Why do NFL teams run so often? (no, seriously: why?)

Thanks to Yurko, Ventura, and Horowitz (2018), we now have reliable estimates for the expected points added (EPA) and win probability added (WPA) of every NFL play from 2009 through 2017. A consistent pattern that has emerged from this data is that on average, rushing lowers a team’s chance of scoring (as measured by EPA) and winning (as measured by WPA). And yet, on 1st and 10, NFL teams ran the ball 53 percent of the time in 2017.

In order for rushing so frequently to be rational, rushes have to carry some tangible future benefit to offset their cost. What are these possible benefits, and do they exist? I’m going to run through explanations that I’ve heard and whether they hold up. As more research comes to light, I’ll keep this page updated.

Hypotheses that are testable

I’m going to start with testable hypotheses. These make a prediction that can be tested in the play-by-play data that I have access to. For example, “rushing sets up play action” is a testable hypothesis as it posits a relationship between two variables that can be measured. An example of a non-testable hypothesis is “teams rush frequently because rushes have fewer turnovers”. While it is true that rushes are less likely to result in turnovers, we can’t know if that’s the reason teams rush. And if it is the reason, it’s not rational because even accounting for turnovers, passes are far more efficient (again, EPA shows this).

I. Rushing benefits the passing game [UNSUPPORTED]

The most comprehensive study I’ve seen of this is from Nathan Ernst on Hawkblogger. He found that 2-4 percent of a quarterback’s QBR in a given game in 2016 can be explained by measures of rushing frequency or effectiveness from the first 3 quarters of that game (with the first 3 quarters used to eliminate game script effects).

Along the same lines, Scott Barrett of Pro Football Focus took a look at 1st half rush attempts versus 3rd quarter adjusted net yards per attempt (ANY/A), and found that only 1 percent of ANY/A can be explained by 1st half rush attempts.

Another common argument is that rushing slows down the pass rush by making the defense expect the run. But even if this were the case, it does not appear to impact passing efficiency, so it is probably not a meaningful effect.

II. Rushing wears down the defense and leads to bigger gains later in the game [UNSUPPORTED]

Example of the argument: “These 5,6,5,7 yard runs early in the game turn into 8,10,12,15+ later in the game.” -Geoff Schwartz, former NFL player

Sean Clement of Field Gulls took at deep dive into this and came away with the following conclusion:

Using yards as our metric, there is not sufficient evidence to suggest that the amount of previous rushing attempts by a team in the game, the quarter, nor number of drives changes the underlying distribution of the length of a run.

“But teams run to setup the pass!” Is usually the next defense offered up for run establishment. This insidious philosophy seems to have no end of goal posts to move. However, as we can see below there seems to be little evidence that earlier running attempts increase the yardage obtained on passing plays.

III. A rushing yard is more valuable than a passing yard [UNSUPPORTED]

Example of the argument: from Tom Osborne, “We believed that a rushing yard was more valuable than a passing yard when it came to winning football games.” (taken from this excellent Danny Kelly piece; follow the link for more)

Danny’s piece noted that Pete Carroll defines explosive plays using a cutoff of 12 yards for rushes and 16 yards for passes, with the clear implication being that rushing yards are more important than passing yards. But when using play length to predict scoring on a drive, rushing yards and passing yards are about the same (in addition, this should be obvious given how EPA is calculated: a yard is a yard).

IV. Rushing carries value by helping set up play-action passing [UNSUPPORTED]

Example of the argument: “Teams have to commit to running the ball first to open up [play action]” -Geoff Schwartz (again)

In a recent post, I looked for evidence that the effectiveness of play action passing is related to rushing volume or effectiveness. My conclusion:

Putting this all together, I cannot find any support for the success of play-action passing being related in any way to a team’s rushing statistics, whether measured by frequency or effectiveness […] After measuring this every way I could think if, it appears that the conventional wisdom that running is necessary for play-action passes to be effective should be questioned.

V. Teams don’t pass at higher volume because it would become inefficient [UNSUPPORTED]

In a piece on Football Perspective last offseason, I looked at season-long quarterback volume and efficiency. Here is what I found:

We can see that it is certainly not the case that the simple act of passing at low volume is sufficient to carry a QB to strong efficiency stats […] To sum up, looking at the last 15 seasons in the NFL reveals no statistical relationship between a QB’s volume and efficiency.

VI. Teams maximize success rate rather than the likelihood of scoring [UNTESTED]

For example, on 2nd and short, teams should probably be throwing down the field but in many cases they do not. Thinking of how to test this. This article from Brian Burke will be a good starting place.

VII. Running the clock gives a team’s defense time to rest and increases defensive efficiency [UNTESTED]

This should be easy to test: how does the points per drive allowed by a defense depend on the amount of game time that has elapsed since they were last on the field?

Hypotheses that are not testable

Here are the hypotheses that I don’t think we can test. This doesn’t mean that they aren’t true, but it means that I can’t think of a way for these hypotheses to be informed by the data I have access to.

I. Lower chance of turnover

As mentioned above, we know passing plays have a higher turnover rate. We know that, despite that, passing is better on average (measured by EPA or WPA or whatever). The goal of the game isn’t to minimize turnovers, it’s to win (by scoring more points than your opponent). So where is the testable hypothesis?

II. Injury risk to QB

Quarterbacks are less likely to get injured on rushing plays than passing plays, but how do we know if that’s the reason teams run so frequently?

III. Coaches do not call games in a way that maximizes the likelihood of winning

A potential explanation for why teams run so frequently might be a simple one: they don’t know any better. This is certainly possible — perhaps even probable — but the goal of this project is to rule out as many alternate explanations as I can before setting on this explanation. Based on recent research, we know that teams cost themselves wins by being too conservative on 4th downs. If they are also costing themselves wins by rushing too often, it wouldn’t be surprising.