Friday, March 23, 2007

Double steals: are managers too conservative?

Here's an excellent study by Dan Fox on double steals, and comments by Tangotiger.

Dan found that for a double steal, the theoretical break-even percentages (1970-2006) were 60% with nobody out, 53 percent with one out, and 78 percent with two outs. Anything less than that, and it's costing the team runs.

With nobody out, teams were successful only 57% of the time, which makes the no-out double steal look like an overused strategy. However, the breakeven percentages assume that it will be the lead runner who's thrown out. But, in real life, it's sometimes the trailing runner. If you take that into account, the 60% theoretical breakeven point is too high. It's probably even less than the observed 57%.

Where it gets more interesting is in the one-out case. Teams were successful there 1587 times out of 2258, or 70%. This is substantially higher than the 53% breakeven point. Dan suggests that managers underestimate the value of getting a runner to third with one out, and that's why they don’t call for the double-steal enough.

(The two-out case has insufficient data – see the

I should mention that these results aren't necessarily inconsistent with perfectly rational managers. Suppose that sometimes, perhaps with really fast runners and a pitcher with a slow delivery, the success rate is 90%. Other times, the chance is only 30%. If managers are smart enough to go for it only when the chances are greater than the 53% breakeven, the overall success rate will be somewhere between 53% and 90% -- perfectly consistent with the observed 70% rate.

However, even if that's the case, we can say that managers are less inclined to call for a double steal than to call for just a regular stolen base. According to "
The Book," the overall (single) steal success rate is 69%, which is very close to the breakeven. The 0-out double steal rate is also close to its breakeven, but the one-out double-steal rate is higher than the breakeven. So, overall, the double steal is used more conservatively than the single steal.

The obvious question is: why are managers more conservative with double steals than with single steals? One answer suggests itself from the
NFL study finding that teams are kicking field goals on fourth-and-short when they should be going for it: that managers are afraid that the risky play will fail, and their reputation will suffer.

If coaches are generally afraid of being criticized, they will call for unusual plays less often than they should. A double-steal is seen as an out-of-the-ordinary play, for which the manager gets a large part of the credit or blame. A regular steal is routine, so credit usually goes to the runner or the catcher, and managers can call for steals without fear of criticism.

If this theory is true, we should find that the kinds of plays for which managers are second-guessed are consistently underused, especially as compared to plays that are similar but uncontroversial. This double-steal data is mildly supportive of that theory.

Anyone have any data on the hit-and-run, or the suicide squeeze?

Labels: ,

At Saturday, March 24, 2007 10:38:00 AM,  Unknown said...

re: measuring the hit-and-run, retrosheet has occasional data for runners being on the move, but it looks very incomplete ...

"If managers are smart enough to go for it only when the chances are greater than the 53% breakeven, the overall success rate will be somewhere between 53% and 90% -- perfectly consistent with the observed 70% rate."

A great point, and of course true for any decision. A manager who is only successful at near the breakeven rate overall, while supposedly not "hurting the team", is probably doing a poor job of evaluating the chances for success, or he would end up with an above break-even success rate.

At Tuesday, March 27, 2007 12:23:00 AM,  Greg Spira said...

The same thing is true for sacrifice hits. Managers are more reluctant to sacrifice with men on first and second than with just a man on first, despite the former being a better play.

Why? I suspect with only one baserunner, managers look to score one run and thus use one-run strategies. With two baserunners, however, they become more interested in a big inning and use one-run strategies less often.

At Tuesday, March 27, 2007 1:32:00 AM,  Phil Birnbaum said...

Hmmm, interesting. Maybe managers are also concerned only with having a runner on second base? I think Dan's article said that they underestimate the value of a runner on third with one out; that would be further evidence that way.