THE BOOK--Playing The Percentages In Baseball

Is the Leverage Index higher during a basestealing event because runners tend to steal more when it’s an important situation… or is it because when there’s a runner on first, the LI is on average 20% higher than the exact same situation without a runner on first?

Oooh, great point.  I should have thought about that.  A perfect opportunity to remind people of the archives, and specifically this article:

http://www.tangotiger.net/archives/stud0069.shtml

In there, I show the LI, by the 24 base/out states.  With a man on 1B, with 0,1,2 outs, the LIs are: 1.38, 1.11, 0.77.  The simple average is 1.09.  I forgot what the distribution of SB events were by out (it’s in The Book somewhere), and it was either 26/33/40, or 40/33/27.  In the former, the weighted LI would be 1.04, and in the latter, the LI would be 1.13.

As well, also in The Book, I should the number of times that the SB occurs based on the score of the game.  However, I didn’t show the number of times that the score of the game happened (so that we could figure out the percentage of times attempted… I don’t think I did anyway).

Anyway, my guess is that of the 1.20 LI, around 1.10 of that is because the guy is on base, and another 1.10 of that is because of the tightness of the game.

I don’ think you can only include first base. There are steals of third as well as steals of second with a runner on third. That would drive the LI up.

Steals of 3B would be somewhat similar to steals of 2B (see link).

Steals of 2B, with a runner already on 3B has an LI of close to 1.6.  Given the rarity of this, this may add .01 to .03 to the total LI.

Studes’ analysis is interesting, as always, but I’d argue it understates the value of SBAs (or overstates the damage, I suppose). 

First, it doesn’t look like it captures advancement on errors on a failed pickoff attempt, which presumably happen in high LI situations. 

More importantly—and this applies to nearly every analysis of SBs and CSs—it is counting some unknown # of CSs that really are the result of blown hit and runs, not true SBAs.  If we knew which CSs (and SBAs) took place when the batter swung at the pitch, and subtracted these out, I suspect the net value of SBAs would be substantially higher, probably close to or better than break even.  I say that because you’d have to think the success rate on these “SBAs” is pretty poor, often involving slow runners who wouldn’t attempt a SB absent the H&R being on.  (Alternatively, you could try to add the positive value of H&Rs where the batter makes contact, in terms of extra bases gained and DPs prevented.  But I’d think that would be even harder to figure out.)

You know, this shouldn’t be too difficult to figure out. If you looked at all instances of a SBA with a swing-and-miss using Retrosheet data, you’d be capturing every hit-and-run, and presumably not too many straight out steals.