Aaron Donald has been the most dominant defensive player over the last five years of football.
We’ll start with that because it is a consensus. Had Houston EDGE J.J. Watt avoided health problems, it could have been contentious—but he didn’t, so it’s been Donald. Since 2014, the year he entered the league, he is second in sacks despite playing DT, not EDGE; first in TFLs by a wide margin; first in QB hits by a comfortable margin as well. The last time he wasn’t the best interior rusher on PFF’s Pass Rush Productivity metric, which accounts both for pressures and sacks generated, was in 2015. He was second.
Conveniently, most of the stats in this argument are based on the pass-rush—only TFLs give a head-nod to running plays. That’s for a couple of reasons. Firstly, there are no established stats specific to the running game by which we can evaluate defensive players—TFLs are really the best individual stat thereof, and they only measure production behind the line of scrimmage. Secondly, it is intuitively and analytically more important that a defensive lineman is good against the pass than against the run. Passing plays are more dangerous than running plays, so from a general view, it is more valuable for a defender to be good against the pass than against the run.
This week, ESPN analytics shared their next step in the interpretation of player tracking data they receive from Next Gen Stats: Run Block Win Rate (RBWR) for offensive linemen, Run Stop Win Rate (RSWR) for defensive linemen. These metrics were used to address the first issue—the lack of a definitive stat for defensive line performance against the run—and they showed that Aaron Donald was an average run defender. The entire argument for Donald’s numbers is laid out in a Twitter thread here by ESPN analyst Brian Burke, for those who cannot access the full article on ESPN.
https://twitter.com/bburkeESPN/status/1303405922803503105
For not just tape-watchers, but just average NFL fans, this is heresy. That Donald is elite is an a priori assumption; it is as unimpeachable as it is fundamental to understanding football at any level. As such, it is difficult to understand the ESPN argument. If a metric thinks Donald is average, there must be an issue with the metric.
And there is an issue with the metric—multiple, even. From a pure play-level perspective, Burke admits “...that an expert watching the all-22 video would probably do a better job grading run blocking and stopping performance than these metrics.” Teaching a machine to understand a football play is ludicrously difficult, especially if you consider how hard it is for a person, intuitive and clever and receiving far more data inputs, to understand the same play. We should expect the machine to categorize plays as wins that we would categorize as losses, and vice versa, as nuance is lost in the process of categorizing and filtering plays and situations.
That’s OK, as it’s true for a lot of metrics. Next Gen Stats defines a “tight window throw” as a passing attempt that has a defender within one yard at the time of completion; I define a “tight window throw” in Contextualized Quarterbacking as a throw which has an ideal placement changed relative to coverage. One, you can teach a computer; the other, you can’t. One might be more stable year over year than the other. This is a blossoming field, with discoveries yet unmade.
Scrutinizing the results of RSWR with a suspicious eye, then, we can take a look at the same sensations that Burke and Seth Walder did when they went to verify the approach of RSWR relative to Donald’s poor numbers. They found that:
- Donald’s tackle rate (9% of running plays) was 130th out of 158 qualifying defensive linemen
- Donald was double-teamed at an “above average but not exceptionally high rate”
- Donald was not avoided in the running game via run direction
- Donald had the greatest vertical push of all defensive tackles (1.3 yards, on average, in 2 seconds after the snap)
There are some connections we can draw from this information. Donald’s RSWR was 29%—just below average—but his tackle rate was wildly below average. As Burke points out in his thread, this actually points to Donald being a better run defender than his tackle rate would suggest—i.e., the point of a metric like RSWR. But we can and should add color to this conclusion by highlighting where Donald’s tackles live: behind the line of scrimmage.
Donald’s 25 TFLs against the run last year once again paced interior defenders—but his total tackles against the run were only 16th (Grady Jarrett and Davon Godchaux, mentioned as top performers in tackle rate, were atop this list). As such, what stands out about Donald in the running game is not how much production he had—but rather, where his production occurred. In that RSWR puts no weight on TFLs relative to tackles beyond the line of scrimmage, nor weight on wins that create TFLs relative to wins that create tackles beyond the line, this effect of Donald is not captured in the metric.
Among players with at least 100 snaps against the run, Donald was third in TFLs/snap behind only New England’s Adam Butler and Buffalo’s Jordan Phillps. He was also third in the rate at which his tackles occurred behind the line of scrimmage, behind Phillips and Philadelphia’s Hassan Ridgeway.
Because of Donald’s low overall tackle rate, we know that this sensation is not about Donald being a sound run defender who is wildly productive behind the line of scrimmage. That argument can be better made for a player like Grady Jarrett, who has an elite tackle rate, but also the second-highest TFL number (19) among interior defensive linemen. Rather, it is about the fact that when Donald makes a tackle against the run, it’s behind the line of scrimmage, and thereby an extremely high-impact play—otherwise, he doesn’t seem to make many tackles. Last year among DTs, Donald was only 27th in EPA against running plays, despite his league-leading TFL numbers.
A compilation of those TFLs from 2019 can be found here. If this is where Donald makes his positive plays against the run, we want to understand how he makes them.
https://youtu.be/7XWfZA1bXnI
On Donald’s stops at or behind the line of scrimmage, we see a common theme. Yes, there are hustle plays and clean-ups, as well as traffic jams generated by a staunch defense on the point of attack. But generally, Donald produces TFLs because of his explosiveness, quickness, and hand usage—the same traits that make him so dangerous as a pass-rusher. This is also reflected in Donald’s vertical push shared by ESPN, which had no defensive tackle getting further upfield in such a short period of time as Donald after the snap.
That vertical push metric, however, is not just for plays on which Donald has tackles or TFLs. It’s for all running plays. As such, on plays on which Donald does not make the TFL or even the tackle—of which there are many, as we know—he is still getting far upfield. Those plays look like this.
https://youtu.be/jL5wOq8OOTw
This is really the crux of the argument, the seesaw on which we teeter. Are these plays good or bad, and how good or bad are they?
There is an indeterminable calculus here, rooted in a few hazy truths and otherwise obscured. One truth is that disruption is production—a common refrain when pressures were incorporated into the accepted lexicon by which to describe a pass-rusher’s ability. Here, the argument is that Donald’s premier ability to get upfield regularly discombobulates blocking schemes and forces running backs to adjust their paths, which is exactly the role that a “squeeze” or “force” player should fill in the running game. It is not his job to make the tackle—rather, to make the read, decision, and angles easier for a second-level defender.
Another truth is that you cannot pick up one end of the stick without picking up the other. The film provides compelling evidence that Donald’s low tackle rate and high TFL rate are both products of his play style, which is born of his physical toolkit: he gets upfield aggressively and looks to disrupt in the backfield, not only because that’s what he does historically well, but because he does not have the physical skill set to do the alternative thing (sit on a block, gap control, maintain the line of scrimmage) historically well.
This choice means a lot of things: Donald is less likely to be able to affect more than one gap (that’s how Butler is so productive). Bigger gaps are more likely to open up behind Donald given how hard he sells out upfield (that’s what we see in the above cut-up). Donald is more likely to affect play-action quickly because of how aggressively he comes upfield against the run (and remember, play-action passes are the best thing to ever happen to offenses).
This is where the limitation of something like RSWR is revealed. RSWR does not assume that each player has an equal responsibility to tackle the ball-carrier—that’s why Donald grades better than expected, given his tackle rate. But it does assume that each player has the same intention to play the run, and that simply is not the case. Donald is penalized within the metric for sacrificing sound run play for the sake of aggressive penetration and potential pass-rush opportunities. The question that looms over us, then, is this: is the juice worth the squeeze?
This is where our calculations become difficult to define or describe. To what degree does Donald need to play so aggressively downhill against the run in order to generate his quick and record-setting sacks? Could a better nose tackle or stack linebacking corps behind Donald mitigate the damaging effects of his play style on the run defense as a whole? If Donald dedicated himself to stopping the run instead of getting to the passer, would he be a better player overall? To that last one, emphatically, I say no.
What Burke and Walder attempted to do was define the success of the individual player within the gestalt of the run fit—a worthy pursuit. Their metric undeniably undervalues Donald’s disruption as a run defender—but that’s OK because even in that undervaluing, they force an important next step: to define the value of an individually successful run fit relative to the alternatives. For many players, it would seem that having consistently suboptimal run fits foretells bad decision-making or risk management—but this is Aaron Donald. His supremacy is assumed, and by his ends, his means are justified.
At the end of the day, this is what it boils down to: star players, and the cautious handling of them. Troy Polamalu once Superman-ed the goal line snap count against the Titans. Ed Reed once speed-turned Peyton Manning in the middle of the field. If Aaron Donald thinks that Aaron Donald needs to play a certain way to continue producing 11-plus sacks/year on the interior, then it is probably best to let him play that way. If a metric says that he’s an average run defender, guess what? He’s still the exact same Aaron Donald he was two days ago—the best defensive player of the last five years, and a self-proclaimed outlier by the very film analysts now stunned to discover that his mind-boggling play has come at a small, generally insignificant, otherwise unimportant cost after all these years.
Filed In
Related Articles
NFL Draft
Arik Gilbert Doesn’t Need Big Workload To Be A Top NFL Draft Pick
- Aug 22, 2022
NFL Draft
2023 NFL Mock Draft: Marino 1.0
- Aug 22, 2022
Written By