This has already been a great offseason for us at PFF, from my colleague Timo Riske revisiting a seminal piece by Cade Massey and Richard Thaler regarding the value of draft picks to Kevin Cole’s look at how each combine drill performs when projecting college players into the NFL. The PFF Forecast with myself and George Chahrouri has had plenty to talk about, with quarterbacks hitting the free-agent market to Tua Tagovailoa’s stock as the draft’s second-best quarterback. We’ve even published a few excerpts from our college-to-pro projection model over that past month.
All of this is coming off a 2019 offseason where we laid bare some of our most controversial ideas, chief among them that coverage is more important than pass rush. While this finding was a bit upsetting for some, it elicited some great discussion (here’s our chat with former NFL safety and current PFF analyst Solomon Wilcots). Additional studies looked at how quality coverage forces quarterbacks to hold the ball longer, while quality pass rush makes quarterbacks throw it more quickly, and how the Ravens broke the NFL in that regard in 2019.
Today, I want to talk about the problem of constructing a defense a bit further. Generally speaking, I’ve thought about coverage and pass rush as collective units — aggregates of constituent pieces. This is helpful in trying to draw broad conclusions, but for this time of year, when teams are building their rosters full of individual pieces, it’s not the most appropriate way to look at it. The Chiefs are explicitly deciding whether to trade Chris Jones or to sign him to an extension — they are only implicitly deciding to change the composition of their pass-rushing and run-defense groups. When Denver acquired A.J. Bouye this week, they didn’t acquire an entirely new secondary, and when his former team likely trades Yannick Ngakoue in the coming weeks, they will not be transferring their entire pass rush with him.
Thus, a problem to address is how “fragile” a defense is to performance by each player within the unit. Since pass defense is really all that matters (kidding, kind of), we will examine how each player, from the highest-graded to the nth-highest-graded player in coverage and in pass rush, affects opponent success in a given game. The two competing hypotheses are that:
- Defense is “star dependent” — outcomes are mostly a function of how the best players perform
- Defense is “depth dependent” — outcomes are more a function of how lower-tier players perform
The first hypothesis is a view that defense is more “antifragile,” that small perturbations to the system have little effect on outcomes. Unless such perturbations affect the star player or players, a defense can still thrive. This is my working hypothesis for how offense works — most non-quarterback perturbations (injuries, ineffective play) will have second-order (at best) effects on outcomes.
The second hypothesis is more aligned with the view that defensive football is fragile — small changes (even on the lower rungs of the defense) affect the wellbeing of the entire defense because the offense has the power to identify and exploit the corresponding weaknesses.
For this study, we will use two sources of data — expected points/success rate data at the play level and PFF game-level grades. The former track how successful a team is offensively and defensively at the play, game or season level, while the latter is our proxy for how good a player is in coverage or pass rush during the course of a play, game or season.