First I looked at the average number of plays faced per drive. Ostensibly, a good defense should face a few number of plays on average per drive because it forces the offense off the field quickly. This is not always the case, defenses that were gashed with an abnormal number of explosive plays would also yield lower plays per drive. My hypothesis was that these big plays were infrequent enough that this phenomenon would not affect plays per drive too much. When you consider the relationship between plays allowed per drive and EPA per play allowed on defense, this hypothesis is more or less correct.
For EPA per play allowed, negative figures are good for the defense because EPA is from the point of view of the offense. In general, lower plays per drive correlates positively with lower EPA allowed per play, to the tune of 12.2 percent. This might not seem like a strong relationship on its face, and frankly it is not, but in the world of football analysis, where there is always a ton of variance, this is somewhat meaningful. But how consistent is plays per drive allowed year after year?
This chart demonstrates how much plays per drive should change in year N after year N-1. Unsurprisingly, teams that yield an especially low number of plays per drive tend to move back towards the "pack" the following year (top left of the viz). Same principle applies to defenses that could not get off the field (lower right). Teams towards the middle do not see as extreme yearly perturbations. What does this mean for 2020?
A theme of this post is that New England is going to have to withstand a ton of defensive regression in 2020. They yielded by far the lowest plays per drive. Interestingly, two awful defenses rank highly here (or very low, depending on how you think about it) in Jacksonville (RIP sacksonville) and Carolina. I thought this might be a function of the aforementioned issue with explosive plays. I considered any play gaining at least 20 yards an explosive play. The Bengals were the worst at yielding these plays. They gave up 92 total, which made up 8.87 percent of plays run against them. The Jaguars were third worst (75 total plays, 7.29 percent of all plays) and the Panthers sixth worst (76, 7.05 percent). The Raiders look relatively inept at being able to get off the field here. Add in the fact that they were second worst at preventing these splash plays (7.55 percent of all plays) and you can clearly see why they were awful on defense last year. Arizona is similar to the Raiders; they could not get off the field and were fifth worst in preventing explosive plays. The Vikings yielded the highest plays per drive on average, yet were still a very effective defense last year. They were very good on not giving up the home run, thus proving to be the epitome of a bend-don't-break unit. Given the talent drain at defensive back this offseason, I would be worried (if I was Minnesota) that they can continue to withstand so many plays on defense. I think they are one of the teams most ripe for a setback on the defensive side of the ball this season.
Next, touchdowns per drive. If you remember my last post, like any other statistic, touchdowns per drive regress heavily on the extreme end year over year. Oddly enough, touchdown rate is more sticky on defense (0.124 R-squared versus 0.076 on offense). You know the drill; context then 2019 data:
The best teams should not be a surprise if you followed the NFL last year. The degree to which New England is better than everyone else is impressive, despite their schedule last year. The worst teams are also not surprising. The Dolphins, Giants, and Raiders stunk. One team worth pointing out: the Bengals are middling here yet were terrible overall as a unit. I talked about their propensity to give up explosive plays last season (not sticky year over year), so if that normalizes a bit in 2020 this can be a decent unit. That, in conjunction with a shot to the arm on offense in the form of Joe Burrow, will make the Bengals an interesting wild card sleeper in 2020 (remember this season there are three wildcards in each conference).
To get an overall view of how effective a defense was in a given year, the best way is to look at EPA per play allowed. For the reasons I outline here (I also recommend reading the link from ESPN that goes more in-depth on the background of EPA), EPA is the best measure we have in evaluating teams on the play-by-play level. It strips out contextual factors that a offensive or defensive unit has not control over and does the best job of distilling which teams are the best and worst at running or preventing plays that contribute to offensive efficiency. Like at measures in football, however, EPA per play is still prone to large swings year over year because we only get 16 games worth of data from each team each season.
A negative change is good for the defense (and a positive is bad). A few notes: The vaunted 2018 Bears defense (with the addition of Khalil Mack) regressed heavily in 2019 (CHI-2019). Their EPA per play increased by almost a tenth of a point per play (for context the average passing play adds 0.07 points per play). Another similarly vaunted defense was the 2017 Jaguars. The 2018 Jaguars unit had by far the most regression after that run to the AFC title game. At least the Eagles spun some of their defensive luck into a Super Bowl after the 2017 season. The main reason for these two setbacks I will address in a minute, and how that applies to 2020.
Check out New England, San Francisco, Pittsburgh, and Tampa Bay at the bottom of the chart. All four of these units took a massive step forward from 2018 and should be expected to take a corresponding step back. The Bucs (with the addition of Brady) and the Steelers (with the return of Roethlisberger and more importantly the absence of Duck Hodges or Mason Rudolph) are going to be aided by improved offenses. New England added Cam Newton, so it is difficult to handicap how the offense will look. The team I would be concerned about (relatively) is San Francisco. I still think they will be good, but the combination of defensive regression and multiple players on offense are coming off career years (most notably Kittle) makes me think that crowning them the NFC West champs before the season starts is not warranted.
Here is an overview of the EPA allowed per play by each team last season:
Remember, negative is good and positive is bad. Miami made a ton of investments into their defensive backfield in the offseason. Even without those investments I would expect some improvement but if Byron Jones plays well, Xavien Howard has a bounce back year, Noah Igbinoghene is effective in the slot as a rookie, and they get some semblance of a pass rush, I think they can be mediocre. Washington is another team that has invested heavily in its defense (mostly in the front-seven through the draft but also by signing Kendall Fuller and Landon Collins in consecutive years) and should improve despite a shoddy secondary. Given the narrative around the Rams last season, I was shocked to see they fared this well. I guess much of the blame should be placed on the offense regressing, but people forget that the Rams would have made the playoffs with the new format this season. They curiously parted ways with Wade Phillips, but if the offense bounces back they can be in contention for that division (which I outlined in my post about offenses). Tampa Bay made a huge jump from 2018 (see more below) but they were an under-the-radar unit last year. With Jameis Winston throwing interceptions like it was going out of style, the defense often found itself in disadvantageous situations with respect to field position, so they gave up points despite success at the per-play level. Tom Brady taking care of the ball should make the unit appear better to the average fan this year. Denver did not give up too many touchdowns last year but are middling when looking at EPA per play. They could not get off the field, however, and were sort of a Vikings-lite with regards to the bend-don't-break style. I would be surprised if they prevent big plays at quite the same rate last year, so if they want to compete for a playoff spot they are going to have to tighten things up play to play.
I buried the lead here, but the main source of positive or negative changes in defensive performance are takeaways. Takeaways were the driver of the Bears success in 2018 and the Jaguars success in 2017. Year over year correlation in takeaways is a microscopic 1.73 percent, effectively carrying no signal and correspondingly all noise. Check out how the magnitude of takeaway regression is affected by takeaways in year N-1:
Its is extremely stark. Looking at the location of the trend line, for every takeaway you gain in a season, you should expect to give that back the following year. It is remarkable how consistently inconsistent defensive takeaways prove to be. How often a defense is able to turn over an offense is not stable year over year so when handicapping defenses, assuming they will have an average number of takeaways (about 20 over a full season) instead of a figure close to the previous year is most prudent. Teams with over 30 takeaways in a year, on average, have 15.5 fewer the next year. Teams with less than 15 have 11.85 more the following year. The 2018 Bears had 34 takeaways and followed that up with 19 in 2019. The 2017 Jaguars had 32 and then a measly 15 in 2018, less than half of their 2017 figure. The Niners had just 7 takeaways in 2018 and then 25 in 2019. That is more than three times the amount of takeaways. When the Niners defense looks worse this year, I would highly encourage you to remember this fact. Who else is going to be affected by this circumstance in 2020?
Pittsburgh sticks out like a sore thumb. In 2018 the Steelers had an awful 14 takeaways. In 2019 that increased to 37, an astounding 23 more takeaways year over year. This was by far the largest change in my sample. This team is really leaning heavily on Roethlisberger returning to form. If he is a shell of himself, this team could really struggle in a very competitive division. The Patriots and Bucs are two other teams that should be a bit less unlucky this season, but with so much uncertainty associated with their offenses, it will be interesting to see how that affects their season win totals. The Chargers and Raiders should be aided by some positive regression. The Raiders in particular were awful last year, displaying the brutal combination of bad per play defense and getting unlucky with takeaways. It was not surprising they gave up touchdowns at the highest clip last year. I was skeptical looking at Denver's EPA numbers but the fact that they gave up so few touchdowns per drive without a lot of takeaways is encouraging. This probably also has something to do with not being able to get off the field. Buffalo was average here and had a great defensive unit. They have a great roster, one of the best if you do not include the quarterback. Not only is the roster great, but it is extremely young. For the first time in ages, the top of the AFC East is going to be a dog fight. The Vikings were incredibly lucky, they had 30 takeaways after just 19 in 2018. I alluded to their inability to get off the field and fortune with explosive plays. The talent drain in the secondary along with this oncoming takeaway regression is a bit worrisome. I am sure they expect newly acquired Yannick Ngakoue to help stave off this occurrence. Seattle is the last team I will touch on. They were middling in EPA allowed per play despite this gaudy takeaway total. But like the Vikings, they traded for a star in Jamal Adams. I do not think one player can really change the fortunes of a defense, so Russell Wilson is going to be tasked again with propping up a mediocre defensive unit.
No comments:
Post a Comment