Cited as a fundamental truth of football, the phrase “establish the run” is often uttered as teams hand the ball off and pound away for little to no gain. The story we are told is that such attempts will make running later in the game more successful, accompanied by descriptors such as “body blows”, or some such boxing term meant to have you and I imagining two massive fighters pounding away at one another. That landing these hits against the defensive line will eventually crack them open and allow either average runs to become longer or allow more explosive runs later in the game.
Tom Cable, the sometimes controversial offensive line coach for the Seahawks, has advocated for “establishing the run” to improve the passing game since at least his time as head coach of the Raiders.
These 5,6,5,7 yard runs early in the game turn into 8,10,12,15+ later in the game. Good start for the Chiefs OL.— Geoff Schwartz (@geoffschwartz) December 17, 2017
So, when the Seattle Seahawks dial up a run on 2nd-and-7 in the first quarter the theory goes this will open up the running game later on -- The reality of the situation is that there is little to no evidence that any of this is true in the modern NFL.
This article isn’t an attempt at proving the negative nor an attempt to discredit the value of running in general, instead I will be showing why we should not assume establishing the run works. And to be very clear, we are talking specifically about the idea that runs get longer over the course of the game because of previous attempts, or that a large amount of running attempts opens up the passing game. This article is not addressing the benefits of having a healthy rushing attack in general, nor am I advocating that Seattle never run the ball. So before I go off slandering football canon, let’s talk through a few things.
After all, what does late game success in the rushing attack actually look like? Most often commentators and proponents of establishing the run posit that doing so leads to longer gains. So naturally, the most straightforward metric would be yards. One might argue that we should use a different, more advanced, statistic in order to capture information such as line to gain, or other situational factors. So, in order to address the case more fully I will present you three different measures of success:
- and WPA. (You can read more about EPA and WPA here.)
Personally, I don’t buy the argument that yards don’t capture the effectiveness of an offensive play in general. While it is true that the amount of expected rushing yards will change depending on play call, given a large enough sample of runs we should see the variations caused by particular edge cases smooth out and the overall trends remain in place. Additionally, are the run plays called in later quarters measurably different from those called in the first? That would almost surely be a question of coaching philosophy and may not show in the league-wide data. In any case, what story does measuring success in yards tell us?
If we are in the fourth quarter, can we expect a change in the average value of a run or perhaps longer explosive runs relative to earlier in the game?
From the above plot, we can see there isn’t a noticeable difference between the median run distance in the 2017 season when partitioned by quarter. In addition to the medians being the same, outside of overtime which is much lower sample size, the structure of the data by quarter looks almost identical. What does this tell us? It suggests the distribution of the length of a run in the fourth quarter is indistinguishable from that of a run in the first. This does not support the theory that running the ball is easier or more successful later in the game.
Quarter, however, may be the wrong way to partition our data. The difference may be because of the number of drives in a game, or the number of previous rushing attempts. Keep in mind on both of the next two charts, that the values on the extreme right are much more rare and therefore have lower sample size. Nevertheless, there does not appear to be a systemic change in run length distribution in either case.
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.
We cannot view any substantial change in the structure of our observations and there does not appear to be any strong evidence that rushing attempts gain a team more yards regardless of play type later in the game. But perhaps this isn’t the whole story. What if you don’t accept yards as a valid metric? Thanks to another Field Gulls staffer, Ben Baldwin, we can look at both EPA and WPA.
EPA is a measure of expected points based on a play, so if previous rushing attempts wore down a defense, we would expect to see higher EPA as the number of previous attempts increases. However, when combing through seven years of data —2009 through 2016— Ben finds exactly the opposite.
In general, the number of expected points decreases as the number of rushing attempts increases. Let’s be careful what we pull from this graphic, this chart should not be interpreted to suggest that rushing has no value, especially late in the game. Teams that have run a large number of running plays could be doing so for any number of reasons. One of which is obviously a team with a lead late in a game is more likely to run the ball in order the reduce the amount of time their opponent has to operate if they get the ball back. Scoring in this case almost seems secondary. But if running the ball early sets up a team for running success later in the game, EPA doesn’t show it from an expected points perspective.
WPA is a measure of win probability, here we see the probability that a team will win as a function of how many running or passing plays they have run throughout the course of the game. So when positive, a play with a positive WPA increases your overall probability of winning a game. Interestingly, we see that win probability appears to increase above zero for rushing after about 20 attempts. However, again there are confounding factors here.
It could be true that rushing leads to a higher win probability. However, it could also be true that greater than 20 rushes indicates a more balanced offensive attack. Additionally, a team rushing after a large number of rushing attempts is potentially just trying to run out the clock. After all, WPA accounts for time remaining in the game. Lastly, some teams have a large number of rushing attempts out of nessessity since they lack a franchise quality quarterback. So, at best there is a slight correlation between positive win percentage gains and running the ball late in the game after earlier attempts. But telling which is the chicken and which the egg is not immediately clear.
So the number of rushing attempts does not have a positive effect on the expected number of rushing yards later in game, and the number of expected points does not increase either, there does seem to be a positive relationship between probability to win a game and late rushing attempts. For those who watch any amount of football this seems unsurprising as teams with a lead often focus on the run to run out the clock in the waning moments of a contest.
I won’t be joining the chorus of those calling for teams to never run, but I do think idea of “establishing the run” is one we ought to be skeptical of. While the idea has permeated the commentary for and philosophy of NFL football, there is little to no evidence to suggest such a relationship.
“But what about the Seahawks? Maybe they do better than the league wide trends!”
I’d love to think this is true, let’s first take a look at the 2017 distribution of run lengths in yards through Week 15.
The relationship almost looks like we finally might have a trend here, except that we need to remember that the rushing attempts of one team over a partial season (though Week 15), especially partitioned by number of previous attempts, is a fairly small sample and we expect to see some variation. Especially given the less than stellar success of Seahawks rushers in 2017 not named Russell Wilson or Chris Carson.
But, what if we look at a more “golden age” of Seahawks rushing, such as the 2013 and 2014 seasons? After all, if establishing the run works, surely it must especially have worked when Seattle had Beast Mode in the backfield. A back still known for battering down defenses, breaking tackles, and running would be tacklers over.
At best, our graphs show that any possible gains from running early start to or completely disappear beyond 20 rushing attempts and even then only a slight pattern is evident in the 2013 season, none in 2014. Much more likely is that the very small movements we see in median rushing attempt length is attributed to random variation in the sample.
This isn’t to say running doesn’t have value, running allows you to control the clock, pace of the game, is less variable than pass receptions, and arguably less susceptible to turnovers. Taking what the defense will give, much like the Los Angeles Rams did using the rushing attack versus the Seahawks, seems like the much better strategy than committing to hoping that somehow the running game will improve on it’s own throughout the course of the game. So, unless evidence comes forth where “establishing the run” leads to better performance somewhere, it’s probably time to let the saying die. The most likely scenario, is that simply the game has changed. And while Seattle’s coaching staff has talked about using an established run as a proven championship technique, the truth is that even in the last decade the game of football has shifted.
Visual of AY/A 1970-2016. From 1979-2003, passers were about equally efficient. From 2004 on, it's like a whole new league. It's why I'd caution about comparing rookies in 2017 to the rookie stats of like McNabb or Peyton. Different sport almost. pic.twitter.com/B36mZSRbWL— Justis Mosqueda (@JuMosq) December 23, 2017
Every once in a while it’s important to challenge our beliefs and evaluate the validity of things we hold as true. Perhaps in the 1970s, in an era of smaller rosters and no in-actives for injuries we would see a measurable effect of establishing the run. Or perhaps the theory still holds at the lower levels of football, it’s just that professional players now are better coached and in better shape than at any time before. This article isn’t to say defenses never break, get tired, or just plane old get beat.
But if establishing the run was ever a verifiable way of improving the offensive performance in the National Football League late in the game, it appears there is little evidence to suggest that it remains so as evidenced yet again by the recent game against the Cowboys. A game in which even early success against the Seahawks failed to materialize substantial benefits to the Dallas offense when they needed them the most.
Ezekiel Elliott rushed for 51 yards in the 1st quarter, 22 in the 2nd, 15 in the 3rd, 9 in the 4th.— Danny O'Neil (@dannyoneil) December 25, 2017
It’s time to accept that maybe this conventional wisdom is either a matter of faith or that establishing the run is an artifact from a bygone era of football no longer valid in the current day. In either case, it would behoove Seattle to look past this method. While improving the run game will certainly help the team be less one dimensional on offense, as evidenced by six of the top 10 DVOA rushing teams making the playoffs this season, it’s clear early run calls simply for their own sake do not improve overall offensive performance late.
Data for this article was obtained by use of the nflscrapeR package. Credit to Ron Yorko.