## Why Do Teams Run The Ball, Part II

Seven and a half years ago, I asked the question, why do teams run the ball? Today, I want to revisit that post, given that a lot has happened over the last seven and a half years.

Let’s begin my analyzing league-wide pass and rush efficiency. To measure rush efficiency, we will use Adjusted Yards per Carry, which is calculated as follows:

(Rush Yards + 11 * Rush TDs + 9 * Rush First Downs) / (Rushes)

For passing, we will use a modified version of ANY/A by also giving credit for first downs. Here’s the formula:

(Gross Pass Yards – SackYardsLost + 11 * Pass TDs + 9 * Pass First Downs – 45 * INTs) / (Pass Attempts + Sacks)

As a reminder, because touchdowns are recorded as first downs, this formula gives us our familiar 20-yard bonus per touchdown.  Ideally, we would incorporate fumble data into this as well, but unfortunately, we can’t determine whether fumbles happened on rushing plays or pass plays when looking at individual or team game logs. 1 As a result, they are ignored in this analysis, but otherwise, this is a pretty good way to encompass the risks and rewards of passing and rushing.2

Now, let’s take a look at the league-wide pass efficiency and rush efficiency rates since 1950.  In black, I have plotted the adjusted yardage gain (using the formula above) for the average NFL pass, while in red, we have the same for each NFL run.  In purple, I have plotted the average AFL pass gain, and used orange for the average AFL rushing gain.  As you can see, there was a time, as recently as the late ’70s, where rushing was more efficient than passing.  But that time has long gone.

To the extent you hear coaches say that running and stopping the run are the keys to victory, know that they are not crazy, just outdated.3 Rushing used to be the most efficient way to move the ball, along with the most consistent. While a higher variance could be the reason teams chose to run instead of pass, that argument holds much less water when passing reaches the efficiency levels present in the modern game.

Which gets us to the next question: how often are teams deciding to pass rather than run? In black, I have plotted the average pass ratio (that is, pass attempts plus sacks divided by total plays) in the NFL for each year since 1950, while in purple, we have the same information for each AFL season. After the 1978 rules changes, the pass ratio in the NFL spiked, and it has been consistently increasing for the last 30 years.

Now, here’s the big question: are teams passing more because they are better at passing? That would make sense, but what does the data imply? I’ve taken the information from the first two graphs to create one last chart. As in the above chart, in black, I have plotted the average pass ratio in the NFL for each year since 1950. This is identical to what you just saw. But now, in red, and against the right Y-Axis, I have plotted the difference between the pass efficiency and rush efficiency averages in the NFL for each year.

That’s remarkable, at least to me. The frequency with which the league passes is nearly directly tied to the delta between pass and rush efficiency. While that makes perfect sense, seeing it in graph form is pretty striking. There are some outliers, of course. In the early ’50s, teams were passing way too much, which perhaps explains why the league seemed to suddenly stop passing in 1956. And from ’99 to ’02, teams may have been passing too much then, too, perhaps unaware that rushing efficiency had increased significantly over the previous decade.

But right now, the two lines are pretty well aligned, which leads to an interesting conclusion. If we re-run these numbers in 7.5 years, and pass efficiency continues to rise, that should mean the historically high pass ratio we currently see should only increase. How long until we finally reach a league-wide 60% passing rate?

1. Although it does appear that fumbles happen slightly more often on passing plays. []
2. Also ignored: penalties, but that’s something I’d like to address in a future post. []
3. Well, I suppose the two ideas aren’t mutually exclusive. []

That last chart is remarkable to me, as well. Does this imply that coaches are highly rational in their play selection, or that they should be passing even more, given that the delta between passing and rushing is at an all-time high?

• Kibbles

Braess’ paradox suggests that the goal is not simply to call the most efficient play every time, but to call plays in a way that ensures maximum efficiency over the entire sample. In other words, it’s not as simple as “Is passing more efficient than running? If so, teams should pass more!” Instead, it’s “Would the one-time gains in efficiency from adding another pass play outweigh the small marginal losses in efficiency across all other pass plays we call? If so, teams should pass more!”

The fact that frequency is increasing in near-lockstep with efficiency suggests that a lot of this sort of optimizing is going on behind the scenes, (whether coaches are doing it consciously or not), which is the big takeaway I’m left with.

Here’s a great paper on pass vs. run ratios and Braess’ paradox: http://www.sph.umn.edu/faculty1/wp-content/uploads/2012/11/rr2012-004.pdf

• Tom

Kibbles –

That article was great, thanks for sharing it. There’s no doubt the rules are helping the passing game, and that QB’s and WR’s are getting more precise, that’s plain to see, but I completely agree with your last statement that some kind of optimizing is taking place, and it appears that coaches are getting better at it.

Excellent read. I guess it’s hard for me to believe the NFL coaches are actually making optimal decisions, but at least in this specific area, it seems like they are.

• sacramento gold miners

Offensive balance, or at least some production from the run game, still has a role in today’s game. If Seattle doesn’t outsmart themselves in February, we likely see more coverage of their offensive approach. And Ben Roethlisberger cemented his place in Canton with help from Levion Bell’s breakout season in 2014.

The running game remains useful in shortening games, bad weather, and helping offenses without strong QB play. An effective running game also helps reduce hits on the QB, Denver figures to run more in 2015 to protect Manning.

• sn0mm1s

Is that 2nd formula supposed to be passing 1st downs?

• Andrew Koche

Hi, Chase. Thanks for the analysis, and it was great to read part I as well.

However, I do think you are overlooking an important strategic component to the decision to run vs. pass. Simply, running the ball effectively opens up the passing game and vice versa. If the defense is going to use nickel personnel all game, you are probably going to see decreased pass efficiency and increased run efficiency, and if they’re going to stack the box, the opposite applies. By having a balance and unpredictability, you may increase your overall efficiency. I know you’re not rejecting this advantage of balance, but it should probably be part of the discussion.

Then there’s the impact of the clock. Even if the rushes aren’t kneels and you are up, you could be making the right decision by doing a low-efficiency run if it will kill the clock. Passing also has a similar niche for saving the clock. I’d be interested to see some analysis of run/pass efficiency as a function of time and score.

• James

“If the defense is going to use nickel personnel all game, you are probably going to see decreased pass efficiency and increased run efficiency,”

Are you sure about that? Put a fullback on the field, and the defense puts a linebacker out there. Put a WR, and the defense plays a DB. One is stacking the box, the other is nickel, but in the end you have the same number and equally matched blockers and defenders on the field.

I don’t necessarily see why a FB blocking a LB would lead to better rushing results than a WR blocking a DB. Now maybe going 3 wide changes the specific personnel (i.e. a run stuffing linebacker comes out for a coverage LB) but in general I don’t see formation having a significant impact. If anything, the one back spread offense was designed in large part to open up the passing game by spreading out the defenders and giving the RB more space to work.

• tonka

Mismatches could force a change in personnel grouping which could possibly strengthen the chances of success for the pass or run.

In the 2013 Patriots vs Saints game. Coach Belichick used cornerback Aqib Talib to defend against Jimmy Graham who had been destroying defenses prior to that game. If that strategy was used against Rob Gronkowski who is a far more capable blocker, the Patriots would have an advantage when it came to the running game due to having more big bodied blockers then run stopping defenders in the box.

During the 2015 Patriots vs Saints preseason game, the Patriots were seen utilizing a split back formation which took advantage of their abundance in pass catching running backs to effectively attack different defensive personnel groupings. If a linebacker was used to cover someone like running back Travaris Cadet (or James White), the Patriots would be able to motion him out onto the line of scrimmage to serve as a wide receiver, a role he had served during his time with the saints, given his familiarity with running routes as well as his 6’1 frame he would create a mismatch versus most linebackers in coverage. If the defense decided to use a defensive back then the Patriots would be capable of running it due to the the smaller box.

Paraphrased it more or less from Rich Hill’s analysis article. http://www.patspulpit.com/2015/8/26/9197881/film-review-two-running-back-sets-for-james-white-and-dion-lewis. Substituted White for Cadet since I REALLY want him to stay since I think this formation works better with him since he creates a bigger mismatch as RB2.

• Duff Soviet Union

The interesting thing is that while running was almost as efficient as passing on a league wide level up until about 2003, it has always been passing that wins more than running and that’s due to variance. Passing done well varies from average a heap more than running done well. And the worst teams are the worst passing teams because passing done badly is worse than running done badly.

People act like the running game being less important than passing is some sort of new fangled trend, but it’s pretty much always been that way.

• Kibbles

Hmm. It seems that NFL coaches, as a class, might actually be pretty smart after all.

• Richie

I know the NFL was going through a lot of changes in the 40s and early 50s. Were there any rule changes in the late 40s that encouraged teams to pass more often in the early 50s, apparently to their detriment?

• In 1947, the NFL made rules changes to penalize defenders for using their hands to block a receiver’s vision.

The 1950 merger had what I call the Paul Brown Effect, in which established teams realized they needed to try to emulate Brown lest they fall into obscurity.

In 1952-53, the NFL made changes to illegal motion penalties in order to give offenses more pre-snap freedom, and in 1954 they made a rule allowing the offense to request a clean and dry ball at an time during a wet game.

Oddly enough, passing percentage took a nosedive in 1956, the year the NFL got rid of the slippery (hard to throw and catch) white ball in favor of the brown leather ball.

• Richie

I was curious who the most pass-happy teams of all time were. I queried teams with 500+ pass attempts and less than 350 rush attempts. 35 teams qualified. On average, these teams went 5-10. Only 4 of them had a winning record (1991 Oilers, 1990 Oilers, 2008 Cardinals and 1995 Falcons). The most pass-happy was the 2013 Falcons who attempted passes (or were sacked) on 69% of their plays. They finished 4-12.

This is an area where game scripts would probably be more interesting. I assume most of these teams passed a lot because they were trailing often. Good teams that pass a lot still end up running the ball late in the game to run out the clock, thereby ruining their pass-happy ratio.

• Duff Soviet Union

You’re right. Chase did actually look at this using game scripts and his list of the pass happiest teams were pretty much all really good. Number 1 was the 1995 49ers the year Rice set the receiving record. The top 10 had 4 Warren Moon Oilers teams, the 2007 Pats, the 2001 Rams, 2011 Packers, an Air Coryell Chargers teams and the 2004 Eagles. No super bowl winners but a whole bunch of excellent teams.

Meanwhile the bottom of the list was filled with bad quarterbacks (2000 Bengals, 1982 Patriots, 2007 Raiders, 2001 Cowboys, 1982 and 1983 Colts, 2012 Chiefs) and, at the very top, bad quarterbacks who could run (2011 Tebow Broncos, 1972 Bobby Douglass Bears). Even the best teams on the “run heavy” list (2009 Jets, 1999 Bucs) were teams who were good despite their quarterback.

I don’t think necessarily that calling a lot of pass plays leads to winning, it’s more that winning games and calling lots of pass plays are both caused by having a good quarterback. On the other hand, “establishing the run” is less of a philosophy teams want to follow and more something you have to do when your quarterback sucks.

• Anonymous

I do think teams should generally run on 3rd and 3 or less and maybe even sometimes on 3rd and 4. I agree with you that teams should rarely run on first down until they get to about the opponent’s 40. (About the point where I’d assume a shorter field might make the pass less effective, although I don’t have any data on that.) One thing I find especially bad is when teams run on second and 8-10 to set up a shorter third down. There is a clear difference between 3rd and 1 and 3rd and 5, but the difference between 3rd and 5 and 3rd and 10 is far from as big as most people perceive. (Third and 5 is converted 44% of the time while third and 10 is converted 35% of the time if I can recall the numbers correctly.) I think it’s best to try for a first down on second and long and use third down as your backup if the second down pass falls incomplete.

• Anonymous

I wonder about the idea of trying for a 10+ yard pass on every play. If the defense figures out that this is what you are doing, then that would make every play like a third and 10. As I stated in my last post, the odds of converting third and 10 is 35%. Your odds of not converting the first down on three plays would theoretically be .65^3, which equal 27.4%. That means you would convert 73% of your first and 10s, which is well above the NFL average of 66%. It might be an even higher likelihood of converting if you adjust for the high likelihood of defensive penalties on long passes. The main problem I can think of with this idea is that the completion percentage might drop below 35% if the defense keeps seeing the same types of plays over and over again.

• Anonymous

I always wonder how the CFL is roughly as high scoring as the NFL with one less down and 10 more yards on the field. Burke suggested once that this is because NFL teams often just wait for third down and don’t try hard enough to convert on first or second down. I think CFL teams go for third down a lot more often than NFL teams go for it on fourth down as well.

• Trepur

It’s also a wider field, meaning defenses have to cover more ground.

• shah8

Eh…

This is the wrong way to go about it.

As some have mentioned, football is situational. However, I do not think that it’s all that obvious wrt to formations and clocks.

I think this has more to do with the overall game having higher and higher level of baseline set for skill. This affects all of the positions, but we see clearest when it comes to higher variance activities. However, I don’t think, precisely, that rushing has declined in value. The *quality* desired from a rushing attack has *increased* while the *quantity* desired from a rushing attack has *decreased*. The gross and efficiency differential between rushing and passing simply reflects where the empty calories are. Your Matt Schaub, Andy Dalton, and other such caliber QBs beating up on bad defenses. The baseline skills have risen such that mediocrities can have their day in the sun, like Fitzpatrick’s six TD day. That doesn’t mean that the differential is *really* as big as it looks. When it comes time to *make plays* that efficiency of passing drops like a rock–just look at Matt Schaub or Andy Dalton in the playoffs, when you have to actually *make plays* passing-wise. When it comes time to *make plays*, high quality rushing attacks tend to be critical, witness what happened to the Steelers because La’Veon Bell wasn’t playing. In general, in the playoffs, the high variance of passing tends to be far more punishing.

Even as the number of rushing attempts decline, the importance of each rushing attempt probably is rising (my guess).

I do think the pendulum will shift a little back the other way, as much of this is also about a matter of how a franchise wants to fail. Thus, a change in aethetics is probable at some point.

• Timothy Potts

You all suck