clock menu more-arrow no yes mobile

Filed under:

Quantifying Football: Davante Adams is a case study in analytics

Adams’ value has been difficult to nail down, but rethinking the denominator to a commonly used stat may make all the difference in the world.

Divisional Round - Seattle Seahawks v Green Bay Packers Photo by Quinn Harris/Getty Images

Football stats are getting more sophisticated. Long gone are the days of using simple volume stats to try and gain insight into player talent or player value. At least that’s the case if you are serious about your endeavor. We are firmly in the early stages of the analytics revolution in football. It took baseball by storm in the 2000s, basketball has rapidly accelerated its acceptance in the past fifteen years, and even soccer, which had been resistant to analytics, has seen a large increase in usage over the past five years.

Football is no different. And the ending will be much the same even if the sports are very different. Very smart people will figure out ways to approximately quantify what is happening on the field, and those who are early adopters will gain significant value from doing so. Then, when they are wildly successful, the entire sport will follow. We see that in baseball, where there are virtually no dumb teams anymore. That is the end-game of the analytics revolution. We are currently in the late 1990s on the baseball timeline, with very few full adopters and plenty of ground still to cover.

In this stage, there is a quite a bit of upheaval. Not just of the old ways of thinking, but even in how we construct the very metrics we use in the “new school” of thinking. Statistics will come and go, and our understanding of what is happening on the field will improve. That doesn’t make the old stats bad. The misconception about analytics is that it comes to the problem already knowing all of the right answers. “Analytics” isn’t really a thing so much as it is a process. The goal is to incrementally get closer to the “right” answer.

Baseball, a sport that is much easier to appropriately quantify given its disconnected nature of players and very single-event structure, still isn’t perfectly understood, even after two decades of intense scrutiny at the highest levels. But in 2020, it is certainly far better understood and teams make FAR fewer mistakes than in 2000. That is the goal of analytics. That’s why changes will occur within football analytics for years to come, and why we should be open to analytics and also, not overly confident in our conclusions at the time.

EPA vs. Incompletion

That long preface is all to bring me to a new statistic that was brought to me from the hellscape that is Twitter dot com. Anthony Reinhard (someone whom I strongly recommend following at @reinhurdler if you do not already follow him) posted a thread with a different way of looking at receiver value:

I strongly recommend going through the thread as it does a good job of explaining why Anthony made the changes he did to a commonly used stat of EPA-per-target. It helps to further flesh this out and also suggests reasons why some statistics that we traditionally use to value wide receivers have been possibly underrating Davante Adams.

Adams’ ill-fitting role

My main complaint with Davante Adams hasn’t even been Davante Adams the player, but rather the way that Green Bay has chosen to use Davante Adams in some instances. One of my least favorite plays that Green Bay runs for Adams (well, it’s really more of a check than a play), is throwing him a smoke screen immediately on the snap. If you’re unfamiliar with what a smoke screen is, or the offensive terminology used that you’re familiar with is different, it is when the quarterback snaps the ball and throws it to the wide receiver immediately, isolating him on the opposing cornerback. It essentially turns your wide receiver into a punt returner.

Our own Paul Noonan talked about why Rodgers and Adams are a poor fit because of Rodgers’ propensity to hold onto the ball too long, but getting the ball to Adams too quickly — like instantly — is also a drag on his efficiency. I have to credit APC’s own Matub for sarcastically saying in the APC writers’ Slack something along the lines of “any time you can turn the league’s best route runner into a punt returner, you just have to do it.” I don’t think my point needs to go any further than that.

With those targets included, Adams’ efficiency is dragged down. Schemed throws can be successful if they utilize what the player does well, but given that Adams best trait is his world-class route running ability, not letting him run a route is kind of defeating the purpose of having Davante Adams on the field.


As Anthony goes on to say in his thread, there are many problems with EPA-per-target as a stat. It punishes receivers for many factors that are out of their control. The schemed play is only one example. Another is that they are damaged when a ball is thrown on third or fourth down and they do not reach the first down line. Is it the receiver’s fault that his route happened to be short of the first down marker?

What EPOI (expected-points-over-incompletion) seeks to do is to reduce the impact of situational variance on the player’s numbers. It isn’t a perfect metric. With the publicly available data we currently have, there are no perfect metrics. The truly amazing data lies behind paywalls with price tags in the thousands and tens of thousands. So we are left to work with what we have.

When creating a statistic, or making a modification to an already existing one, the leaderboard should make sense. It also should have an interesting entry or two, otherwise it isn’t adding anything new to the discussion. The reason I chose to highlight EPOI here is because it does both. The leaders here do make sense. It’s a who’s-who of elite receiving targets. It also gives us a few guys who we may be underrating. This is where analytics can thrive. It’s not about us completely re-inventing the wheel, but about always getting a step closer to the truth.

The debate of Davante Adams’ true value has been something that has confounded me for awhile now, and I’m not strongly committed to one thought or the other. Seeing Adams rank so strongly here does make more sense than how he has ranked in something like DVOA for years (as a fringe WR1). What I hope we can see from Green Bay in 2020, if we get a football season in 2020, is that the Packers will be able to more closely align his DVOA with his EPOI. If GB can cut the fat off of his targets by getting him the ball in that sweet spot range between two and three seconds after the snap, eliminating the Rodgers hero-ball and those godforsaken smoke screens, the Packers offense could make a good jump and Adams could be in line for a monster season.