Re: What did this article conclude?

Practically, what the article tells someone building a team is that

1) A good pass rusher last year will likely be a good pass rusher this year. So, if I add a historically good pass rusher to my team the chances that he is a bust is less than a coverage player.

2) A good coverage player this year is harder to identify than a good pass rusher this year. Coverage players have more variance in their play than pass rushers.

3) A good amount of the variance in the coverage player's performance is due to the offense he is facing.

One point that I think should be discussed is the difference in grading pass rushers versus coverage players. A pass rusher receives a positve grade for a pressure--beating one guy. The coverage player has to force an incompletion or not get beat in coverage. In my opinion, pressures are much more binary than forcing incompletions. Did the defensive lineman quickly get past the blocker assigned to him? While forcing an incompletion is much more gray--how often have you seen a coverage player in good coverage get beat by backshoulder throw? or see a tipped ball caught? etc. Coverage is subject to a higher variance of outcomes given the same variables in my opinion.

4) Passing offenses can adapt to great pass rush but not great coverage. This is key. Good passing offenses will release the ball before the great pass rush affects the play. Good passing offenses will target the weak link in coverage

5) Bad coverage is more costly than bad pass rush. Pass rush is easier to manufacture through scheme(multiple players working together) than coverage.

If I were a GM I would add pass rushers through free agency. And use my draft capital on coverage players--and I would take alot of them. If I pay a guy a top free agent deal I want to minimize my risk. I do that by selecting the pass rusher in free agency. With my cheap rookie deal money I want to take shots at the coverage players.

That's what I got from the article.

If you are not familiar with EPA the below is a good explanation:

[www.advancedfootballanalytics.com]