Which Teams Should Run More?

If I read one more of these articles in the Baltimore Sun, I think I'm going to throw up.

...Ray Rice consistently ranks as one of the most productive and dynamic running backs in the NFL, capable of eluding defenders on the ground and through the air. He benefits from running behind bruising All-Pro fullback Vonta Leach, who's regarded as the most devastating lead blocker in the game. Yet, the Ravens have fallen to 19th overall in the NFL in rushing yards per game with an average of 104 yards per contest. And Rice ranks 12th on the rushing chart individually with 524 rushing yards...

The article goes on to add:
Blah blah blah. Run more blah. Need to find their identity blah blah.

Ok, I made that second part up. Actually, I have to give the author credit for attempting at least a surface level statistical analysis. But the fetish that sports columnists have with the running game has been one of the most enduring false narratives in the sport of football. It's time to put it to rest. I get that offenses have to run to keep defenses honest and to set up the passing game, but they don't have to run as often as they generally do to accomplish those things.

Let's take a look at offensive Expected Points Added per Play by play type. EPA/P accounts for all play outcomes--sacks, interceptions, fumbles, gains, losses, first down conversions, and so on--in other words, everything. Naturally, an offense would want to do more of the kinds of plays that gain them more and do less of the kinds of plays that gain less. Although this should be intuitively understood, zero-sum game theory proves that the optimum total production is when payoffs are equalized across strategy options.

Teams with high differences between pass payoffs and run payoffs should probably be running more often. And although you might think that teams with higher run payoffs than pass payoffs should be running more, that may not be true due to the passing paradox: Underdog teams that are poor at passing may need to do it more often to generate high variance outcomes.

The table below lists each offense's passing and running EPA per play, their total EPA per play, the difference between pass and run EPA/P, and their proportion of pass plays. I limited the analysis to 'normal football' where the score is relatively close and time is not yet a factor in play selection. The teams are ranked by the difference between pass and run payoffs. Teams at the top of the list should be passing more, and the teams at the (very) bottom should be running more. (Click on the table headers to sort.)


OffensePass EPA/PRun EPA/PTotalDiffPass %
OAK0.22-0.240.030.4660%
PIT0.30-0.120.150.4263%
SD0.22-0.180.060.4059%
NO0.31-0.080.190.3968%
ATL0.31-0.060.190.3766%
NYG0.23-0.070.110.3060%
HST0.330.060.200.2752%
DEN0.17-0.060.080.2361%
PHI0.05-0.17-0.040.2357%
BUF0.240.020.130.2353%
WAS0.310.110.210.2047%
GB0.200.040.150.1667%
NE0.250.100.180.1554%
DET0.09-0.070.020.1560%
TB0.180.030.120.1556%
CAR0.210.070.140.1454%
CIN0.12-0.010.070.1359%
MIN0.130.010.070.1252%
NYJ0.10-0.010.050.1156%
TEN0.120.020.080.1061%
JAX0.01-0.08-0.030.0956%
IND0.130.060.100.0856%
ARZ-0.04-0.11-0.060.0762%
CLV0.02-0.040.000.0660%
MIA0.04-0.010.020.0555%
DAL-0.01-0.04-0.020.0361%
SEA-0.02-0.02-0.020.0147%
BLT0.060.070.07-0.0164%
KC-0.11-0.06-0.09-0.0544%
SF0.110.220.15-0.1156%
SL-0.020.110.03-0.1359%
CHI-0.060.120.01-0.1959%
NFL0.11-0.010.060.1158%

Notice that Baltimore is actually doing it right. They're as balanced as could reasonably be expected between run and pass payoffs with a difference of only 0.01 per play. Just because most other teams are doing it wrong, it doesn't mean the Ravens should follow suit. Seattle is another good example. Even though they are on the other end of the spectrum in terms of run-pass proportion, they are just as balanced in terms of relative payoffs.

You could make the case that SF, SL, and CHI should be running more often than they have, particularly SF and CHI thanks to their tough defenses. Otherwise, there aren't any other teams that you can say should be running more often, and many that should be passing more.

The simple lesson here is that when thinking about run-pass balance, you need to consider both payoff and proportion, not just proportion.

Note: You can always see an updated graphical plot of offensive (and defensive) EPA/P for passes and runs here.

  • Spread The Love
  • Digg This Post
  • Tweet This Post
  • Stumble This Post
  • Submit This Post To Delicious
  • Submit This Post To Reddit
  • Submit This Post To Mixx

25 Responses to “Which Teams Should Run More?”

  1. Jason says:

    I didn't read the full article you quote but the usual argument is that running more also benefits the passing. I know you are pointing out that by switching from passing to running you are switching from Passing EPA/P to Running EPA/P (which for Baltimore is almost identical) but couldn't a shift in proportion also raise both EPAs? In that case the switch to running would be wise.

  2. Anonymous says:

    Brian, Do you think you could post the to date play by play data?
    Thanks

  3. Anonymous says:

    Great post, thanks!

  4. Anonymous says:

    In a Nash Equilibrium (a.k.a. game theoretically optimal) strategy, the expected payoff for each of *the opponent's* viable options is equal.

    From a game theory point of view, what the table is (ostensibly) telling us is that the defenses Chicago, Saint Luis, and San Francisco have faced have - in aggregate - tried to defend against the pass too much. While Oakland, San Diego, and Saint Louis' opponents were too worried about the run.

    A statistical test of the run-pass balance would have columns for the payoffs against defensive approaches they faced. Short of game footage, I haven't seen data that actually includes that information, so producing that sort of chart could be a lot of work.

    In plainer English:

    When the payoffs for running and passing are the same - as they've been for Baltimore's - that tells us that the opposing defense, not the offense, is 'well-balanced'.

    N.B.: It's worth pointing out that you want to exclude 'silly' choices when equalizing the opponent's payoffs. Nobody in the NFL is going to worry about a trick punt on first and 10.

  5. Josh says:

    Could you run a regression to figure out what % of overall success is tied to your passing success? It seems like that's pretty much all that matters.

    Also, what would you honestly suggest New Orleans do? They pass more than anybody in the league, and their difference between passing and rushing is still fourth highest in the league!

  6. Jerome Manson says:

    Never really looked at it this way before. Seems like some insight into what teams look into and break down.

  7. Joseph says:

    @Josh:
    What the Saints need to do is get a defense. As a 30 yr Saints fan, they have been passing a lot to try to score enough to "bail out" the defense, or to try to keep up with the other team's scoring. On the other hand, with their running plays, they need to have better selection with which RB they use (FREE PIERRE THOMAS!), which play they use, which formation they use, and which down & distance they run in. They truly are a team that needs to run a little bit more (say 65% pass, 35% run, vs. 68-32 now) to open up their passing game. Because while their passing game has been good, it has been GREAT for the past 3 yrs, and now it's not.
    In their SB year, they were very good at running, no matter what the situation. They salted away many games in the 4th Q because they were picking up 4-5 yds in situations where the D was EXPECTING the run. This year, they have been too predictable--and that is where, IMO, they miss Coach Payton.

  8. Adam H says:

    "I get that offenses have to run to keep defenses honest and to set up the running [passing?] game, but they don't have to run as often as they generally do to accomplish those things."
    Offenses also have to PASS to keep the defenses honest, and I would think these two considerations would roughly cancel each other out.

  9. Brian Burke says:

    to Anon above re Nash--

    You wrote: "When the payoffs for running and passing are the same - as they've been for Baltimore's - that tells us that the opposing defense, not the offense, is 'well-balanced'."

    Not quite. The NE tells us that *both* the offense and defense strategy proportions are optimal. But since the offense's choices are distinct while the defense has a continuous spectrum of strategy dispositions, I'm focusing on the offenses choices here.

    Think of it this way--the offense's payoffs are simply the inverse of the defense's payoffs. So in your own explanation, the defense's opposition (the offense) is balanced.

    (Sorry. I realize I'm not explaining this well.

  10. Anonymous says:

    > Think of it this way--the offense's payoffs are simply the
    > inverse of the defense's payoffs. So in your own
    > explanation, the defense's opposition (the offense) is
    > balanced.

    Assuming the defense isn't being stupid, we can say that the EPA/Play (that is to say the payoff) is close to the Nash Equilibrium EPA/Play. That doesn't mean that we the run/pass proportion is close to the NE strategy.

    Let's say we ignored an average half of Baltimore's passing attempts. Then the run/pass payoffs would still show as 0.07/0.07, but the pass percentage would be closer to 50% than 64%. Is 50% also an optimal passing percentage for Baltimore?

  11. Anonymous says:

    The hypothesis that if pass epa/p should equal run epa/p not only has not been proven, but it seems from that table to be incorrect.

    The teams that do this most closely, Baltimore, Dallas, Seattle, Miami, and KC have a record of 17-19.

    The most successful teams, chi, atlanta, houston, are way off from matching their run/pass epas.

    thus, the conclusion is suspect.

  12. Brian Burke says:

    That's a misunderstanding of the concept, and a very important point. I'm not claiming teams with balanced payoffs are the best offenses. I'm claiming they are 'optimized'--in other words, they are doing the best they can with what they have. (KC is never going to be a very good offense no matter how balanced their payoffs are, but they might as well be as good as they can be.)

    Not to mention, those teams you cite all have excellent defenses.

    (By the way, sort the list in order of pass %. You'll find NO, GB, ATL, PIT...many of the very top *offenses*.)

    Look at it in terms of the marginal play. If team X expects to average .2 EPA per pass and .1 EPA per run, would it improve or reduce its performance by switching 1 run play to a pass play?

    Ok. Now how about switching 2 run plays to pass plays?

    And so on...

    When should team X stop swapping passes for runs? When defenses adjust to the point that their respective expected payoffs equalize.

  13. Brendan Scolari says:

    Hey Brian, anything encouraging to read into the fact that offenses are passing 58% of the time this year?

    I remember your your preseason article about how passing is becoming more and more attractive in the modern NFL, andbut teams have been stuck at passing 56% of the time for a number of years.

    So there's seemingly been a little progress in regards to optimal playcalling, although I'm not sure if pass% drops as the weather gets colder which would make the current percentages deceptive.

  14. James Dollinger-McElligott says:

    While I think this analysis a good start I think one potential flaw is that it only focuses on aggregate numbers and ignores situational factors. While I'm only familiar w/ the buffalo bills situation it would not shock me if it generalized to other teams. If you look at the buffalo bills they have 8th greatest differential in pass epa / play vs run epa per play but the 6th lowest pass percentage. The naive conclusion one would draw is that they should throw a lot more often. However it seems to me that a large %age of the bills epa/pass play has to do w/ the high percentage of throws that result in touchdowns (second highest in the league behind GB). Furthermore 9 of these 15 touchdown passes have been 20 yards or less suggesting that a large %age of the pass epa comes from throwing the ball when the Bills get close to the goal line. When you look at total yards per play buffalo has the lowest differential in the league gaining only 1.3 yards per pass play more than per run play. This suggests to me that the optimal strategy for buffalo is one that has a more balanced run/pass ratio until the Bills get close to the goal line in which case throwing much more often to take advantage of the teams ability to get touchdowns by throwing the ball which I think pretty closely mirrors the play calling this year.

  15. Dave says:

    I can't believe Miami is that far down the list. Their running game has been terrible lately. I wonder if it's a function of their early success with it, and possibly also the early struggles with passing.

  16. Anonymous says:

    Teams could probably stand to pass a bit more but I think multiple things are at work and that we are pretty close to the optimum mix.

    1. The more times you pass the more chance you have of your QB getting hurt. Likely for multiple games. If you have a star QB this is a significant loss for your team.

    2. Playing into number 1 teams with large leads should be content to waste clock by running the ball more even if it means less epa/play.

    3. Brian you posted this awhile back

    http://www.advancednflstats.com/2010/09/deep-vs-short-passes.html

    "All things considered, short pass attempts yield a net negative average EPA. Run plays average a better EPA than short attempts. It's not simply that passing has generally superior payoffs than running. It's that the entire advantage of modern passing comes from deep passes. Teams should probably be passing deep more often, and running and passing short less often"

    Where the majority of the passing advantage comes from deep passing. You can't just dial up a completed long pass every play. Your line has to be able to hold up, receivers get open, etc. Easier said than done.

  17. Jordan Dietrich says:

    Every time I read something like the cited article, I typically hear "Well, this team is N and 0 when they run the ball more than X times, so they need to get it to their running backs early and often".

    That line of thinking typically makes me twitch and want to write the commentator a letter explaining the difference between correlation and causation, and that maybe, JUST MAYBE, they aren't winning because they run so much, but they run so much because they are winning.

  18. tmk says:

    Thanks. Ive been reading your "anti-running" manifestos over the year and this post best explains previous analysis.

    I agree that the game has shifted and teams need to adapt. However if you can beat Indy buy rushing for 225 yds in 2010 why not?

    I think one question you have to ask is when do you abandon your game plan?

    Best laid plans rarely survive first contact with the enemy, which reinforces the importance of momentum, however early execution doesn't ensure a win.

    I think coaching staffs have to have faith in the game plan, even when players don't execute properly, otherwise why would they get a contract instead of a pink slip.

    I would be interested to see the s.r. of teams that "change it up" mid game, and whether it resulted in a win

  19. Brian Burke says:

    To clear up a few misconceptions:

    -my analysis was limited to what I call 'normal' football, score within 10, 1st-3rd quarters minus the last 2 min of the 1st half and any drives that ended in expiration.

    -QB injury is a valid consideration, but is very hard to quantify. They pass a lot more in the CFL and the QBs survive.

    -It's true that deeper passes are what gives passing its advantage, and they are difficult to accomplish. But they are not accomplished 100% of the time when you don't attempt them.

  20. Anonymous says:

    "That's a misunderstanding of the concept, and a very important point. I'm not claiming teams with balanced payoffs are the best offenses. I'm claiming they are 'optimized'--in other words, they are doing the best they can with what they have. (KC is never going to be a very good offense no matter how balanced their payoffs are, but they might as well be as good as they can be.)"


    but in no way is that conclusion justified. As I pointed out, if you look at the results, you show that the best offenses are "not optimized" and that lousy offenses are "optimized". That is a strongly counterintuitive result.


    "Look at it in terms of the marginal play. If team X expects to average .2 EPA per pass and .1 EPA per run, would it improve or reduce its performance by switching 1 run play to a pass play?"

    Again, a hypothesis that has not been proven. If every single added pass play increases total EPA, why not pass 100% of the time? Because, total EPA does always increase with an added pass, obviously.

    You try to even out Pass EPA and Run EPA. What you want to do instead is to maximize total EPA. While simple game theory may suggest that you equalize the two play types, there is no reason to assume that it applies to the NFL.

    have you verified how Pass/Run EPA changes with the independent variable being % of pass plays executed?


    Also, I'm a big fan of the site. Great work. Lots of interesting stuff to read.

  21. Anonymous says:

    Brian,
    Your work is very interesting and I enjoy reading it. I know this isn't completely related to this article, but I'm curious about this: have you written an article/ explored the possibilities that better rushing teams perform better in the red zone, and by effect score more than their pass heavy counterparts that get into the red zone and are down around the goal line and can't punch in a run? And are some defenses, like the Miami Dolphins of this year come to mind, who have a great run defense, but a mediocre pass defense but play a sort of bend dont break style and dont allow many first downs, better in the red zone and ultimately let up fewer points because of their ability to stop the short pass? Is red zone rushing and defense a "skill"? I know it isn't for QBs, or it lies below the level of detection, telling by your article. Moreover, is it just luck that the Dolphins are poor agianst total yardage passing but don't allow many first downs. Im not sure if youve explored this or written an article about, or if it even makes any sense. Just let me know.

  22. Anonymous says:

    At other Anon:

    "As I pointed out, if you look at the results, you show that the best offenses are "not optimized" and that lousy offenses are "optimized". That is a strongly counterintuitive result."

    It's not that counterintuitive. Most teams pass between 55 and 65 percent of the time, and the best offenses, for the most part, are the ones which pass the ball well. So teams who pass the ball well perform well on offense, but could perform even better if they passed the ball more. Whereas teams who are struggling to pass the ball wouldn't benefit as much from passing the ball more.

    Coaches don't think in terms of statistical optimization; most coaches' have a certain base run/pass percentage that they like and make slight adjustments based on personnel. Therefore, many teams have an "optimized" offense as a direct result of the fact that they aren't very good; the generally accepted 55-65 pass percentage in the NFL happens to be optimal for not-so-good teams, but not optimal for good offenses who pass well.

  23. Andy says:

    Think of the distinction like this - what if all plays going to the left side of the ball gained 0.25 EPA/play and all plays going to the right side of the ball gained -0.25 EPA/play. If you were the coach would you call plays running right or left?

    Would you "set up" left side by running a few plays to the right first?

    You might say that passing and running are more fundamentally different - for starters passing is more risky than running, where left and right are essentially the same. But I would say they aren't different - all the uncertainty and riskyness is already contained in that EPA/play number. It knows about interceptions, sacks, penalties, incompletions. We need to trust the number.

    Also, nobody is saying to pass 100% of the time. In fact, we can't even say what percentage teams should be passing at. Except for this: MORE. At least for 27 teams,more.

  24. Upadhyay Vishal says:

    Your blog is very well-informed. . You have plant a very significant concern. I liked it. Thanks for sharing with me.

    www.alltimeplay.com online soccer manager, management, free, soccer game, fantasy football

  25. Upadhyay Vishal says:

    Excellent post, one of the few articles I’ve read today that said something unique

    www.alltimeplay.com online soccer manager, management, free, soccer game, fantasy football

Leave a Reply

Note: Only a member of this blog may post a comment.