As long as we’re still talking about the Colts vs Pats game
Matt Yglesias, as he tends to do, lays down some common sense-style support for Bill Belichick. Summation: If you can avoid it, don’t give the Colt’s the ball because they have a good offense.
OK, sure. There’s no way I would fight that logic as a piece of generally good advice. However, on Sunday night the Colts were still suffering some pretty important offensive losses, including a faulty Joseph Addai. And…as good as that logic is generally, it certainly wasn’t the case on Sunday that the Colts were scoring at will. The fact is that with 2 minutes to go, the Pats were leading….
And if you don’t want to give the Colts the ball deep in their own territory, you certainly don’t want to run the risk of giving the Colts the ball deep in your own.
Which is not to say that I think that Belichick was wrong, necessarily. It’s hard to fight Mankiw on this:
Some strategies that fail ex post might be optimal ex ante. Randomness is a fact of life, even if Patriots’ fans do not fully appreciate it.
But let me give it a shot.
What happened to the Patriots is not necessarily randomness. The kind of econometrics used here can only be guidance and not because of “randomness.” The statistics only measure what has tended to be the case over time without any causal explanation. The fact is, that going for it on 4th is what should likely be the default position, especially with short yardage…and doubly especially in short yardage in enemy territory.
However, in each of the measured instances of when going for it worked or didn’t work there were observable, on-the-ground, causal factors to explain the result. These are not captured in the general statistics. Modeling can help us peer through the trees to determine the contours of the forest, and that can be helpful. But it isn’t everything. Even if we control for things like particular franchises against other particular franchises (Pats vs Colts) there are still players and coaches and injuries and fatigue that are explanatory variables on the field at the time that Belichick made his decision. Following the guidelines of the probability analysis with disregard to the instantiated team he was actually playing is irrational and, as it turned out, it was the wrong choice regardless of whether it was rational.
Give the idiosyncratic and heavily injured Colts the ball at the 20 yard line with 2 minutes to go. That’s rational regardless of the statistics.
Comments
2 Responses to “As long as we’re still talking about the Colts vs Pats game”
Agreed, Big Dog – those variables on the ground are important. I always grit my teeth going into football math articles, because they tend to draw surprising conclusions (which I like) out of non-specific, possibly not applicable data (which I loathe). To carry on with my endless poker metaphors, a lot of card players say poker is an art as much as a science (or as much as it’s math). But really, the “art” in a game like poker is just math that takes a lot of situational variables into account, which a more general purpose model cannot. No poker book can adequately explain what to do when your opponent overbets all-in on the turn, unless it were to devote 500 pages to the myriad possible situations in which that action might arise. You could come up with a statistics-based percentage that tells you always calling in that spot is profitable with a certain strength of hand – and in fact, it would work out that way, assuming you play enough poker. But a better poker player than you has a rough grasp of that statistic, PLUS a bunch of variable situations that inform his decision for greater profitability.
So going for it on 4th down and 2 or less yards might generally be the correct move, but might specifically not be. Or maybe it’s a wash – in which case maybe you go with the move that gets you criticized less the next day. That’s not a Belichickian philosophy, but it’s not a bad basis for a decision when the outcome might be basically the same. (I’d love to write a long and boring post getting into all those numbers in really dirty detail, because I think every attempt I’ve read so far has been general to a fault – but fortunately for y’all, I have a job and like playing video games and watching TV.)
One last thing though. You said, “Following the guidelines of the probability analysis with disregard to the instantiated team he was actually playing is irrational and, as it turned out, it was the wrong choice regardless of whether it was rational.” Maybe you misspoke, or I’m misinterpreting, but I don’t like this kind of conclusion. By saying “as it turned out” you imply that the correctness of the decision can be determined by the outcome. That’s not true when chance plays such a large role in the result. Going back to poker again, if I know that calling in a certain spot wins the hand 80% of the time but loses 20%, calling is ALWAYS correct even on the 20% of the time that I lose – absent information that alters my understanding of the situation, that is. If Belichick could prove (with better math than I’ve seen thus far) that the 4th down call was at least even odds with punting to win the game, then his decision is correct, even though the try failed and the Colts won.
Also, there is the best dissection of those statistics in the Yglesias comments (http://yglesias.thinkprogress.org/archives/2009/11/to-punt-or-not-to-punt.php). Some commenters are dummies but some are very astute. I highly recommend all the comments by “Mo,” as they get into the kind of situational thinking that needs to be applied for this to be done properly.