During Sunday’s NCAA contest between Notre Dame and Texas, the Longhorns faced a second-and-goal from the Irish one yard line midway through the second quarter.
“I wouldn’t be surprised to see a run,” said ABC announcer Todd Blackledge, opining on what type of play the Longhorns should try. “However, I will say that second down is the down to throw if you want to throw.”
Blackledge’s comment is classic football-think, behavior that’s been suggested for decades with no known empirical basis. However, thanks to the supple data of Armchair Analysis, it’s the type of behavior that easy to check and quite possibly validate (at least using NFL data). Thus, the two questions I’ll attempt to answer are:
How do coaches call plays in goal to go situations? Second, how should they call plays in goal to go situations?
Armchair’s database contains each NFL play since 2000, which I filtered into goal-to-go plays that occurred on first through third downs. I also cut out fourth quarter plays, as to worry less about the effects of varying late-game behavior.
How do coaches call plays? Here’s a barchart showing the percentage of plays which are passes, separated by down and distance (Note: called passes include sacks).
With one yard to go, roughly one in four plays are passes, which is roughly the same on first, second, and third downs. Across other distances, coaches are fairly consistent in their desire to run on first down and throw on third down, with second down decisions roughly a 50-50 split.
Interestingly, there is no noticeable spike in teams calling pass plays on second down, at least not relative to their behavior at other downs and distances. So, while football coaches love to talk about throwing the ball on second and goal, they aren’t necessarily acting that way.
Perhaps the more interesting question is how should coaches call their plays in goal to go situations. The answer is not straightforward. For example, passing plays are more likely to yield touchdowns on longer plays, but they’re also more likely to yield negative plays and plays of no gain.
One possibility is to consider the drive’s eventual point total as the outcome and to work from there. Using the same set of plays, I used the drive’s result (categorized as a touchdown, field goal, or neither) to estimate the average point total given each play call at the various down and distances.
Here’s a chart of expected points, separated by runs and passes.
Each point in the chart reflects the expected point total given a run or a pass at a certain down (1st, 2nd, or 3rd) and distance (x-axis). The size of the circle is proportional to the number of plays called at each point.
Interestingly, across most downs and distances, running plays offer slightly more of a return than passing plays. The differences aren’t overwhelming, but they are consistent, generally in the neighborhood of one or two tenths of an expected point. Notably, there’s no evidence to back up any claim that teams should pass the ball on second down – if anything, it’s the opposite.
These results are somewhat surprising given Kovash and Levitt’s seminal paper on NFL team behavior, which implies that teams should pass more than they currently do. One possibility is that the shorter field lessens the abilities of teams to throw the ball. Anecdotally, and as an example, teams seem to call way too many fade patterns.
All together, what did we learn?
First, there is no obvious truth to the theory that teams are passing more on second down and goal to go situations than we would expect. Second, there’s no evidence to the theory that they should be passing more often than they already are.
If anything, there’s a drop in efficiency on passing plays in goal-to-go situations, which may be showing up in the form of fewer expected points. However, I’m cautious of reading too much into this conclusion, given the nature of this analysis (it’s on aggregate, and doesn’t account for game and play specific factors) and the inherent difficulty in categorizing a team’s decision to run or a pass.