March 4, 2013
Of Dogs, Men, and Stolen Bases
This is one of the nerdiest pictures of myself that I have. It's me in 2006 in Moscow, Russia, at Universitet Lomonosov (aka Moscow State University) standing next to a statue of Ivan Pavlov, the man who discovered the stimulus-responses conditioning reaction in his famous experiments with dogs. Every time I look at this picture, I still salivate. How d'you like that?
Pavlov's big idea, which seems obvious now (all the big ones do), was that dogs (and eventually, people) can form associations between a stimulus and a response that have nothing to do with one another. At the time, it was assumed that animals and people reacted to their immediate surroundings. So, when Pavlov, whose real aim was to collect saliva samples from dogs for work on understanding the digestive system (work for which he won a Nobel Prize!), saw that the dogs were salivating "too early", he wondered why. What he figured out was that the dogs were learning that hearing the footsteps of his assistant coming down the hall meant mealtime. And so the dogs started to salivate. Pavlov eventually replicated the experiment with his famous bell (and other noises).
Thus was born the field of classical conditioning. The theory was eventually extended. Reward a behavior, and dogs (and humans) will do it more often. Punish the behavior, and dogs (and humans) will do it less. The more times you do it, the more ingrained the pattern becomes.
Managers have a funny job description, when you think about it. They don't hit or pitch or field, but they do get to make all of the strategic decisions, including whether the runner on first should stay put or try to steal. Often the runner gets the blame or the credit for being safe or out, but everyone knows who gives the sign.
It must feel nice when the steal succeeds. I'm generalizing from when I hit the "steal" button in a sim game and it works, but there's every reason to believe that it feels good in real life too. And it must be painful to realize that your decision just cost your team a baserunner and an out. So, I got to wondering whether managers were a little less likely to hit the "steal" button if earlier in the game, they had been punished for this behavior by suffering a caught stealing.
Warning! Gory Mathematical Details Ahead!
I isolated all instances from 2003-2012 in which there was a runner on first and no runner on second. I did this at the PA level. A batter who leads off with a walk and then stands there while his three teammates make outs had three chances to steal, and his manager went 0-for-3 on sending him. For each manager, I looked at how often he hit the "steal" button overall. This will serve as a control within each manager and more easily allow us to determine whether a caught stealing messes with his tendencies.
I also looked at the speed score (my own home brew) for the runner on first, the inning (entered categorically), the number of outs (entered categorically), and the difference in the score at the time. Finally, I looked to see whether there had been a caught stealing earlier in the game.
I threw all of these runner-on-first-not-on-second situations into a big logistic regression and found that... managers were actually more likely to attempt a steal if a runner had been caught earlier.
More. Apparently, they didn't learn their lesson. The dogs would not be proud.
Grunt with me
Maybe because managers aren't dogs.
If there's something that I've come to understand about baseball, it's that it's impossible to understand the game without understanding the psychology of the human male. Yes, over time, the rational mind will see that if a strategy isn't working, you should stop using it. However, the human male is not known for feats of rationality, particularly in a case where he has challenged another man and come up short. Men, in general, don't take that sort of thing well, and are more likely to try again. (I was debating whether to make a Tim Allen "More power!" joke, but then I realized that half of the audience here might not have been around when that show aired).
A few years ago, I ran a study looking at how steal-happy individual managers were. I found that there were significant differences between managers, and that once you controlled for extraneous variables, they were pretty stable from year to year in their tendencies. I wondered what the “reaction to a caught stealing” findings would look like if I ran them for individual managers. I selected for all managers active in 2012 and re-ran the same basic logistic regression for each manager's data (going back to 2003 if available), and only for him. I then looked at how powerful the "previous caught stealing" variable was in predicting the chances of a steal attempt being made. In theory, this should tell us how much of an effect a caught stealing has on that particular manager.
For all managers, the effect was positive and significant, but there were variations in how significant. Reporting on the effects of a logistic regression is tough, but here's what I did. I set all of the other variables in the equation equal to a 5% chance that the manager would steal. I then looked at what the simple fact of a previous caught stealing would have.
The managers most likely to try to chase down that first caught stealing?
(Ozzie Guillen was no. 6)
The managers least likely to let that previous caught stealing bother them?
Anyone up for some qualitative analysis of those two lists?
It looks like the first list is filled with managers known for being rather fiery gents, while the second has several men who are known to be more introspective types. And while it might seem that this is just a neat little parlor trick with no practical use, I'd argue quite the opposite. Whether or not it's a good idea to be more aggressive with stolen base attempts is an empirical question. But by treating each manager as his own data set, we've empirically shown how a manager tends to react. And maybe this sort of measurement correlates with other things that a manager does. Do high-reaction managers get more out of players? Do more cerebral managers? Or maybe we just have a nice little way to profile a manager's personality. Either way, we might just have a new tool to run some rather fascinating analyses.