situations are displayed:

I’m not surprised that the mostly likely situation for attempting a steal is with the lead-off man on and the score tied in the first inning. That’s pretty much what lead-off men are designed to do, get on base and steal if they have the opportunity. Most of the other high steal attempt situations are late in the game with two outs and the score fairly close, but not trailing. In fact, in the top 34 rows, the only trailing situation in is in the third inning, with two out trailing by one. My guess is these are situations in which the lead-off man reached the second time around, and it’s worth him getting in scoring position for the big guns. The low attempt situations mostly come late in the game with teams trailing by at least two runs.

What impresses me here is that managers really do know how to use the stolen base. They tend to run when one run is important, or early on when their best thief reaches base. Looking at the list, there are very few situations in which I’d ask, “Why are you running often there?” The consensus manager, if you will, has a pretty good grasp of when to run.

The next installment examines which players run often, and which players cling to first.

As always, I’m interested in your feedback. The next installment will combine inning, outs and score difference into a complete model. You can follow the series here.

Please consider supporting this work with a donation to the Baseball Musings Pledge Drive.

The information used here was obtained free of charge from and is copyrighted by Retrosheet. Interested parties may contact Retrosheet at 20 Sunset Rd., Newark, DE 19711.

Posted by David Pinto at 8:31 pm | Base Running, Defense, Probabilistic Model of Steals | Permalink | No Comments