Football

Is it football season already? It seems like I just got over the epic disappointment of the Superbowl. (Yes, I'm a Pats fan) So, in honor of football season, I think it's worth highlighting one of the major trends to affect the sport over the last few decades: the ascent of the passing game. Since 1960, quarterbacks have managed to increase their average gain per pass attempt by nearly 30 percent, from 4.6 yards to 6.5 yards. (Running backs only get about 4 yards per attempt, a number that hasn't changed in thirty years.) Furthermore, even as quarterbacks have gotten more yards per pass, they have managed to throw fewer interceptions. In 1980, passes were picked off more than 6 percent of the time. By 1995, the rate of interceptions had been halved, which meant that passing the ball wasn't any statistically riskier than rushing.

Given these statistics, it shouldn't be surprising that NFL teams are passing more than ever before. Nevertheless, they still aren't passing enough. That, at least, is the conclusion of Ben Alamar, a professor of sports management at Menlo College. He argues that the NFL exhibits a "Passing Premium Puzzle". (This is the sports version of the Equity Premium Puzzle, which is the mystery of why investors hold so many low yield bonds when stocks perform so much better over the long-term.) Alamar notes that, despite the significant increase in the expected utility of the passing game, coaches still run the ball about 46 percent of the time. While this represents a decrease from the 1960's - the average NFL team use to run the ball more than 58 percent of the time - a perfectly rational coach would almost always choose to pass, since passing represents a higher rate of return. So why do coaches still run the ball? Alamar admits that a successful running game contributes to the success of the passing game, since you want the defensive backs to have to worry about the possibility of a run. But he isn't convinced that coaches need to run the ball quite so often. "For all of their planning and late nights," Alamar writes, "NFL coaches do not act in a fully rational manner." Just like investors choosing bonds over stocks - stocks have a much higher rate of return over the long run - coaches are swayed by the illusory perception of risk, or what's often known as risk aversion. Although passing the ball isn't statistically riskier than running the ball, it feels riskier, as the ball is lofted into the air and is up for grabs. The end result is that teams gain fewer yards than they might otherwise, simply because the brain isn't good at accurately calculating risks versus rewards.

More like this

Might want to edit the 2nd and 3d sentences in the 2nd paragraph...

The behavioral paradigm here is Choice & Matching; the coaches, as you describe them, are exhibiting undermatching--responding on the richer schedule (passing) proportionally less than the reinforcement rate would predict.

Undermatching typically happens when there is little cost for switching from one schedule to the other. As you note, the running game averages 4 yards per carry; as Woody Hayes famously observed, "four yards and a cloud of dust" on each play guarantees a string of first downs, and eventual scoring. Yes, the 6.5 yards per passing play is more efficient, but as running will also work, there is no reason not to engage in undermatching.

In addition, there is no cost in time to switch between a running lineup and a passing lineup; if it required a time-out to get different players on the field, switching would be more costly and matching or even overmatching would likely take place. Quite arguably, it is not that "the brain isn't good at accurately calculating", but simply that the cost of switching is not significant, and coaches' behavior responds to the contingencies at hand.

Speaking of fans...as a fan of this blog, I was delighted to see a tall stack of Proust Was A Neuroscientist paperbacks on the front table of the JHU Barnes and Noble when I walked in the other night. Your... umm.. flurry of self-promotion, or maybe innate talent + hard work, paid off, Jonah.

What anon is missing is the counterstrategy of the defense. If the defense sees that the offense is very likely to run, the results are usually far less than 4 yards per rush.

And 4 yards per carry means a mean of 4 yards. In today's football it is +9, +2, no gain, +12, +3, +1, -1, +6. The chances of eventually getting something like +3 and no gain on first and second down are very high, then the defense can be made against the pass, giving 6.5 yards and usually forcing a punt.

With the spread formations and "West Coast Offense" it is just as likely that the passing game sets up the running game by spreading the defense.

By natural cynic (not verified) on 04 Sep 2008 #permalink

Keep in mind that teams utilize the run in order to control the *Clock*. Once your team is winning, it behooves you to run more often, as the clock will keep running in between plays, with no risk of incomplete pass (clock stops).

Coming from a university perspective, you might have noticed that USC's (and other universities) defensive strategies include players that are agile and fast, and can defend well against the passing game. As these defensive players move into the pros, the passing success rate there might fall.

I still won't invest in stocks because I make decisions for emotional reasons, and despite recent neuroscience research that ties emotions with rational decision-making, I stink at stock selections.

Jonah, I'm disappointed that someone who likes empirical research hasn't done any on the relative rates of returns on stocks versus bonds.

Over the past 10 years (end date 12/31/07 or 6/30/08, doesn't matter), U.S. Treasury bond indices have outperformed the S&P 500 Stock Index. UK gilts have also outperformed the FTSE stock index over the past 10 years.

10 years is an awfully long time to wait for the "long run" you cite in which stocks have "a much higher rate of return."

Still, we shouldn't rule out stocks outperforming over the next 10 years, so investors DIVERSIFY across asset classes. Similarly, NFL coaches DIVERSIFY their teams so they can deliver a better overall game result. Alamar (and Lehrer) should think in shades of gray, not a binary "must do 100% passing or 100% rushing" way.

Source data for performance data above is the 2007 Barclays Equity-Gilt Study, and The Journal of Portfolio Management, Winter 2003 vol. 29, Article #2. (They're not linkable, sorry.)

By Bond investor (not verified) on 05 Sep 2008 #permalink

It seems to be conventional wisdom that running the ball is more taxing on the defense than is passing the ball (because it more physical or something). Thus there may be some benefit in the fourth quarter to have run a lot earlier, especially wrt time of possession, keeping their defense on the field and their offense off the field.

There are so many underlying factors that go into this that it's not really reasonable to say that coaches are running too much. Aside from the usual "passing sets up the run, running sets up the pass" that would encourage an equal or near equal balance of each type of play, you also have clock and score considerations to think about before every play.

You see this kind of shaky analysis a lot on pre/post game shows, though in a bit of a different style. A lot of people say "when team X rushes for Y yards, they win" when in actuality it's "when team X wins, they rush for a lot of yards." Essentially, rushing in a football game runs the clock, reducing the number of plays that either team will run from then on, and making it more likely that the current score will reflect the final score. So when a team gets ahead (often by effectively passing), they then run to end the game faster and make it less likely that the other team comes back. THAT is why you see high rush totals for good teams. Just because a rush play is less likely to generate a high yardage does not mean that is necessarily a worse play.

Of course, you'll see statistical oddities all the time. Last year's Patriots are one of those oddities. For whatever reasons Bill Belichick had, he ran up the score consistently (in my opinion, bad sportsman form. of course, I hate the Pats, so make of my opinion what you will). He utilized the passing game to score points far deeper into games than almost any other team would. As such, the Pats passing stats were astronomical and their running stats were mediocre all season.

That considered, coaches aren't wrong for running. Passing sets up the run, and running sets up the win. Scoring points does not necessarily win the game, and you can get in trouble if you don't pay attention to the clock. Just ask the 1992 Oilers.

I think we have to remember there are downs. Because you only get 3-4 shots at making ten or more yards, the average of a run or pass isn't the only consideration. The variance becomes an issue, no?

Passing wears out a defense more; it forces the defensive line to rush the passer more, while the offensive line can (mostly) stand there and keep them away from the QB. The Cowboys use this strategy. They pass a lot during the first half (I think it's something like 60-70%) to wear out opposing defenses (the Cowboys have a massive O line) and grab a good lead, then the second half features a lot of more successful running to preserve the lead and run the clock out faster.

I'm typically skeptical of these sorts of analyses of the game, since they usually sound like they treat each play as a discrete event unconnected to the rest of the game.

Btw, if you're looking for serious statistical analysis of football by people who really understand the game, check out footballoutsiders.com.

Just a pedantic point, but it's not really risk aversion that's being exhibited. It's true that there is an effort on the part of coaches to avoid perceived risk, but if what you mean by 'risk aversion' is what you discuss in the sentences following its mention, then it's not what economists mean by the term. Someone is risk averse if they behave in the following way. Suppose you have two choices:

A a 10% chance at a reward worth 100 points (of whatever) and a 90% cahnce of nothing,

or

B a 100% chance at a reward woth 10 points.

In each case, the expected value of the reward is 10 points. Someone who is risk neutral will be indifferent, but someone who is risk averse will choose B.

The normal explanation for risk aversion (although this isn't uncontroversial) is that risk averse individuals assign a diminishing marginal value to the reward item. In football, this may be a rational behaviour, or it may not be. Because a football field is a fixed length and one needs only 10 yards to get a first down, it wouldn't be surprising if there's a somewhat odd valuation given to yardages of different lengths.

Yep, if you could guarantee 4 yards per rush, you'd rush every down, because you'd get a first down every time.

Obviously this isn't the case, but I certainly think it's likely that rushing yards are far more "reliable" in terms of distribution... with passing you may have a higher average, but you end up with a lot of zeros as well.

And on 3rd down, the opposing team will gladly allow you a short completion... perhaps falsely inflating the average somewhat since those are essentially "throwaway" yards that the D is allowing.

Many other valid considerations also noted above (clock management, situation of game dictating playcalling, etc.)... the analysis seems way oversimplified here, and certainly doesn't support the end assertion (that teams run more than they pass simply because the brain isn't good at accurately calculating risks versus rewards). Interesting discussion, unfounded conclusion.

An interception is not the only potential risk from passing the ball. Passing plays are far more likely to result in zero or negative yardage through an incomplete pass, sack or holding penalty than running the ball which rarely results in a loss or penalty.

Regarding wearing down the defense, while an individual pass play may require more energy from the defense (I can't confirm nor refute this) a team that can successfully run the ball can wear down the defense by requiring more plays. A twelve play drive is more tiring for a defense than an eight play drive of the same distance.

And I'm sure, by the professor's calculations, that pitchers should only throw strikes, since throwing a ball couldn't possibly have anything to do with making an out.

Like baseball, football is a game of misdirection. Fooling a defense by mixing run and pass is the essence of its strategy. Two runs for no yards may be the precise reason why the third call, a play-action pass, goes for big yards.

It sounds like the professor learned about football from a book translated from the original Greek.

Brilliant! A university professor analyzing football's passing and rushing games in vacuum

I'm a statitician and human performance psychologist (Ph.D. 1964). The serious mistake made here is in neglecting the issue of personnel. Some teams have quarterbacks who throw the ball well and receivers who catch well and run well after making the catch. Other teams have great running backs and offensive linemen who create holes for them to sprint through. Rarely does a team have it all.
A coach's job is to make the best use of the personnel available to him. If I were coaching a team today that had O. J. Simpson and Jim Brown as running backs in their prime, and Bart Starr at quarterback, Starr would be handing the ball off to one or another of his backs 80% of the time, throwing only occasionally, and then short sideline passes. Sort of the way Lombardi did it 40+ years ago, except that he had Hornung and Taylor as his running backs.
On the other hand, with Montana throwing and Rice catching...not too much running.

By Nat Ehrlich (not verified) on 07 Sep 2008 #permalink

If you are averaging the gains from passes and runs among a historical grouping, I suppose you could be left with the conclusion that passing is indeed better and that teams should pass more.

Still, a defense centered on the fact that the offense is going to favor passing will be in a better position to defend against the pass. If the offense sets the expectations of a passing game just right, they might get in one hell of a running game.

I have always cringed when hear about "keeping the other team's offense off the field" with ball control (usually meaning running the ball). Statistically, if you keep the ball longer per possession, that just means less opportunities to score for both teams. Maybe you get the other team's high-powered passing game out of synch, maybe, but otherwise no real advantage. Discuss.

Jim,

Like I said above, running a lot ends the game faster. If you have a lead, you run to end the game ("control the clock"). If you pass and fail, you give the other team more opportunities to catch up. Winning a football game is not necessarily just about scoring as many points as you can. There are relatively few Browns-Bengals barnburners and Patriots score running up humiliations in the league.

ESPN did an article about this and they added in another element in the question why coaches do not go for the two point conversion more: peer pressure. There are generally held beliefs in all professions. In medicine it is called "standard of care." If an owner had a head coach who goes for two every time, he will lose games that he might have won (going for one) and will be ridiculed by other owners; ditto the GM. "Is your guy nuts? Nobody goes for two on every touchdown!" Coaches are not automatons and they will tend to go toward the mean. The mean is "no two pointers" and, likewise, "run to set up the pass."

One last pro-passing argument. Barry Sanders was a fun runner to watch, but not a "great" runner because of one thing: negative rushes. If you give up three sacks in a game, that is alot. But negative rushes are more prevalent and keep the per rush average down.

Yeah, if you want to talk about actual risk aversion in play-calling, the decision to punt far too often is a better one. There's been some pretty convincing analysis that suggests teams should almost never punt. You'll see a variation of this idea with some of the better offenses once they get inside the opponent's side of the field. When it's too far for a field goal, but too close for a good punt, they'll often go for the fourth down conversion. Invariably, the commentators talk about how risky it is.

Another thing the original discussion seems to neglect is the 4 downs system. It doesn't matter what you average per pass if the series doesn't get you a first down. If, as others point out, you can run for 4-5 yards reliably, then you'll just do that until the defense can stop it.

You are ignoring some significant differences between passing and running. Yes, passing is better at getting more yards quicker, but running gives you two distinct advantages:

1) Running the ball uses more time, as an incomplete pass stops the clock. A passing play also is more likely to end out of bounds, also stopping the clock. If you are ahead in the game, you want to run the ball more to chew up the clock, giving your opponent fewer opportunities to catch up and forcing them to pass, thus cutting their offensive options down and making defense easier.

2) Running the ball wears down defenses over the course of the game. This also bunches them up closer to the line, making it easier to make big passing plans later in the game. They call it "grinding it out" because it weakens the defense over the course of the game much more than passing does.

So, their are metastrategies in play when choosing to run over pass that trump the pure "yards per play" formula.

Believe it or not, football is an extremely complex game and this debate just demonstrates the nuances of it.

First of all, changing the ratio from 58-42 to 46-54 is a big change (even over 40 years). Watch a team that runs 58% of the time and another 46% and you will be watching two very different approaches.

Second, the average gain is not the only thing to consider. The median is key too. When you are third and two, or fourth and inches, there is a much higher rate of "failure" to a pass, defining failure as not getting the first down.

The high average gain for passes, is weighted heavily by passes that gain 20, 30 or 60 yards at a time. And the possibility of high gain passes weights the averages too. You get the ball on your own 30. You run three straight times for a first down and then throw a 50+ yard for a touchdown. By your measure, you are running the ball 75% of the time, but that only tells part of the story.

You need the jab and the haymaker. Finding the exact right mix is a delicate science.

By Seth Berkowitz (not verified) on 07 Sep 2008 #permalink

Wouldnt passing also wear out the offense more if the WRs and TEs end having to run routes constantly?

FELLAS, haven't seen this in any of the L.A. papers, but here's an interesting piece that ran on the Florida Times-Union website (http://news.jacksonville.com/pieceofdiroc/2008/09/07/after-the-storm/). It refers to Florida's defeat of Miami Saturday night:
RB Emmanuel Moody did not play because of a coachs decision. He suffered an ankle injury last week, but Urban Meyer said Moody is not injured. He also said Saturday night that he must get Moody more work in the game plan. There are rumblings that Moodys attitude isnt the greatest and that hes not exactly happy about sharing carries with Jeff Demps, Chris Rainey, Kestahn Moore and Brandon James. If thats the case, then he has no one to blame but himself. UF doesnt have a feature back and never will under Meyer, and he should have known that when he was trying to decide where he was going to end up after leaving Southern California

By arealrascal (not verified) on 07 Sep 2008 #permalink

Alex,

That's not what i am referring to. Running down the clock if you happen to get a lead is one thing, but I mean as a game plan going in, such as to keep a Peyton Manning "off the field."

The play action fake has to be mentioned specifically here. The run sets up the pass in general, but the run REALLY sets up the play action fake. This is the play where the quarterback pretends to hand off the ball, holds the ball on his hip, waits for the RB to hit the line, and then quickly dumps the pass off to a receiver.

Teams that are good at the play action fake require a few things:

1) A QB who is good at hiding the ball and acting like he handed it off (obviously). This often involves keeping one's back to the defense for a second or two when you can legally be crushed (brave).

2) A RB that will run very hard, and can also hide the fact that he doesn't have the ball.

3) WRs/TEs that consistently run routes at full speed even on running players. Receivers that slack off on running players will give away the fact that its a play-action-fake when suddenly they run hard on a running play.

When you have these three things, the play-action-fake can be devastating.

-Michael Hartman
Blogging about Online Gaming and Virtual Worlds:
http://www.muckbeast.com

Excellent discussion!

Having a bit of fun with "the Daily Transcript"'s post on the new search engine "pubget", I took a look at "matching" and "football", and it seems that I have been unaware of a study showing exactly what my previous point (Anon comment #1) was.

Comment #3 notwithstanding, the counterstrategy is already accounted for in the mean yardage per play, so that particular objection is null. The difference due to variability turns out (see below) to be a significant factor, so good call, natural cynic. Anyway, I wanted to point out Reed, Critchfield, & Martens, (2006), JABA, "The Generalized Matching Law In Elite Sport Competition: Football Play Calling As Operant Choice".

Abstract:

A mathematical model of operant choice, the generalized matching law was used to analyze play-calling data from the 2004 National Football League season. In all analyses, the relative ratio of passing to rushing plays was examined as a function of the relative ratio of reinforcement, defined as yards gained, from passing versus rushing. Different analyses focused on season-aggregate data for the league as a whole, game-by-game data for the league as a whole, and game-by-game data for individual teams. In all analyses except those for a few individual teams, the generalized matching law accounted for a majority of variance in play calling. The typical play-calling pattern reflected undermatching (suggesting imperfect sensitivity of play calling to yardage- gained reinforcers) and a bias for calling rushing plays. Bias was found to be a function of both the relative risk of turnovers and the relative variability in yards gained associated with passing versus rushing plays. The external validity of the matching analyses was supported by significant correlations between parameters of the generalized matching law and team success on offense and season winning percentage. These results illustrate the broad applicability of the generalized matching law to problems outside of the laboratory.