I must share a major disclaimer before jumping into to this: Pass Coverage performance as a whole, regardless of what metric you use, is by far the most volatile of datapoints. Hence, ANY model built to predict outcomes based on coverage ability (using the past to predict the future) must be taken with a hefty appetite for variance. This is said in the most macro of senses. That is, compared to other advanced statistics coverage data tend to be more elastic. As Pro Football Focus said in a relevant study in 2019:
“Coverage players who grade well one year are more likely to grade well the next year than those who do not, but the uncertainty is substantial. Second, and likely more importantly, your defensive success is largely a function of the offenses (and, more specifically, the quarterbacks) your team faces – something it has little control over even when sound personnel and scheme decisions are made.”
Essentially, how well a player is at covering a receiver is NOT as predictive game-over-game or even season-over-season, as other player performance metrics like pass pressure rate, broken/missed tackle rate, etc. Not to mention, figuring out WHO will cover a player, and how often is riddled with subjectivity. With that said, despite the fact it seems I’m telling you to stop reading, there is still some insight to glean from a properly calibrated model, WHEN YOU CONSIDER IT AT THE MARGIN (i.e., think when you choose a starter and have a 50/50 decision between WRs). Each week, FantasyPros plans to unveil our WR vs CB piece, utilizing our friends at PFF’s database, while applying our own model to determine marginal advantages for WRs in a given week for fantasy football, leveraging what we know to be predictive triggers between players.