From the middle of the second millennium to the pontificate of John Paul II, the Roman Catholic Church had an office of the promotor fidei (“promoter of the faith”). While this prelate’s role varied slightly from century to century, his primary function remained consistent: to ensure no one was elevated to the sainthood injudiciously. This meant seeking out and presenting the best possible arguments against each Servant of God, no matter how revered he or she may have been. Thus the office earned the nickname advocatus diaboli – the devil’s advocate.
On Friday, my colleague Neil Paine laid out a compelling argument for why the Seattle Seahawks’ Richard Sherman is likely a better cornerback than the Arizona Cardinals’ Patrick Peterson, who has lately been claiming otherwise.
Paine primarily uses four pieces of evidence:
- Pro Football Focus’s play-by-play ratings favor Sherman.
- Passes thrown toward Sherman have had much worse outcomes (by a variety of metrics, including how often they’ve been intercepted) than passes thrown toward Peterson.
- Teams throw fewer passes toward Sherman.
- Seattle’s pass defense is better than Arizona’s.
This is about as open and shut as statistical arguments in the NFL get (though I might add that Sherman has had four forced fumbles to Peterson’s zero). And let’s not forget that Sherman also made probably the greatest individual play by a cornerback in recent memory, accompanied by the greatest post-game interview.
But before we canonize the man, let me play advocatus diaboli.
Peterson claims his job is harder than Sherman’s. Paine puts this aside to focus on tangible statistical outcomes, which is probably the right approach to establish who is objectively more likely to be the better cornerback (especially with the balance of evidence so decidedly in Sherman’s favor).
But while we may not be able to test a counterfactual — How would Sherman do on Arizona’s defense, and vice versa? — we can examine the stats for hints that Peterson’s theory is at least plausible.
I found one particularly interesting stat: Peterson made 151 of his career 161 tackles (94 percent) by himself, compared to 138 of 167 (83 percent) for Sherman. This suggests the possibility that Peterson carries a heavier individual burden than Sherman, which could be one reason why Peterson’s stats don’t seem as impressive.
Pretty thin gruel, I know, but that’s the point: Statistical arguments in football often rest on a razor’s edge. That’s partly because nearly every statistical sample in the NFL is tiny (these two players have played only 48 regular-season games each). Beyond that, there are three major reasons to be skeptical of statistical comparisons, even open-and-shut ones, between players:
- Opponents adapt their strategies to the quality and types of players they’re facing. For example, they may throw toward a tough cornerback less often, or only in certain situations.
- Teams sometimes ask for more from certain players. Better cornerbacks may be given added responsibilities so the rest of the defense can focus on other things. To give an absurd scenario, imagine if one corner could cover two receivers adequately but not spectacularly. His stats might not look impressive, yet you’d be able to send a free blitzer on every down. How valuable would that be?
- One player’s stats are highly dependent on others. All the others. A cornerback’s stats depend on support from the rest of the secondary, but also on how well the defense pressures the quarterback (into making bad throws, for example). Shutting down the running game forces opponents to pass in less-efficient situations, such as third and long. A good offense also helps a defense. It creates more late leads, and thus more desperate and obvious plays by the opposing offense as it takes (necessary) risks to try to get back into the game.
To sum up: Statistics are extremely unreliable indicators of player value, especially over short periods of time and in limited situations. How many flash-in-the-pan defensive standouts have we seen? How many free-agent signings go sour because the player is nowhere near as effective in a different system? (Other than quarterbacks and Deion Sanders, I’d like to say most of them.)
In this case, both Sherman and Peterson have only three years of play for us to evaluate, and each has played in only one situation on one team. Thus, a single confounding possibility — that Peterson has considerably more difficult responsibilities than Sherman — if established, could make the entire multifaceted case for Sherman collapse.
With that, the devil rests.