In yesterday's NFC playoff football game, the New Orleans Saints were leading the Carolina Panthers 31-26 with two minutes remaining to play. The Saints faced 4th-and-two at midfield, and the Panthers had used all their timeouts. In this situation, I think most coaches would choose to punt: the idea would be that the other team, in this case the Panthers, would then get the ball with less than two minutes to play, no timeouts to stop the clock, and perhaps 90 yards to cover for a game-winning touchdown--a poor prospect. Considering it from the other side, if you try for the first down--only two yards to go--and succeed, you've won the game, whereas if you fail you've shortened the field for the other team by around 40 yards.
The Saints spurned the punt and went for it on fourth down. They called a passing play and their quarterback, under pressure from the Panthers' pass rushers, threw toward a covered receiver about 15 or 20 yards down the field. The pass was intercepted and then fumbled out of bounds. It was a poor decision to intercept the pass, since if it had just been knocked to the ground the Panthers would have taken possession at around the fifty instead of at their own 30-something yard line. But it's hard in the moment not to do something that you practice doing and is almost always the best play.
The question about spurning the punt remains. In yesterday's game, the Saints held on for the win when the Panthers, after moving the ball to about the Saints' 20-yard-line, threw a couple of incompletions in the end zone before running out of time and downs. In the view of the television commentator Troy Aikman, the Saints' defense got their head coach "off the hook" for having tried for the first down on 4th-and-two. I suppose the opinion people have about this depends mainly upon their intuition about what is likely to happen--and, of course, since something does actually happen the door is always open for Monday morning quarterbacking from fans who did not have to make a decision in real time.
What interests me is the possibility that one need not rely exclusively on "gut instinct" to decide on the best strategy in a situation like the one the Saints faced yesterday with two minutes left in the game. When following the progress of a game on a website such as espn.com, there is a win percentage figure that constantly calculates the probability that one team or the other will eventually win the game. As the game goes back and forth, the figure fluctuates with the ebb and flow of events. I assume the figure is based on computer modelling of what actually happens in NFL games, and this in turn suggests that there is a "right" answer to the question about whether it was wise for the Saints to go for it on 4th-and-two yesterday. To illustrate, we can just assign probabilities to certain events. For the Saints, the salient ones would have been the probability of making a first down on 4th-and-two, the probability of a touchdown being scored in less than two minutes when a team starts at their own 10-yard-line and can't call a timeout, and the probability of a touchdown being scored in all the same circumstances only with the offense starting at midfield. Data from all NFL games played allow one to approximate these probabilities. If we set them at 70%, 10%, and 30%, respectively, then determining the best strategy is just a math problem. Go for it, because your chances of winning are slightly better! If you punt, you have a 90% chance of winning (I'm discounting the possibility of unlikely events such as the punting team winning after yielding a touchdown). But if you go for it, you have a 91% chance of winning. You can satisfy yourself this is true by considering what you'd expect to happen in, say, 100 games. Seventy times you'd win by picking up the first down on 4th-and-two. In the other 30 games, you'd win 21 times by stopping the other team from scoring a touchdown after they take over on downs at midfield. I cooked the numbers to make things turn out nice.
I also put the word "right"--"right" decision--in quotes above because of course it's impossible to set these probabilities with precision. The computer model could tell you what happens, on average, in this or that situation in all NFL games, but decisions are made in particular games with particular players who have their own particular strengths and weaknesses and degrees of tiredness and health, etc., etc. If I had to guess, I'd say that yesterday the Saints' coach reasoned something like:
If I go for it, the outcome may be determined by my offense playing against their defense. But if I punt, the outcome is going to be determined by their offense playing against my defense. I feel better with the former.
The use of "feel better," if apt, indicates that these are indeed decisions based on gut instinct rather than data, which is defensible if, as in my example, data modelling renders a close verdict. That is, it's six of this or half a dozen of that, so I'm going with what my gut tells me is best for my team in this game. My last point about this, however, is that from everything I can tell the data does not yield a close verdict. If in an NFL game a team is in a quandary over whether to punt or go for it on fourth down, the default position should be to go for it. Putting it another way, NFL teams punt way too much: they always punt when the data says they should, and they almost always punt when the data indicates that going for it is the optimal strategy. One of the early papers on this subject was written by David Romer, an economist at the University of California, Berkeley. I'm sure he's a football fan, but, as the abstract to his quite math-y paper from 2005 makes clear, he's really interested in the widely accepted view that people who run firms know how to run the business efficiently so as to maximize results:
This paper examines a single, narrow decision—the choice on fourth down in the National Football League between kicking and trying for a first down—as a case study of the standard view that competition in the goods, capital, and labor markets leads firms to make maximizing choices. Play-by-play data and dynamic programming are used to estimate the average payoffs to kicking and trying for a first down under different circumstances. Examination of teams’ actual decisions shows systematic, clear-cut, and overwhelmingly statistically significant departures from the decisions that would maximize teams’ chances of winning. Possible reasons for the departures are considered.
An entertaining summary for those put off by mathematical symbols and the like is here. Evidence struggles to overturn the dead weight of conventional wisdom. A leading reason may be that it takes guts to resist it. If the Saints had lost the game after punting on 4th-and-two, the coach probably would have been asked in the post-game presser whether he wished he had instead tried to make a first down. But if they had lost the game after going for it on fourth down, he would have been asked about nothing else, and people would be calling for his scalp.
Comments