Football Outsiders

Innovative Statistics, Intelligent Analysis

2017 Play-Action Offense

by Bryan Knowles

Pop the champagne and unfurl the banners, because 2017 saw play-action come roaring back with a vengeance!

… What, I'm the only one who celebrates Play-Action Day? You guys are missing out.

Over the past decade, play-action passing peaked from 2012 to 2014, in what historians will surely call the Read Option Era. Teams averaged play-action on 21 percent of their passing plays in those years, as the Colin Kaepernicks and Robert Griffins roamed the Earth. That rise was created almost solely by what we called the "Big Six" -- Washington, Seattle, Philadelphia, Carolina, Minnesota and San Francisco. The rest of the league stayed more or less flat, and when defenses adapted to the read option and peak of the craze had died off, play-action frequency fell with it.

The frequency of play-action passing in 2015 and 2016 was roughly the same as it had been in 2010 and 2011 -- advanced from the primitive days of the '00s, but fairly flat overall. It looked like the bump was a temporary thing. The read option wasn't exactly a Wildcat-esque fad, but it was a temporary trend that bumped everyone's numbers up a bit.

Enter 2017.

This chart shows the change in play-action frequency over the past decade. The orange bars around each data point represent the standard deviation in each year. The larger the bars, the wider the spread of play-action around the league. The bars around 2012 are the largest, because it had the largest gap between the teams that used play-action the most and the ones that used it the least. That indicates that, although the average play-action rate had increased then, it wasn't a uniform increase; there were some teams using it much more frequently than others. The bars around 2017 are much smaller, because the increase in play-action is more evenly distributed around the league. Rather than being, say, just the RPO teams or just the Shanahan/McVay school pushing play-action, nearly every team is using play-action more frequently. The increase in league-wide play-action rate was really a joint effort.

Twenty-four teams saw their play-action rate increase from 2016 to 2017. A dozen teams saw their rate increase by at least five percentage points. The five teams that increased the most went to the playoffs, including all four conference championship teams (plus the Rams). It really was the best year for play-action in recent history, even if it was fractionally less successful overall this year than it was in years past.

The following table lists play-action data for the 2017 regular season only. Offenses are sorted by descending rate of play-action usage (as a 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, 2017
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 30% 1 8.8 69.0% 1 8.9 69.0% 1 6.4 24.8% 11 2.4 44.2% 7
LAR 29% 2 9.0 44.3% 7 9.1 44.4% 8 6.4 24.2% 12 2.6 20.0% 14
NE 28% 3 8.1 39.1% 10 8.1 39.1% 10 7.0 50.5% 1 1.1 -11.4% 27
PHI 27% 4 6.3 15.0% 20 6.3 11.8% 22 6.5 45.5% 3 -0.2 -30.6% 30
CIN 27% 5 7.5 18.7% 19 7.5 18.0% 19 5.2 -3.1% 24 2.3 21.8% 9
BAL 25% 6 6.4 -6.3% 29 6.4 -6.5% 28 4.7 -2.6% 23 1.6 -3.7% 23
SF 24% 7 7.0 27.5% 16 7.1 26.9% 16 5.8 6.1% 17 1.2 21.4% 10
DEN 24% 8 6.8 -6.3% 28 7.0 -11.2% 30 5.2 -17.1% 31 1.6 10.9% 18
SEA 23% 9 7.6 22.5% 18 7.6 20.8% 18 6.0 34.1% 7 1.6 -11.6% 28
MIA 23% 10 7.5 45.6% 6 7.6 45.8% 5 5.2 -13.7% 28 2.3 59.4% 1
TEN 23% 11 8.9 44.2% 8 9.1 45.5% 7 5.2 -8.0% 27 3.7 52.2% 2
JAX 23% 12 8.7 50.9% 4 8.9 53.0% 4 6.0 1.7% 20 2.6 49.2% 3
ATL 22% 13 8.7 46.9% 5 8.7 45.6% 6 6.9 26.0% 10 1.8 20.9% 12
KC 22% 14 6.8 38.6% 12 6.9 38.2% 11 7.3 35.5% 6 -0.5 3.1% 21
IND 22% 15 7.0 1.9% 26 7.1 1.3% 25 5.1 -14.8% 29 1.9 16.7% 16
GB 22% 16 5.7 5.7% 25 5.4 -0.2% 26 5.4 5.3% 19 0.3 0.3% 22
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
BUF 22% 17 7.0 14.1% 21 7.2 16.6% 20 5.4 -6.5% 25 1.7 20.6% 13
HOU 22% 18 7.7 38.8% 11 7.6 37.8% 12 5.8 -7.4% 26 1.9 46.2% 5
DAL 22% 19 6.7 25.7% 17 6.7 23.3% 17 6.2 5.9% 18 0.5 19.8% 15
WAS 21% 20 8.3 55.1% 3 8.4 53.5% 3 6.0 6.3% 16 2.3 48.7% 4
NYG 21% 21 5.8 8.2% 24 5.8 5.0% 24 5.4 1.6% 21 0.4 6.6% 20
TB 21% 22 8.1 36.4% 13 8.2 34.6% 13 6.5 29.8% 8 1.6 6.6% 19
NO 20% 23 7.0 11.8% 23 7.0 11.5% 23 7.6 42.9% 4 -0.6 -31.1% 31
DET 20% 24 9.4 60.6% 2 9.4 59.5% 2 6.3 15.6% 13 3.1 45.0% 6
CHI 20% 25 6.0 12.0% 22 6.1 14.3% 21 5.6 -15.0% 30 0.3 27.0% 8
CAR 19% 26 6.8 28.8% 15 6.9 30.1% 15 5.9 7.6% 14 0.9 21.2% 11
LAC 18% 27 8.4 40.5% 9 8.4 40.5% 9 7.3 45.5% 2 1.1 -5.0% 24
ARI 17% 28 6.3 -20.0% 32 6.3 -20.0% 32 5.7 -2.4% 22 0.6 -17.6% 29
CLE 16% 29 6.1 -11.9% 31 6.1 -16.1% 31 5.3 -26.1% 32 0.8 14.2% 17
OAK 14% 30 5.9 -7.7% 30 5.9 -7.5% 29 6.6 26.3% 9 -0.7 -34.0% 32
NYJ 14% 31 5.9 1.3% 27 5.6 -1.9% 27 5.9 6.6% 15 0.0 -5.3% 25
PIT 11% 32 6.6 31.5% 14 6.7 32.0% 14 7.6 37.8% 5 -0.9 -6.2% 26
NFL 22%
7.3 25.9%
7.4 25.0%
6.0 11.6%
1.3 14.3%

What's behind this spike in play-action? Some of it comes from Al Michaels' favorite new buzzword, the run-pass option. While RPO plays differ from traditional play-action passes -- the offensive line is generally run blocking on RPOs; sometimes the decision is made before the snap and the play-fake isn't involved; and sometimes you actually hand the ball off -- the end result will often be a quarterback faking a handoff to a running back and then throwing a pass, which will pump the play-action numbers up.

This evolution make sense with the influx of collegiate quarterbacks with backgrounds in spread-option offenses, rather than "traditional" schemes. More and more teams are mining the college ranks for strategic upgrades (sorry, Andy, but people who love football's strategy can enjoy college football, too) in order to find the best plays for the players they're getting, rather than trying to squeeze this generation of players into pre-existing boxes. It's not surprising to see these sorts of package plays making an impact at the pro level. While RPOs made their NFL debut long before last season (check out Michael Vick and the 2004 Falcons), they really burst into the spotlight in 2017, and we should expect to see more of them in 2018. Copycat league, and all.

That doesn't explain all of the rise in play-action, though; otherwise, we'd expect to see the teams that heavily bought in on RPOs in 2017 to top the list, and everyone else to still be back in the same pack as previous seasons—a new version of the Big Six. That's not what's happening. The Rams, Patriots, and 49ers are all near the top of the play-action leaderboards, and they almost never touch RPOs. No, there's more going on here.

Play-action passing makes passing easier. Defenders have to respect the possibility of the run, and so hesitate in their pass rush and their coverage. It makes sense, then, that in a year when so many of the best teams in the league had inexperienced or just lower-quality players at the quarterback position, we'd see coordinators dialing up more play-action to give them every benefit they could.

Spotty, inexperienced and unproven quarterbacks are all over the top of the play-action leaderboards. The teams that saw the biggest jump from 2016 to 2017 include:

  • The Rams and Eagles, who saw the biggest and third-biggest increases, respectively. Both were starting second-year players coming off of negative DVOA seasons as rookies.
  • The Vikings, with the second-biggest increase, had Case Keenum, who had never had a passing DVOA better than -19.5% in a full season entering 2017.
  • The Jaguars and Ravens, ranked fifth and sixth, had veteran quarterbacks who had finished outside the top 20 in DVOA in both 2015 and 2016.
  • The Packers, ranked seventh, had to spend a lot of the season with Brett Hundley under center.
  • The Broncos, ranked eighth, shuffled their way through a trio of uninspiring quarterbacks.

If you don't have the luxury of having a Tom Brady, Aaron Rodgers, or Russell Wilson under center, you'll want to do everything in your power to make life easier for your quarterback. Play-action passing is one way to do just that.

Of course, some teams do have the luxury of having a Tom Brady under center. The Patriots saw the second-biggest jump in play-action in 2017, and it's not like New England had to scheme around Brady's weaknesses. Brady saw his average depth of target rise from 8.1 to 9.6, and play-action passing does help buy time and space for the deep ball, so that might be part of the reasoning behind the Patriots using more play-action last season. There's probably a more simple reason, however. They're good plays, Brent.

DVOA on play-action passes was at 25.9% last season, while DVOA on all other passing plays was just 11.6%. That 14.3% gap is slightly down from last season, but is still the second-highest gap we have ever recorded. Play-action passes are completed at a higher rate. They're (generally) thrown further downfield, producing more yards (though stick a pin in that, we'll come back to it in a moment). They work regardless of whether your team is actually any good at running the ball. The misdirection provides a benefit in all but the most obvious passing situations, and even then it doesn't actually hurt. Ben Baldwin had a couple of great pieces this offseason going into great depth on these topics, and I encourage everyone to re-read them, but in short: play-action works for most teams most of the time.

Twenty-two teams had a higher DVOA with play-action than without it. Half the league saw their DVOA increase by at least 15 points, while only four teams saw their DVOA drop by that much.

Interestingly, two of the teams that saw their play-action frequency increase the most had two of the largest drop-offs while actually using play-action. New England's -11.4% difference isn't great, but it's also not the end of the world; they averaged an additional 1.1 yards per play out of play-action, their 8.1 yards per play was 10th highest in the league, and they still were at nearly 40.0% DVOA with play-action. That's good enough that I wouldn't be overly concerned with the play-action passing game if I were Josh McDaniels, and besides: they had been better with play-action than without it every other year since we started tracking this. I wouldn't be concerned over what could just be a one-year fluke.

Philadelphia's a more interesting case, especially with all the hype over the RPO. It's worth noting there's a significant difference between Carson Wentz and Nick Foles here. With Wentz, the Eagles had a 29.6% DVOA with play-action and a 53.8% DVOA without it -- still very high, and with great results with play-action anyhow. Under Foles (in the regular season), the Eagles had a -0.1% DVOA without play-action … and a -45.0% DVOA with it. Still, if RPOs are so good, then how come Philly's DVOA dropped in play-action? That wouldn't pass the eye test.

The thing to note with Philadelphia -- and with the other two teams to be at least 30 percent worse with play-action, Oakland and New Orleans -- is that they are three of the five teams to actually see their yards per play drop when they go into play-action. I think that's the culprit for the drop-off -- and that's actually just fine in their offensive scheme.

Part of the value of play-action passes is that they generally go deeper. Play-action passing is more effective on passes of a given distance (as measured by yards per play) starting on throws that travel more than 10 yards down the field, and appears to be more effective at all further distances. Play-action passes work in part by pulling linebackers closer to the line of scrimmage and then throwing it behind them to the vacated space.

But Philadelphia's average depth of target using play-action falls. It goes from 9.9 yards to 8.9. That's because of the run-pass option. You can't run deep routes on RPOs in the NFL. You're asking your offensive line to run block, meaning they're exploding upfield quickly. If you run anything but the shortest of routes, that's asking for an illegal man downfield penalty. That means you're running a lot of screens and quick slants; fast developing routes that mean you get the ball out of your hands quickly to ensure a legal play.

Play-action passing has a higher DVOA than non-play-action, but deeper passes have a higher DVOA than shorter passes in general. If your play-action game is set up to throw short and your non-play-action game is set to throw long, you'd expect to see your play-action game have lower DVOA.

So when you see Philly with such a drop-off in their play-action DVOA, it's not that they were bad at play-action. They had the most play-action passes of 8 yards or fewer, and their 26.3% DVOA on those plays ranked 11th and was higher than their 12.2% DVOA on non-play-action plays of the same distance (excluding sacks in both cases). It also doesn't take into account the fact that sometimes, when you run a run-pass option, you opt to run. Those plays don't get included here, because they're not play-action fakes if you actually hand the ball off. Those are just called "running plays," and generally, they're going to be good ones -- they'll come more frequently against less-loaded fronts, because if the defense is loading up the box, you'll just pass the ball instead.

I would go so far as to wager that, if more teams start to emulate Philly's success with the run-pass option, we'll see the gap between play-action success and non-play-action success continue to shrink. Not all of the benefits of play-action come from its deeper depths of targets, but some of it does. If more and more short play-action plays comes into vogue, we'll likely see play-action passing DVOA go down -- and, as long as it brings with it better results than non-play-action passing in the same distances as well as an improved run game, teams will be just fine with that.

Some other quickfire hits from the table before we go:

  • Even with Philly's RPOs producing shorter play-action in general, six teams still managed to have fewer yards per play than their 6.3. For Cleveland and Oakland, this was not intentional -- they tried to throw it deeper; they just weren't particularly good at actually completing passes in 2017. The Packers and Jets were among the league leaders in RPOs, explaining their results. That leaves the Giants and the Bears, both of whom should lay off the play-action running back screens going forward.
  • The Steelers are proof that there's more than one way to win a football game. While 10 of the 12 playoff teams used more play-action in 2017, Pittsburgh actually saw their play-action percentage drop from 14 percent to 11 percent last season, finishing dead last in the NFL. Pittsburgh has not been particularly adept at play-action the last three years, with a negative differential in 2015, 2016, and 2017. They did, however, see their DVOA on play-action jump from 12.7% to 31.5% last year; perhaps they were more selective in when they used it, rather than relying on more of Todd Haley's god-awful bubble screens.
  • The other playoff team to drop was Atlanta. Kyle Shanahan's teams have been in the top 10 of play-action percentage every single year this past decade, but he took his talents to San Francisco this past offseason. Atlanta had the single largest play-action dropoff in 2017, going from 27 percent to 22 percent. Unlike Pittsburgh, Atlanta offensive coordinator Steve Sarkisian would be well advised to bring some play-action back; Atlanta had a 20.9% DVOA difference when using play-action.
  • Minnesota led the league with a 69.0% DVOA out of play-action, but that's the lowest total to lead the league since 2012. At least the Vikings shouldn't expect too much of a drop-off with Keenum gone; Kirk Cousins and Washington ranked third in play-action DVOA last year. It will be interesting to see which side gives, though -- the Vikings ran play-action more than anyone else; Cousins ranked 20th in the league. They'll probably meet somewhere in the middle.
  • Washington should have been using more play-action a year ago! Cousins was quite good using play-action with both Shanahan and Sean McVay as previous coordinators, and Washington had the fourth-largest increase between using play-action and not.
  • Cousins aside, there has not been a lot of consistency in a team's play-action success from year to year. Take Seattle, for instance. Over the past three seasons, their DVOA difference in play-action has gone from -30.6% to 32.3% to -11.6%. I mean, you tell me.

Later this week, we'll flip to the other side of the ball and see how defenses handled all this extra play-action.

Comments

14 comments, Last at 22 Jul 2018, 11:46am

1 Re: 2017 Play-Action Offense

by Aaron Brooks Good Twin // Jul 17, 2018 - 3:16pm

Did Philly use RPOs to replace bubble screens?

Login or Register to post comments

4 Re: 2017 Play-Action Offense

by Bryan Knowles // Jul 17, 2018 - 4:14pm

"Replace" is a strong phrase, but not entirely inaccurate.

Philadelphia threw about 60% as many bubble screens last year as they did in 2016. They also used play action on half of them, as opposed to under 20% in 2016 -- those are both RPO related effects, most likely.

Login or Register to post comments

2 Re: 2017 Play-Action Offense

by Dan // Jul 17, 2018 - 3:19pm

Could the charting distinguish RPO passes from traditional play action plays, and give separate numbers for them?

Login or Register to post comments

3 Re: 2017 Play-Action Offense

by Vincent Verhei // Jul 17, 2018 - 3:33pm

Not without being in the Philadelphia huddle and knowing what play was called.

Login or Register to post comments

6 Re: 2017 Play-Action Offense

by jtr // Jul 18, 2018 - 10:35am

Right, and as with the read-option craze a few years ago, a lot of the time the threat of the read is more important than actually executing the read. Chip Kelly mentioned a few years ago that a lot of the plays the media labeled "read options" were actually designed RB runs; the threat of the QB occasionally pulling the ball was usually enough to keep the backside end from crashing, even if the QB wasn't actually making the read on a given play. Likewise in Philly, they could simplify things for Foles by having him actually have a defined read before the play, with opposing linebackers still slowed down a bit by worrying about RPO's.

Login or Register to post comments

8 Re: 2017 Play-Action Offense

by Dan // Jul 18, 2018 - 1:18pm

It should at least be possible to chart when the offense is using an "RPO look" or a "read-option look".

Play action is designed to create a certain kind of confusion/uncertainty in the defense, RPO plays are designed to create a slightly different kind of confusion/uncertainty, and RPO looks where the QB doesn't actually make a read are designed to create the same kind of confusion/uncertainty as actual RPO plays (while simplifying things for the offense). So there is some value (IMO) in putting RPO plays and RPO-imitation plays which don't have an option into the same bucket for analysis.

Login or Register to post comments

9 Re: 2017 Play-Action Offense

by The Ninjalectual // Jul 18, 2018 - 3:10pm

How do you define an "RPO look"?

Login or Register to post comments

10 Re: 2017 Play-Action Offense

by Bryan Knowles // Jul 18, 2018 - 3:45pm

The best way to define it would be if the offensive line -appears- to be run blocking, and the receivers -appear- to be running routes.

That's very, very subjective (are they running routes because they might get the ball, or are they running routes to trick the cornerbacks? Is the line going forward to open up a lane for the running back, or are they just establishing a pocket?), and any attempt to do so would involve a fair amount of guesswork and projection. There are borderline play-action cases (usually from rather lazy attempts at play-action), but there are fewer of them, and I'd say less subjectivity there.

That being said, I'm not saying it's NOT something I'll be keeping an eye out for in 2018. Just that making a definitive call one way or another is always going to be difficult without knowing the playcall. If there's one thing I would like the analytics world, in general, to embrace, it's uncertainty -- knowing what we don't know is just as important as knowing what we know, and false precision does no one any favors.

Login or Register to post comments

11 Re: 2017 Play-Action Offense

by Vincent Verhei // Jul 18, 2018 - 6:53pm

If we had time to pass for each play, you could probably separate RPOs from play-action passes. Any play-fake pass thrown within, say, 2 seconds of the snap is probably an RPO. Anything longer than that is definitely a play-action pass.

Login or Register to post comments

5 Re: 2017 Play-Action Offense

by jtr // Jul 18, 2018 - 10:32am

As far as Pittsburgh goes, I think I understand why they didn't use play action very much, and why it was less effective than their normal pass plays. PA limits the routes that running backs can run, since they have to commit to the run fake before they can run a route. Since he's such a weapon as a receiver, Pittsburgh would rather have Le'Veon Bell more involved in the pass patterns than having him execute a fake, and the DVOA splits suggest that they were on the right track with that.

Login or Register to post comments

7 Re: 2017 Play-Action Offense

by Bryan Knowles // Jul 18, 2018 - 12:54pm

It's a solid theory, although then you have to ask yourself why Sean McVay and the Rams were so happy to use play action when Todd Gurley is such a weapon as a receiver. More than one way to skin a cat and all that.

The other odd thing about the Steelers and play action is that they were the only team in the league to use it more often on second down than first down. This was true last year, as well. THAT one is a tough nut to crack.

Login or Register to post comments

12 Re: 2017 Play-Action Offense

by LionInAZ // Jul 20, 2018 - 2:23pm

Except for the Browns (who sucked no matter what they did), the 6 teams that used PA the least had better DVOA without it. Suggests to me that quality of play fake execution might be an answer.

Login or Register to post comments

13 Re: 2017 Play-Action Offense

by OSS-117 // Jul 22, 2018 - 9:10am

Roethlisberger just doesn't like playaction. Can't really blame him, the years of getting his head caved in the second he turned back around when Arians was OC and their garbage OL. Roethlisberger would hardly bother with executing the play fakes the past few years. And sometimes not at all on plays that were suppose to be PA. But if you compare early last season to sometime after the bye, you saw Roethlisberger making good much more of an effort on the play fakes. I don't think it will reflect in the results, but the film is night and day in the effort.

They don't run many RPOs, at least with Roethlisberger. Read-option was something they did add last season. But used little, and at least the first half of the season, Roethlisberger wouldn't bother with the mesh.

Idk. About the time Randy Fichtner came down from the booth to play middleman between BR and Haley is about when BR quit half-rumping it. Who knows how all the percentages break out this season. Gone back to buddy-OC, so probably gut everything Roethlisberger doesn't like. And he doesn't like PA. Hard to imagine it getting any lower. But when Jones/Rudolph/Dobbs eventually takes over you'll see Pittsburgh vault to the top half of the league. Read option wasn't added to the offense last year for BR's benefit.

Login or Register to post comments

14 Re: 2017 Play-Action Offense

by The Ninjalectual // Jul 22, 2018 - 11:46am

The only QB to full-rump it is of course Mark Sanchez.

Login or Register to post comments