Writers of Pro Football Prospectus 2008

05 Jul 2016

2015 Play-Action Offense

by Sterling Xie

As we move closer to releasing Football Outsiders Almanac 2016, we'll preview some of the charting data which appears in the book. We have already done some of this with passing plus-minus, receiving plus-minus, YAC+, and adjusted interceptions. Now play-action passing is next on the docket. Last season saw some unusual trends across the league, as well as one offense which enjoyed by far the largest improvement we've seen the past five years. Information on play-action usage comes courtesy of ESPN Stats & Information.

League-wide, the rate of play-action usage dropped to just under 19 percent, the lowest play-action rate since 2011, after holding steady the past two seasons at around 21 percent. The standard deviation in play-action usage also fell, dropping from 5.6 percent to 3.8 percent in 2015. Play-action success held steady, inching up very slightly from a 24.0% DVOA in 2014 to 24.5% last season. The yards per play-action play also went up, from 7.5 to 7.9. The league DVOA on non-play-action passes rose a little more from 12.3% to 14.0%, but play-action passes will probably always be a little more successful because of their situational usage. Coaches tend to call play-action passes on less obvious passing downs and they often lead to deeper throws, which usually generate higher DVOA figures.

The following table lists play-action data for the 2015 regular season only. One change to this year's table: We're putting back the section which excludes scrambles and looks at only play-action passes. We got rid of it last year because only one team had a difference of greater than 4.0%. This year, four teams met that threshold, and two more were above 3.0%. Offenses are sorted by descending rate of play-action usage (percentage of dropbacks). The sections next to usage list results on all play-action plays (including scrambles), play-action passes only, and normal passes. The final section shows the difference between standard play-action and normal passes. We subtracted normal pass DVOA from play-action pass DVOA, so a high ranking in the difference column indicates the offense performed better with play-action.

Play-Action Usage, Offenses, 2015
Offense PA Pct Rank With PA (Pass/Scram) With PA (Pass Only) No PA Difference
Yds/Play DVOA Rank Yds/Play DVOA Rank Yds/Play DVOA Rank Yds/Play DVOA Rank
MIN 27% 1 7.4 29.0% 12 7.5 31.1% 10 5.6 3.0% 22 1.8 25.9% 10
CAR 27% 2 6.4 8.8% 28 6.4 4.2% 29 7.1 37.1% 4 -0.7 -28.3% 29
SF 25% 3 6.9 7.5% 29 6.7 -1.5% 30 5.6 -4.1% 25 1.3 11.6% 19
STL 24% 4 9.0 33.8% 9 9.1 31.9% 9 5.0 -28.4% 32 4.0 62.2% 1
BAL 24% 5 6.7 5.0% 30 6.8 5.2% 28 6.0 7.2% 20 0.7 -2.2% 24
SEA 24% 6 5.7 25.4% 17 5.6 21.9% 17 7.6 56.0% 1 -1.9 -30.6% 30
HOU 22% 8 8.0 17.0% 20 8.0 15.6% 21 5.4 3.1% 21 2.7 13.9% 17
ATL 22% 7 8.8 30.3% 10 8.9 30.7% 11 6.2 0.8% 23 2.7 29.5% 9
PHI 21% 9 9.4 13.3% 25 9.5 14.0% 25 5.5 -5.8% 27 3.9 19.1% 14
TEN 21% 10 7.1 11.8% 26 7.2 11.7% 27 5.7 -10.2% 30 1.4 22.0% 12
CIN 20% 11 7.9 46.5% 5 8.1 48.9% 5 7.1 51.4% 2 0.8 -4.9% 25
KC 20% 12 6.1 15.6% 21 6.2 16.0% 20 6.6 25.6% 11 -0.5 -9.9% 27
DEN 19% 13 7.8 14.2% 23 7.8 14.5% 23 5.8 -6.1% 28 1.9 20.3% 13
MIA 19% 14 8.6 35.1% 8 8.6 34.2% 8 5.7 -5.0% 26 2.9 40.1% 4
WAS 19% 15 10.3 58.6% 3 10.4 58.4% 3 6.4 25.0% 12 3.9 33.6% 7
TB 19% 16 9.2 49.2% 4 9.7 50.0% 4 6.5 8.5% 19 2.8 40.6% 3
Offense PA Pct Rank With PA (Pass/Scram) With PA (Pass Only) No PA Difference
Yds/Play DVOA Rank Yds/Play DVOA Rank Yds/Play DVOA Rank Yds/Play DVOA Rank
NE 18% 17 8.4 40.1% 6 8.4 40.1% 6 6.8 34.2% 6 1.6 5.9% 21
OAK 18% 18 8.1 28.8% 13 8.2 29.3% 12 5.9 18.8% 14 2.2 10.0% 20
DET 18% 19 8.1 37.8% 7 8.2 39.5% 7 5.9 13.8% 15 2.2 24.1% 11
NO 18% 20 6.2 15.4% 22 6.2 15.4% 22 7.3 33.6% 7 -1.1 -18.2% 28
ARI 17% 21 10.5 70.7% 2 10.5 71.7% 2 7.6 35.8% 5 2.9 34.9% 6
CHI 17% 22 7.7 18.5% 19 7.6 17.9% 19 6.7 24.8% 13 1.1 -6.3% 26
NYG 17% 23 7.9 29.1% 11 7.9 29.1% 13 6.6 10.3% 18 1.3 18.8% 15
BUF 17% 24 9.9 72.3% 1 10.2 72.4% 1 6.2 13.6% 16 3.7 58.7% 2
IND 16% 25 8.0 27.1% 15 8.1 26.6% 15 5.3 -9.1% 29 2.7 36.2% 5
GB 16% 26 5.2 -10.8% 32 5.0 -15.1% 32 6.4 26.6% 9 -1.2 -37.3% 32
NYJ 16% 27 7.1 25.9% 16 7.2 25.9% 16 6.6 26.6% 8 0.5 -0.7% 23
CLE 15% 28 7.8 11.7% 27 8.0 14.1% 24 5.4 -3.0% 24 2.5 14.7% 16
JAC 15% 29 8.0 23.3% 18 8.0 20.3% 18 6.3 10.3% 17 1.8 13.0% 18
DAL 15% 30 7.4 13.7% 24 7.4 12.5% 26 6.1 -16.8% 31 1.3 30.5% 8
PIT 14% 31 8.3 3.6% 31 8.4 -2.7% 31 7.7 39.0% 3 0.6 -35.4% 31
SD 10% 32 7.6 27.3% 14 7.6 27.3% 14 6.4 25.9% 10 1.3 1.3% 22
NFL 19%
7.8 25.3%
7.9 24.5%
6.3 14.0%
1.5 12.3%

As you'll see if when you read FOA 2016, the Bills were a bit bipolar offensively, with splits that were alternatively excellent and horrendous. Play-action was part of Buffalo's good side, as Tyrod Taylor led the league's best play-action offense last season. You might assume that including scrambles probably helped push Buffalo ahead of Arizona, but the Bills still finished first in the passes-only column with a DVOA basically indistinguishable from their standard play-action DVOA. The next logical step would be for Greg Roman to increase Buffalo's play-action usage in 2016, given Taylor's proficiency on deep passes (81.1% DVOA, fourth among qualifying quarterbacks), and the fact that the Bills only ranked 24th in play-action percentage despite their efficiency. Roman used plenty of play-action during Colin Kaepernick's peak seasons, as San Francisco ranked sixth and fifth in play-action usage during the 2013 and 2014 campaigns, respectively.

The most fascinating team on this chart is probably the Tampa Bay Buccaneers. As alluded to in the lead, the Bucs improved their play-action DVOA by a whopping 105.0% from 2014, by far the largest year-over-year jump we've seen since we started writing articles about this topic five years ago. Helpfully, Tampa Bay also increased its play-action usage more than any team from 2014, taking advantage of this huge improvement. Upgrading from Josh McCown to Jameis Winston obviously doesn't hurt, but Winston's top three targets (Mike Evans, Vincent Jackson, and Austin Seferian-Jenkins) were hurt, inconsistent, or both for large stretches in 2015, and an undrafted rookie (Donteea Dye) was frequently the No. 2 or 3 target in the passing game. Tampa Bay's DVOA on normal pass plays improved substantially as well, though by nowhere near the same magnitude. The Bucs are banking that their young offense will continue growing with Dirk Koetter still in place, as the retired Logan Mankins is the only 2015 starter missing from the projected 2016 lineup. You can debate Lovie Smith's unceremonious dismissal, but this type of evidence helps explain the rationale of why Tampa wanted to keep Koetter around so badly.

By play-action frequency, the Vikings topped the league at just over 27 percent, which wouldn't have even ranked in the top five in 2014. Still, this fits like a glove over the narrative of the 2015 Vikings: With Adrian Peterson reassuming his role as the central focus of Minnesota's offense, the passing game became the offense's change-up, allowing play-action to be more surprising and increasing Teddy Bridgewater's chances of hitting a receiver downfield.

This idea of running to set up play-action is something you'll hear on just about every broadcast. Of course, the very first article on this site debunked the idea of "establishing the run." Moreover, the fact that the best play-action offenses tend to also be the best non-play-action offenses further rejects the premise of needing to run the ball more for play-action to work. In fact, correlation between play-action percentage and play-action DVOA in 2015 was actually negative (-0.15).

Nevertheless, offenses still ostensibly believe that running does indeed set up play-action to work more effectively. Four of the top five teams in rushing percentage last season were among the top six in play-action percentage, and the correlation between the two was a rather strong 0.45. As it turns out, this relationship has been growing stronger throughout our dataset (which dates back to 2007), implying that coaches are believing this premise more and more:

Correlation between
Rushing% and PA%, 2007-15
Year Correlation
2007 -0.16
2008 0.37
2009 0.00
2010 -0.04
2011 0.27
2012 0.44
2013 0.28
2014 0.28
2015 0.45

The correlation peaked in strength last year, but only just barely, and it continued a pattern of "run-heavy" teams continually using more play-action. It's certainly not bad to use lots of play-action; as we mentioned, most teams tend to fare better on play-action dropbacks than normal ones. Still, it's interesting to wonder if teams that use lots of play-action do so because of the running game, or because they happen to have strong-armed quarterbacks who are more likely to hit those downfield passes. Some of the top play-action users from last season fit into the latter category, but many of those teams, like the Seahawks and Panthers, fit into the former category as well. The dip in play-action percentage along with the dip in rushing percentage might also suggest that the two are linked.

Then again, don't tell Jeff Fisher that running doesn't set up play-action. The artists formerly known as the St. Louis Rams had the league's largest gap between their play-action and non-play-action DVOA, finishing dead last in the latter category but a much more respectable 10th in the former. The Rams took full advantage of this split, using more play-action than all but three offenses. The "Bear Raid" offense Jared Goff ran at Cal is wildly different from what he'll run in Los Angeles, but one common thread is that play-action is a prominent aspect in both systems. You'll likely see the Rams near the top of the play-action usage rankings again next year.

We've touched on the teams did well and/or used lots of play-action, but what about those on the other end of the spectrum? Stunningly, Green Bay and Pittsburgh were the two worst play-action teams in the league by DVOA. Both the Packers and Steelers were top-10 teams on normal dropbacks, yet went downhill when adding in the play fake. Green Bay was somehow the only team to finish with a negative play-action DVOA. This looks bad in isolation, but it's actually highly unusual for no other offense to be this bad at play-action. Seven teams in 2014 had a negative play-action DVOA, and the 2015 Packers are by far the best 32nd-place play-action offense we've had in nine seasons. The Steelers and Niners join Green Bay in the red if you throw out scrambles.

Seeing two very good passing offenses post poor play-action DVOA figures encapsulated a larger and unexpected trend from 2014. Remember how I said earlier that good play-action offenses tend to do well on normal dropbacks as well? That premise fell apart in 2015. The correlation between the DVOA in the two categories was 0.16, far lower than the typical 0.35 to 0.70 range we typically see. The play-action rankings were far weirder than the non-play-action rankings, where no one would be surprised to see the likes of the Rams, Cowboys, and Titans inhabiting the basement while the Seahawks, Bengals and Steelers sit at the top. In fact, by DVOA, overall passing offense had a much higher correlation with non-play-action passing (0.96) than with play-action passing (0.40).

It's hard to think of a reasonable explanation for this, and maybe we don't need to. Packaged plays are steadily creeping into more playbooks, and those plays often involve the types of read-option concepts which make them noticeably different from normal dropback pass calls. In theory, a quarterback could be noticeably better at executing packaged plays than he is at normal dropback passes. But they're not yet ubiquitous across the league, as the vast majority of play-action passes are of the traditional variety. The more efficient (and most likely) way to explain this is just to call it a one-year blip (and no, payment for this piece was not determined by word count).

What can offenses with new coaching staffs expect in 2016? Cleveland's play-action usage dropped more than any team from 2014 to 2015 after going from Kyle Shanahan to John DeFilippo, but we should see the pendulum swing back the other way considering that Hue Jackson used a fair amount of play-action in Cincinnati. The Eagles used tons of play-action under Chip Kelly (though less so last season), but Doug Pederson's offenses in Kansas City were always middling in terms of play-action usage. Moreover, Pederson hired Frank Reich as offensive coordinator, and Reich's Chargers finished last in play-action usage during both of his seasons as coordinator there. Don't expect much change in San Francisco or San Diego, though. The Niners were a heavy play-action usage team last season, and you can already see Kelly designing rollouts for either Kaepernick or Blaine Gabbert. Meanwhile, Ken Whisenhunt is back in the saddle for the Chargers. The last time he was their offensive coordinator in 2013, San Diego finished … last, again, in play-action usage. Finally, Adam Gase's offenses have always finished in the middle of the pack in play-action percentage, which is where the Dolphins already were anyways.

Later this week, we'll flip over to the other side of the line and check out how defenses fared against play-action.

Posted by: Sterling Xie on 05 Jul 2016

12 comments, Last at 07 Jul 2016, 8:38am by Jason_PackerBacker


by sandbun :: Tue, 07/05/2016 - 2:24pm

If a team is using PA then the QB is looking away from the defense for a while, so it would be harder to see if someone was breaking through the line. Seems like that would increase the likelihood of the QB being sacks. Was that looked at at all?

by tuluse :: Wed, 07/06/2016 - 4:42pm

I'm not sure this 100% true in this era of option handoffs.

by speedegg :: Wed, 07/06/2016 - 4:53pm

Not necessarily. Players (HB, FB, etc) still have blocking assignments in play-action and coaches will tell them if a blitz is coming to abandon the fake and fulfill your assignment. The exception would be the Dallas Cowboys where your (former) Fullback forgets his pass protection responsibilities and allows Romo to get pounded (sorry Cowboys fans).

Here's an old video from Eagles Film Room about the HB abandoning the fake and going straight to his pass protection assignment. Greg Cosell talks about it around the 2:45 mark. Yes, it's old. Yes, it's Rex Grossman as a QB for Washington, but the principles are still the same.

Link: http://www.philadelphiaeagles.com/videos/videos/Film-Room-Redskins-Passi...

by sandbun :: Wed, 07/06/2016 - 11:03pm

Cowboys and Vikings with AP.

by billprudden :: Tue, 07/05/2016 - 2:41pm

I feel like such a fool asking this, but can somebody explain to me how league average for DVOA for passes is 14%? Shouldn't the average be zero?

Or is it that total DVOA, including rushing plays, sacks, etc balances to zero?



by Aaron Schatz :: Tue, 07/05/2016 - 4:48pm

Whenever we list team DVOA, we're basing that on all plays that are included in DVOA. For defense, that includes both runs and passes. For offense, that includes runs, passes, AND certain other plays that penalize an offense without crediting a defense, such as aborted snaps, delay of game, and false starts.

Since passing is generally more efficient than rushing, that means that average leaguewide passing DVOA is positive and average leaguewide rushing DVOA is negative. This effect is even stronger on offense because offensive DVOA also considers a series of plays which are ALWAYS negative (aborted snaps and certain penalties).

I hope that answers your question. I need to add this answer to the FAQ page at some point.

by billprudden :: Wed, 07/06/2016 - 4:06pm

Gotcha, and thanks.

As a follow-up, an article about the gap between rushing DVOA and passing over the years, and in response to particular rule changes, would be interesting... especially if that gap is steadier than we all think!


by jtr :: Tue, 07/05/2016 - 3:38pm

You looked at PA% vs PA DVOA and Run% vs PA%, but I don't think either of those get to the heart of the "establish the run" mantra. What's the correlation between Run% and PA DVOA? That's the correlation that will actually tell us whether running the ball makes play action more effective.

by Sterling Xie :: Wed, 07/06/2016 - 6:03pm

Fair critique, and I definitely did think about including PA DVOA vs. run %. Since you asked, the correlation last year was -0.08. Super weak and actually the opposite of what you might intuitively expect (as we found with PA % vs. PA DVOA, mentioned in the article). The reason why I didn't do this was because I was more trying to answer the question, "If a team runs a lot, they might assume play-action will work better. Does this mean they'll end up using play-action more?" Under that premise, it's taken for granted that PA DVOA will increase with more rush attempts (even though, of course, that's not really true).

by Jason_PackerBacker :: Thu, 07/07/2016 - 8:38am

I may be misreading this thread, but wouldn't a better alternate comparison be between run DVOA and PA DVOA? That would allow you to compare effectiveness. If I'm up against a running team I am confident I can stop, I won't necessarily key any harder against apparent run plays -- I'll just keep doing my thing, allowing the structure of the defense to remain uncompromised. However, if I am getting totally gashed on most runs, even if they are not a large percentage of the offense's plays, I might start cheating.

Who knew that so much would ride on one collarbone?

by MilkmanDanimal :: Wed, 07/06/2016 - 5:53pm

Is there any real tendency in play action success from year-to-year? There was clearly a big upgrade at QB, RB (with a healthy Doug Martin), and the OL last year in Tampa, but that big of a jump is pretty crazy. I would like to think that portends well for the future?

by Sterling Xie :: Wed, 07/06/2016 - 6:09pm

Nope, it's pretty inconsistent from year-to-year. Scott mentioned this a couple years ago for this column: http://www.footballoutsiders.com/stat-analysis/2014/2013-play-action-off...

It's easy to assume that a team like Tampa which upgraded its offense big time would see a jump in PA DVOA. But then why did teams like Seattle and Carolina, who saw big time jumps from their QBs, get worse? Just something to consider.