Thread by @DerekCarty

/THREAD Why "TE vs. Arizona Cardinals" is PROBABLY NOT A REAL THING

I looked at all defenses vs. TE from 2000-2018, comparing the 1st half of the year to the 2nd half. If a D is actually good/bad vs. TE (instead of getting lucky/unlucky), they should continue being good or bad all year, right? But that's not really what we see...

*Yards per Target*

When targeted, TE yardage has a VERY loose relationship from one half to the next. It's statistically significant, it matters, but it's mostly noise and heavy regression should be expected.

"What about the absolute worst teams in the 1H?", you say? "I only care about extremes like the Cardinals!" Well, over 19 years, only 2 of each year's worst teams vs. TE failed to improve. On average, they improved their yards-per-target to TE by a massive 3 yards in 2H.

*TDs per Game*

Another very loose relationship overall, and the absolute worst 1H teams allowed .4 fewer TD per game to TE in the 2H

*Target Rate*

Still not huge, but teams that are attacked w/ TE in the 1H do tend to continue being attacked in 2H to a degree. Perhaps there's something the other numbers don't capture. Or maybe it's just the lingering perception that leads teams to continue targeting them.

ARI has allowed 29% of their targets to TE -- the 3rd-highest rate in the league, 10% above the league average of 19% -- which is the biggest thing we can point to for why this could continue being a thing to some degree. But it's also not a guarantee because...

Strangely, the teams that allowed the absolute most targets to TEs in the 1H regressed almost all the way to the mean in 2H, which is interesting given the link we found overall. Shows that even for something decision-driven like targets, there is plenty of noise.

We also need to account for context. Teams w/ bad numbers against TEs tend to face an uneven share of good TEs. Arizona has faced the 6th most talented mix of TEs in the league. And DJ Swearinger, the guy getting beat by TEs for big plays in the early weeks, has been released.

Maybe ARI is truly awful vs TE -- they're an outlier, there's some scouting/scheme thing that impacts our prior, etc. But based on data alone, we have to conclude they're non-awful. If the sum total of your "TE vs. ARI" take is "look at the flow chart", that's not good enough.

(Legitimate question: Why does everyone call this stuff "flow chart"? *These* are flow charts. I've never seen football data presented this way. I've seen tables, although I suppose "color-coded tabular representation of data" doesn't quite roll off the tongue as well...)

People hate when you bring up "variance" and "sample size", especially before regression hits. "It's been working, how can it be wrong?!" But it doesn't matter what has already happened. It matters what *will* happen, and context/regression improve our ability to guess that.

Bottom line: Over two decades of NFL, teams as bad as ARI vs. TE have gotten a lot better going forward that same season.

Defense matters, but it's very noisy. If it's your main reason for liking/disliking a player, this data suggests you're doing it VERY wrong.

People love defense vs. position stuff and will probably be upset reading this. If you disagree, I'd love to know why and to hopefully learn something. But if the stance is "Just look at the numbers!" that's not really good enough. That data is not really reliable enough.

Evan Engram will probably be chalk this week. He'll be a good play (especially if Shep and/or Saquon are out), but if I play him, it'll be because he's a talented, high-volume player... not because I'm jamming whoever is facing the Cardinals.