The Epistemology of Baseball (or, Joe Torre and an actuary walk into a bar...)

... where, of course, they watch a baseball game. Let's say it's the bottom of the 9th inning, the game is tied with no one out, and there's a runner on first base. Should the home team bunt?

The bartender asks Joe Torre (the famous baseball manager), who replies: "You should always bunt in this situation. Everyone knows that!" And he's right - if, by "everyone," you mean people commonly recognized as baseball "experts": players, coaches, and sports journalists.

The actuary would likely give a different answer, something like "Meh," or "Bunting would be silly." That's because, historically, teams that occupied the position the bunt is designed to achieve (bottom of the 9th, tie game, runner on 2nd, one out) have won a modestly lower percentage of their games than those that occupied the position the home team is now in. In fact, historical data shows that many commonly-employed baseball tactics often decrease the probability that the team employing the tactic will win. Yet these tactics persist. (You can play along at home here.)

None of this is news to readers of Moneyball, which described the inefficiencies that exist (now to a lesser degree) in the market for baseball players. There is a vast amount of evidence, across a wide range of predictive domains, that formal, data-driven methods of prediction out-perform the kinds of impressionistic judgments that so-called experts often make. Yet despite the strength of this evidence, experts in the relevant fields often ignore or outright reject the evidence. And how would you expect Joe Torre to feel when told that, despite his vast experience, he can't perform as well as an actuary who has never even attended a game?

What does this have to do with law? At least a little, as I'll suggest in a later post.   

Comments

[driade]

So, I'm a bit confused. The stats you cite measure the success of teams who have runners on first or second (and one out) regardless of how those runners got to second? Don't we want to know how often the runner on second scored when they were advanced by a bunt? Might that be a different percentage, after all, than the percentage of runners who scored after a one-out double? Also, isn't it possible that Torre's teams are more successful with the bunt in this circumstance than other teams? In other words, could the conventional strategy be much more effective when correctly employed and disastrous in the hands of the inexpert? Bottom line, the stats may not be helpful if we don't know exactly what they measure, and when do we ever know exactly what they measure?

[Mark Weidemaier]

Driade – It’s certainly true that some teams and players are better at bunting. It’s worth noting, though, that the comparison between pre- and post-bunt states assumes a successful bunt. Thus, even a bunt with a 100% chance of success puts the bunting team in a position from which – judged by historic win-rates – it is less likely to win. So it’s probably more accurate to say that the conventional strategy is "less ineffective" when deployed by an expert.

And I don’t think we do need to know how the runner got to second base. Bunting involves a choice between State A (bottom 9th, tie game, runner on first, no outs) and State B (bottom 9th, tie game, runner on second, 1 out). Since baseball started collecting play-by-play data, thousands of teams have occupied those two states, and we know exactly how frequently they won: 71.6% from State A and 70.4% from State B. If you’re suggesting that a team’s chances of success depend on how it reaches State B, I suppose that’s possible. But extremely unlikely, I would think. (And easily testable by someone with access to the underlying data.)

You’re right, of course, that we lose information by aggregating data. For that reason, a manager who made decisions based only on the data would make some errors. To take an extreme example, win-expectancy tables make it look really stupid to intentionally walk the first hitter in the bottom of the 9th inning. But if that hitter is Pujols, and Weidemaier is on deck, it’s a no-brainer. A good manager would identify these situations where the aggregate data points in the wrong direction. Even then, however, the manager would ideally decide based on some more relevant sub-set of the data, not because of intuition or hunch, or because “It was good enough for Casey Stengel, and it’s good enough for me!” The broader point is that managers also make errors by ignoring or choosing to override the data, and they tend to make a lot more of them.

[generic viagra]

Yet despite the strength of this evidence, experts in the relevant fields often ignore or outright reject the evidence

[WELLSPENNY34]

All people deserve very good life time and mortgage loans or car loan will make it much better. Just because freedom is based on money state.