2019 Rate of Adjusted Pressure

Guest column by Ajit Kirpekar

Welcome to the 2019 Rate of Adjusted Pressure (RAP) numbers! This is my annual attempt to measure opponent-adjusted pressure rates created in the spirit of Football Outsiders. And given the uncertainty around football these days, what better way to whet our appetites for an upcoming season that's in jeopardy than to look back at last year.

If you've never seen these articles before, RAP is a statistic that measures a team's ability to generate or avoid pressure. Pressure, for those unfamiliar, is a more granular statistic for measuring pass rush.

For those who missed the origins and methodology behind RAP, see this link. The tl;dr version: it starts with a model that predicts pressure, then re-runs the model with a subset of variables set to the yearly average. The subset of variables are all opponent-related, such as the quarterback and the quality of the team. In case you missed last year's article, see this link. Reminder, all data comes from Football Outsiders' game charting data (collected first by FO staff and volunteers, then in conjunction with ESPN Stats & Info, and since 2015 by Sports Info Solutions).

I like to begin these articles with a bit of housekeeping. Every year I tweak the model slightly. This year the big change was how the quarterback position is incorporated into the model. In previous iterations, I used the quarterback's name directly. While there are pros and cons to this approach, inevitably as the sample of quarterbacks grows, it becomes too unwieldy to do it this way. So for this iteration, I replaced the quarterback's name with a suite of metrics representing the quarterback, such as their average historical pressure rate, hit rate, depth of target, etc.

With that out of the way, I want to address a point that was raised in the comments in last year's article along with some related thoughts. There was a discussion about how strong the opponent adjustment should be, and if it's too small, how valuable is RAP? Full disclosure, the strongest current opponent adjustment for the full sample is around +/- 40 pressures, with a typical adjustment between five and 15 pressures (about a game's worth). Now, is that a lot or a little? And if it's too low, would adding more factors increase the degree of adjustment?

The answer to the latter is almost certainly yes, but is it worth adjusting for more things? That question raises multiple thorny issues. First, there's the philosophical question of just what should we be adjusting for? AGL adjustment for every team? What about temperature or field type or weather conditions? Those are worth discussing, but there's also a practical problem. Adjusting for lots of factors has ripple effects. Neural Networks, the model I use, have deep interconnections between the various factors and thus small changes can have very large impacts that are hard to disentangle. In fact, the results can be unintuitive at first for reasons I will get into below. All that to say, I tend to be parsimonious in general, and if that weakens the potential opponent adjustment, then so be it.

Anyways, onto 2019. Let's begin by looking at the very best offenses.

Offenses

It's hard to disentangle quarterback play from supporting casts, but the Saints provide a fascinating natural experiment. In last year's article, I extolled Drew Brees' excellence at avoiding pressure as his teams almost always end up at the very top of these tables year after year. Well, this year was different. Although the Saints once again finished first in RAP (they were slightly worse this year than last). Drew Brees started 11 games with a RAP of 19.4%, while Teddy Bridgewater started five games with a RAP of 22.9%. As mentioned above, Brees was excellent once again but so was Bridgewater, whose pressure rate would have ranked him just below Dallas for sixth-best offense in terms of RAP. That result surprised me somewhat, so I was curious to see if perhaps Bridgewater had been Brees-eque historically. Well, no, as it turns out. His historic pressure rate prior to 2019 was 33.7%, placing him in the bottom 10% of quarterbacks who have thrown at least 200 passes. A shocking result indeed.

I guess the other mild surprise is seeing Cincinnati, Carolina, and Jacksonville high atop this list despite fielding some terrible offenses the whole year. A point I have emphasized in prior articles: just because you are good at avoiding pressure doesn't mean you have a good offense.

Here's the list of the worst offenses in 2019:

I normally cover big trades in the beginning of the article, but this felt like a good spot to discuss last year's deal between Houston and Miami. The headline-stealing trade for the 2019 season was the Texans acquiring Laremy Tunsil and change for two first-round picks plus a second-rounder and change. The net result, well, is probably not a surprise from the Dolphins perspective; Miami finished 26th in 2018 with a RAP of 36.7%, but fell to next-to-last in 2019 even though their RAP improved to 32.5%. For the Texans, we should not deny the benefit. They were an unthinkable horror show in 2018, finishing last in the league with a RAP of 41.6%. They made a dramatic improvement in 2019, finishing 29th with a RAP of 30.9%. They still weren't good, mind you, but nonetheless there was real tangible progress. The question is how much of that goes to Tunsil, as it was unlikely the Texans were going to repeat that league-worst performance in 2019? And is he worth the enormous investment in draft capital and now salary?

As for the Dolphins, the first thing that's important to realize is that 2018 was a bad year for the worst pass-blocking teams. With that out of the way, though their ranking dropped in 2019, the Dolphins managed to improve their RAP in spite of the roster bloodletting. If Tunsil was there, might they have improved their season? Probably, though to what degree depends on which quarterback was starting. Ryan Fitzpatrick, to steal a line from Mike Tanier, is a perfectly cromulent quarterback when it comes to his historic pressure rate, while Rosen is among the very worst.

Now, for the best defenses of 2019.

Defenses

Baltimore had a good year. A very good year indeed. Curiously, quite a few of these teams—Baltimore, Pittsburgh, Los Angeles (Rams), Dallas, and Chicago—were also top-10 performers a year ago. The big surprise here is Jacksonville, which had a good year with opponent adjustments pushing them up to second place. In fact, they had the second-strongest opponent adjustment next to Chicago.

The team I want to focus on, though, is New England, which attracted a lot of discussion as they climbed up the historic DVOA rankings. They had a notoriously easy schedule for a big chunk of the year. Remember the discussion above about opponent adjustments? New England was my test case as I specifically wanted to see why the opponent adjustments were "only" worth -1.5%.

Indeed, the schedule in the first half was comically easy. All three of the Patriots' division opponents were among the worst teams in RAP this year, and they also got to play the hapless Giants. So what gives? I don't have a complete answer, but there is something subtle going on. While football metrics are capturing snapshots in time, we also have to remember that football isn't played like a static sport. Even before coming into a game, teams adapt their game plans week to week in advance of the opponent they are facing. And of course, circumstances force further adaptations.

Imagine you are the Jets. You are on the road in Foxborough with Luke Falk as your starting quarterback. What sort of offense are you likely to run? Then, when you inevitably fall behind a lot, do you start dropping back and heaving long passes? Maybe you should, but it turns out the Jets didn't—at least, not according to the charting data. In fact, it's amazing how much play calling changes as the score widens, both from the favorite and from the underdog. When teams are down big, especially in the second half, they are likely to throw shorter, with more screen passes and dumpoffs, and in fact shy away from accruing massive pressure. Teams are unlikely to gamble on third-and-long. Teams understand what New England is doing and likely adjust their game plans accordingly. That helps to inform somewhat why the opponent ratings don't ding New England harder than expected.

Finally, here's a look at the worst performing defenses:

Did I mention it was a bad year for Miami? No? Well, it was bad. It was really bad. We don't have to talk about that anymore; it was just bad. That's it. Hopefully next year will be better.

In the past, I have mentioned how pass defense is pretty weakly correlated with pressure rate (and RAP). This year was a bit of an exception as most of these teams were bad at pass defense. They were also teams that had losing records, so it's not like they were given many chances to tee off on opposing quarterbacks. That reality is a bit of a blind spot for RAP because while I can adjust for the opponent, the probability of getting pressure when you are down big is unlikely to begin with. Remember, game conditions matter as much if not more than who you are playing.

I wonder how many people would have guessed Seattle or Los Angeles would be this bad given their pedigree. It turns out, both weren't that great in 2018 and regressed badly a year later.

It didn't attract the same attention as Houston's other deal, but the Jadeveon Clowney trade was also an interesting one. The Texans seemed un-bothered trading away a talented player for peanuts, but year to year, Houston regressed just a bit, while Seattle regressed badly despite acquiring Clowney for a song. That raises an interesting question about Seattle's need and Clowney's value.

Full Rankings

Here are the full rankings:

Ajit Kirpekar is a data scientist based out of San Francisco, California. Despite this, he roots for the Indianapolis Colts. You can follow him on Twitter @akirp.