I’d heard about this analysis, but hadn’t taken the time to go through it. But after reading this article at Wired, I’ve downloaded it and decided to go through it and see if I can make heads or tails of it.
From the article, there are a couple of key graphs:
So let’s say you are NFL coach, and you have a fourth and three on your opponent’s 30 yard line. Romer could tell you that 1) you have a 60 percent chance of getting a first down, and that teams with 1st downs inside the thirty yard line score a touchdown 40 percent of the time, for an expected point value of 1.7 and 2) that field goal attempts from the 32 yard line failed almost 65 percent of the time, which meant that going for a field goal only had an expected point value of 1.05. In other words, it’s almost twice as effective to go for it than to attempt a field goal.
That’s the gist of how the study on 4th downs was done. Now, for the payoff:
So what do most coaches do? They consistently make the wrong decision. According to Romer’s analysis, teams would have been better off going for it on fourth down during the 1st quarter on 1100 different drives. Instead, coaches decided to kick the ball 992 times. This meant that NFL coaches made the wrong decision over 90 percent of the time. Romer summarized his counterintuitive results: “This analysis implies that teams should be quite aggressive. A team facing fourth and goal is better off on average trying for a touchdown as long as it is within 5 yards of the endzone. At midfield, being within 5 yards of a first down makes going for it on average desirable. Even on its own 10 yard line – 90 yards from a score – a team within three yards of a first down is better off on average going for it.” Romer conservatively estimates that a more aggressive approach on fourth downs would make a team 5 percent more likely to win the game.
So, NFL coaches have been making bad 4th down decisions for all of time. The article goes on to lament that, since the paper had been published, 4th down decision making has not changed one bit in the NFL. Thus, proving that humans are incapable of taking useful information and applying it.
I’m all for counter-intuitive results, but I’ve got a few questions. For instance, how does Romer account for the fact that, if a coach goes for it on 4th-and-3 from their own 10 and doesn’t make it, the opponent has a virtual certainty of scoring a touchdown? Also, should we really expect such a huge change in decision making for a measly 5% improved chance to win a game? Does he assume that each team has a 50-50 shot at winning the game? If so, that’s a big problem. While I believe in the old “On any given Sudnay” maxim in football, I also believe that good teams are good for a reason and they’re chances of winning are substantially higher than 50-50 week in and week out.
Also, I wonder if he attempts to account for momentum? Not converting on 4th down is a mental hit in football. Does something like that show up in the math?
Football has been played long enough that I tend to think that 4th down decision making has evolved the way it has for a reason. I’m willing to listen to the possibility that it’s wrong. That’s why I’ll read through the paper to try and make sense of it. But I’ll state straight out that I’m skeptical. Math is one thing. The real world is something else.
3 replies on “Going for it on 4th Down”
I think this theory is best left for video games. In my opinion, the major factor that he fails to account for is that a spreadsheet is as close as he has ever come to a football field. Everyone has a plan until they get punched in the face.
Possibly. One major conceit, in my view, that he mentions in the paper is he assumes the teams are evenly matched. You and I both know that’s rarely the case. It also appears that his sampling is from the 1st quarter because late game decisions change the dynamics too much. To make sweeping statements about NFL coaches choosing poorly seems a tough sell to me based on that.
I’m still slogging through the paper.
Well, from a statistical point of view, I can tell you that I took a lot of advanced statistical classes in college, statistically, more than I should have had to have taken. I don’t know why, but someone made me.
Essentially what you learn from every statistical class that you take, no matter the level, you can make statistics say whatever you want them to say. That is why as soon as I see someone prove something through the use of statistics, I figure that there is a legitimate statistical chance that someone has statistically proven the exact opposite case with their own statistics…statistically speaking, of course.