2015 Play-Action Defense

by Sterling Xie

After looking at play-action offense earlier this week, we're flipping over to the other side of the ball and looking at how defenses fared against play fakes. Again, all play-action data below is courtesy of ESPN Stats & Info.

Even though offenses experienced some weird trends using play-action this season, the league-wide defensive numbers remained largely on par with what we've seen in past seasons. Most strikingly, the correlation between play-action DVOA and non-play-action DVOA on defense remained surprisingly low at 0.38 for 2015. That's almost exactly where it has resided the past two seasons, as the correlation was 0.34 in 2014 and 0.38 in 2013. Recall that the correlation on offense was all the way down to 0.16 this season, even though the relationship is usually similar or stronger on that side of the ball. As usual, though, there wasn't necessarily a ton of consistency on defense, as faring well against play-action didn't typically ensure that a defense fared well overall.

However, defenses did fare noticeably worse against play-action overall. The average play-action DVOA versus passes and scrambles last season was 20.6% (that figure doesn't change meaningfully if you exclude scrambles and only look at play-action passes). Since we started tracking play-action stats in 2007, only the 2010 season, when defenses posted a 22.2% average DVOA against play-action, has seen a higher number. DVOA against non-play-action dropbacks rose slightly to 6.6%, and just like 2014, only nine defenses were better against play-action than normal dropbacks.

The table below lists play-action defensive numbers for the 2015 regular season. Unlike with the offensive play-action table, we're not bothering with the column which excludes scrambles to look at DVOA on only play-action passes. Most defenses were slightly better when getting rid of scrambles, but none by more than -3.9%. League-wide, the average defense against play-action including scrambles had a DVOA of 20.6% and allowed 7.8 yards to play. Remove scrambles from those play-action plays, and those numbers go to a 20.1% DVOA and 7.9 yards allowed per play.

Defenses are sorted by ascending DVOA against play-action, with the best being at the top. Sandwiched around that is play-action rate and DVOA against non-play-action dropbacks. The final column shows the difference between play-action and regular dropbacks, with a lower number there indicating a defense which was better against play-action.

No one should be surprised to see the Broncos topping the chart for both play-action and normal dropbacks. Denver especially lapped the competition against play-action, where it has an argument as the best defense we've measured in our nine seasons of tracking this stat. The 2015 Broncos don't own the lowest play-action DVOA during that span; that honor belongs to the 2009 Bills. However, 2009 was a really strange season for play-action, with an unusually low league-wide average of 12.5% and more variability than we've seen in any other year. (Let's be honest, it may have just been a strange season for our game charters and the way they marked play-action.) The 2009 Packers also posted a lower play-action DVOA than what Denver managed last season, while the Super Bowl champion Saints of that season had nearly the same figure at -38.8%. However, when comparing league-leading DVOA figures to the league average based on standard deviation, no defense has been more dominant versus play-action than last year's Broncos.

Denver also faced the lowest percentage of play-action dropbacks, so opposing offenses at least knew that nobody in orange was getting fooled. When poring over the Broncos chapter in Football Outsiders Almanac 2016, you'll get used to statements like that. Wade Phillips' unit led the league in what felt like every single defensive category, and play-action certainly didn't qualify as one of their very few weaknesses. Denver may not have been the best defense of the DVOA era, but it wasn't that far off.

The Raiders and the Rams were two of Denver's closest play-action challengers, a huge step up from where both defenses resided in 2014. After ranking as two of the four worst play-action defenses two seasons ago, both jumped up into the top five, easily the two largest improvements from 2014 to 2015. That the Rams were so bad against play-action in 2014 is surprising to begin with, and their improvement seemed to represent an example of regression to the mean, as they made only minor changes to their already talented defense. Next season could be trickier, though, with rangy free safety Rodney McLeod no longer around to take away many of the deep routes offenses will turn to off of play-action. Oakland, on the other hand, has continued a multi-year process of upgrading its pass defense, adding Bruce Irvin, Reggie Nelson, and Sean Smith to a defense that just posted its best DVOA since 2010. The Raiders overall had some shaky underlying indicators, which might be cause to tamper down some of the preseason hype, but play-action defense certainly didn't fall into that category.

The Raiders and Rams were two defenses that made due against play-action without the aid of an excellent pass rush, as both ranked in the bottom half of the league in adjusted sack rate. This was actually more the norm rather than the exception last season, as just three of the 10 best play-action defenses -- Denver, Kansas City and Minnesota -- also ranked in the top 10 in ASR. Last year, I looked at the correlation between ASR and play-action DVOA and found that there was a moderate relationship usually hovering around -0.20 to -0.25. This year, though, the correlation between the two was just -0.08. While not unprecedented, it's the second-weakest correlation we have out of nine seasons, trailing only the -0.02 mark from 2011.

Looking for any unifying quality among last year's best play-action defenses is a bit of a stretch. The logical place to turn would be the secondary, and some defenses such as the Panthers and Jets had multiple cornerbacks with strong charting stats. However, numerous others like Oakland, St. Louis, and (surprisingly) Denver got by despite mediocre charting stats from their starting corners. Heck, not every good play-action defense was even a good defense to begin with. The correlation between play-action DVOA and overall defensive DVOA was 0.65 -- fairly strong, but far lower than the 0.91 correlation between non-play-action DVOA and overall DVOA. This mirrors the trend we saw on offense, and probably shouldn't come as a huge surprise given that there are nearly five times as many passes and scrambles which come from normal dropbacks.

The flip side is that it's not particularly difficult to explain why most of the worst play-action defenses fared so badly. For the most part, those defenses were also mediocre or worse at everything else. Among the bottom-10 finishers in play-action DVOA, only the Texans and Packers finished better than 19th in overall defensive DVOA. The historically bad Saints defense finished last against both play-action and normal dropbacks. Just as the Broncos faced the lowest rate of play-action dropbacks, New Orleans faced the highest, with opponents using play-action on 23.9 percent of their dropbacks against the Saints. New Orleans wasn't too far behind the pack in play-action, though that was mostly a byproduct of severe incompetence from Tennessee, Houston, and San Francisco. No defense since 2009 and only six overall have posted worse raw DVOA totals than what the Saints did last year. However, the Saints lagged much further behind on normal dropbacks, finishing 2.5 standard deviations worse than the league average in that category. Only the 2009 Jaguars, who accumulated a 43.2% DVOA, fared worse in the past nine seasons.

In terms of splits, the Bills had the largest difference in favor of play-action, with a play-action DVOA that was 30.0% lower than their non-play-action DVOA. That difference dwarfed the 2014 49ers, who led the league two years ago with a -21.6% difference, though you only have to go back one more year to find the 2013 Browns, who posted a -34.1% difference. Buffalo was a top-five defense against play-action, but it was the other four-fifths of dropbacks where Rex Ryan's unit underachieved. The Bills regressed more than any other defense against non-play-action dropbacks, as Buffalo saw its DVOA rise 39.8% from 2014, when it posted a league-best -19.4% DVOA.

But generally, Ryan's Bills were the outliers for teams installing new defensive systems. We've already discussed the successes of the Broncos under Phillips and Raiders under Jack Del Rio. The Jets can probably attribute their success to Mike Maccagnan's offseason spending, but Todd Bowles' aggressive scheme did produce top-10 results by both dropback types. Joe Barry transitioned Washington to a 3-4 scheme with more one-gapping techniques than what Jim Haslett employed and helped them improve against both types of dropbacks as well. And while both Chicago and Atlanta regressed versus play-action, they also improved in the more robust non-play-action sample. The Falcons and Bears faced significant retooling projects when implementing schemes from Dan Quinn and Vic Fangio, respectively, as both teams had personnel left over from the previous regimes that was incompatible with the new defenses. Each front office pinpointed defense as the focal point of its offseason. Atlanta took a more draft-oriented approach than Chicago, which has doled out significant money for free agents like Jerrell Freeman, Danny Trevathan, Pernell McPhee, and Akiem Hicks the past two years. Given the ongoing makeovers, further improvement in 2016 would hardly be a surprise.

Play-action may only represent 20 to 25 percent of all dropbacks in any given season, but it's clear offensive coordinators take notice of which defenses bite on those play fakes more often. The correlation between play-action percentage and play-action DVOA was 0.66. This shouldn't be too stunning, as play-action tends to produce bigger plays that are easier to notice on tape. It takes a subtler eye to notice the safety who doesn't pass off the receiver leaving his zone in time, but anyone can point out the dummy who rushes forward for a handoff that never materializes while a receiver sprints right past him. Curiously, though, the correlation between those two stats on offense was -0.08, suggesting that offensive coordinators are much worse at recognizing when their own unit happens to execute play-action well.

Maybe some of that is purposeful. After all, play-action is akin to a change-up pitch, and nobody wants the other side to recognize when that change-up is coming. Still, given how defenses consistently struggle against play-action year after year, it's possible we haven't yet reached the point of diminishing returns for offenses who might fear that using too much play-action will curtail its effectiveness.