Long Drives and the Running Game
Guest column by Ben Baldwin
Proponents of committing to the run game argue that running the ball is the best way to sustain drives and keep a team's defense off the field. After drafting Rashaad Penny in the first round last month, Seahawks coach Pete Carroll explained that running the football "helps our defense." But is that really true? Given that 33 percent of dropbacks gain first downs compared to only 22 percent of rush attempts, we already have reason to be skeptical. Sustaining drives requires obtaining first downs, and the best way to get first downs is to throw the ball.
(Side note: as I explored on Football Outsiders earlier this year, there is no evidence that long drives carry any subsequent benefit for a team's defensive performance after accounting for field position. So even if rushing does help to sustain drives, it doesn't necessarily follow that defensive performance will improve.)
For this piece, I again examined public play-by-play data from nflscrapR from 2009 to 2017. I excluded the final three minutes of each half since time becomes a factor. Code is here.
First, let's take a look at whether longer drives tend to have more rush attempts. This is straightforward: they don't.
This figure includes the results of all drives of at least five plays, with a sample size of at least 20 drives. Regardless of how long drives last, the rate of rushes hovers between 41 and 42 percent. Already we can conclude that there is nothing magic about rushing in terms of generating long drives.
Below are series and drive results when teams rush or pass on first down, broken down by field position:
The advantage of passing on first down is roughly constant in terms of earning a first down, scoring points, and (for plays outside the opponent's 40-yard line) avoiding punting. Since about 44 percent of offensive snaps take place on first down, this is already a big chunk of game situations where passing is more beneficial than rushing.
For drives that start with very poor field position, drives that begin with a pass tend to last longer. This is likely partly because teams that are pinned very deep in their own territory prioritize avoiding turnovers more than scoring or moving the ball.
Brian Burke's earlier work shows that series that open with a passing play are more likely to earn a subsequent first down. Here is what Burke wrote in 2012:
"Coaches focus on making sure that their third-down plays are as successful as possible, which means using the first two downs to gain relatively assured bits of yardage using running plays to set up third-and-short, rather than third-and-long, situations. This mindset is intuitively seductive, but it's ultimately self-defeating [...] The best third-down situation isn't third-and-1 or even third-and-inches. It's converting on first or second down, before ever reaching third down."
For the remainder of this piece, I will look at plays that occur 20 to 90 yards from the opponent's end zone. Here is the average in this range for each of the outcomes under consideration on first-and-10:
While first downs almost always come with 10 yards to go, there is more variance on second downs. Here is how the rush/pass decision affects series and drive results based on yards to go on second down:
Regardless of distance, teams that pass are again more likely to score and less likely to punt. On second-and-4 and greater, passing is also associated with longer drives; for second-and-3 or less to go, passing and rushing are pretty similar. Teams that run on second-and-2 yards or less to go are more likely to earn a first down in that series.
Here are the results on third downs:
On third down, rushing becomes favorable over a wider range of yards to go. While rushing is better than passing in terms of earning a first down on the series with 3 yards or less to go, it is favorable in terms of avoiding punting with 8 yards or less to go, and in terms of plays per drive with 6 yards or less to go. This is likely because teams rush so infrequently in these situations (for example, only 10 percent of the time on third-and-4, and even less on further distances) that defenses do not expect it.
Let's be generous and say that plays on second-and-3 or less and third-and-6 or less (including fourth downs) are rushing downs. Only six percent of NFL plays since 2009 were rush attempts in these rushing down situations. For maximizing likelihood of scoring on a drive, the only situation in which rushing is about as good as passing is third-and-3 or less. Only three percent of offensive snaps since 2009 were rush attempts on third or fourth down with 3 or fewer yards to go (and as Football Outsiders has shown, handing the ball off isn't even the best play in some of these situations; QB sneaks are better when the distance is very short). Rushing downs are high-leverage plays, but they are so rare that investing in success in these situations is unlikely to be an efficient use of resources.
Furthermore, teams that are successful when rushing in short-yardage situations are not necessarily successful rushing in other situations. The correlation between a team's rank in rushing success rate (using the 45/60/100 measure) with 3 or fewer yards to go and its rushing success rate with more than 3 yards to go is 0.21. (The correlation is similar if we remove fourth-quarter plays, so this has nothing to do with playing from ahead or behind.) This is statistically significant, but not especially large. A possible explanation for this weak relationship is that the short-yardage samples are so small that success rate in these situations is mostly noise, but that is not a good argument for investment in the run game.
To review, here are our three main findings:
1) It is not the case that long drives disproportionately have greater rush/pass ratios.
2) Rush attempts that come in situations where running the ball is better than passing the ball in terms of extending drives are extremely rare.
3) There is not strong carryover between rushing success in short-yardage situations and rushing success in other situations.
Justifying an investment in the run game by saying it will help extend drives is misguided. If you want to get first downs, throw the ball (unless it's a short-yardage situation).
An economist by trade, Ben Baldwin uses large datasets to try to learn about human behavior. He covers the Seahawks for The Athletic. Reach him on Twitter at @guga31bb.
22 comments, Last at 22 May 2018, 10:29pm
#1 by ammek // May 16, 2018 - 5:01am
This is very interesting indeed. Thank you Ben and FO.
I'd be fascinated to know whether the success rate for rushing on third-and-short differs according to formation. My sense, watching the Packers, is that when they try to run out of a two-back or heavy formation, they regularly fail to convert. If there's any substance to that, it would suggest the 'surprise' element is really important, and little would be gained by running more often on those downs.
#2 by billprudden // May 16, 2018 - 7:58am
I've often watched 3rd-and-shorts when the offense lines up in 11 (1x RB, 1x TE, 3x WRs) and the defense lines up with 4x rushers, 5 cover guys, and 2 medium to deep safeties, and wondered why that isn't a default run play every time. The offense has 6 blockers plus a running back and the defense has only six potential tacklers. RB ought'a convert a short run every time. Instead, it seems the offense throws throws throws...
#3 by JMM // May 16, 2018 - 9:02am
There is another way to measure "long drives", that is by game time elapsed. Shortening the game can be an advantage, particularly if one team has a less successful offense or if the team has a lead. It would be interesting to see if there is a significant difference between drive time and number of runs. If more passing on a drive extends the number of plays but leaves more game time, will it matter that it had an additional 1st down?
There is an unexplored trade off between game time and field position. This shows when teams will take a time out (worth 40 seconds of game time?) rather than give 5 yards of field position. If passing gives an extra 1st down per drive and improved field position , how much game time does it cost, or leave?
#6 by jtr // May 16, 2018 - 12:16pm
I would be curious to see how productive running plays are during late game run-out-the-clock situations. Just anecdotally, seems like we see an awful lot of times where a team with a lead in the 4th quarter gets into really predictable run-run-pass sequence every first down and ends up running into a lot of 3&longs because the defense is stuffing the obvious first and second down runs.
#7 by justanothersteve // May 16, 2018 - 12:31pm
Let's be generous and say that plays on second-and-3 or less and third-and-6 or less (including fourth downs) are rushing downs. Only six percent of NFL plays since 2009 were rush attempts in these rushing down situations.
Emphasis is mine
I like the article but this reasoning has me confused. First, I would never consider third-and-6 a rushing down even if rushing is a moderately successful strategy. I'd never consider anything more than third-and-2 a rushing down (and even third-and-two would be iffy) as rushing success in those situations is largely derived from the surprise element in a passing formation. IMO, this is conflating statistical success with actual practice. Also, how many of the longer first downs are from scrambling after the QB bails from a called pass play? QBs like Rodgers and Wilson will heavily skew the small number of mid-distance (3-6 yard) third down passing plays as they will usually either run for the first down running or get sacked (which doesn't count as a run) if they don't pass.
And why include fourth downs at all unless it's on the opponent's side of the field? Most coaches will punt even if it's fourth-and-inches on your own side of the field and once you approach the opponent's 20, the coach is more likely to try a FG on fourth down. The trend to go for it on fourth-and-short between the opponent's 30 and midfield is one of the best trends in recent years, but that really is the only situation where teams often use a standard play formation rather than a kick/punt team.
Just curious as to why this was done. It seems like you're trying to fit the statistics to fit your narrative. There's a lot of really good analysis here. But this part just seems forced.
#8 by guga31bb // May 16, 2018 - 12:52pm
That was me being *as generous as possible* to the running game. If you think the six percent number is too high and should be lower, that's fine: it only strengthens my case! And for the 4th downs, I'm only including 4th downs where a non-special-teams play was run (as you can imagine, pretty rare).
I should have mentioned this in the article but didn't (my mistake), but I classify scrambles as pass plays since they were called dropbacks (since as in all the pieces I've done on FO).
#18 by Aaron Brooks G… // May 16, 2018 - 9:09pm
Then the problem is the bit about in these rushing down situations.
You’ve stated 6% of potential rushing plays (this subset of all plays) actually are rushing plays.
What you meant is 6% of all plays (the superset) are rushes in this subset of downs and yards to gain.
The first is much more straightforward. The second is true, but misleading.
#9 by Will Allen // May 16, 2018 - 1:24pm
I kind of get tired of the simplistic thinking that this series is putting forth, masked as a new school approach to football. When someone writes", If you want to get first downs, throw the ball (unless it's a short-yardage situation).", they are ignoring the question of "What promotes a greater chance of any individual pass play being successful, other than having more talent than the defensive players?"
This parallels the frustration I had with the last piece, which sort of implied that pass rusher fatigue was irrelevant to successful defensive performance.
#15 by Noahrk // May 16, 2018 - 5:46pm
It's clear that if passing were 100% of the plays, passing efficiency would go down. But as long as passing efficiency is even a little higher than running efficiency, teams can stand to pass more. The run and pass distribution where both running and passing efficiency are exactly equal should point to the right balance between run and pass, I think.
#16 by Will Allen // May 16, 2018 - 6:02pm
Eh, this analysis ignores the fact that vast differences in personnel, and vast differences in personnel of opponents, makes generalized rules which are useful very hard to formulate. If you had a good qb, and didn't pass more than 95 % of the time against the 2007 Vikings, you were an idiot. If you adopted that same strategy against the 2013 Seahawks, except thi.js time you had a good run blocking o-line, but an average qb, you were similarly idiotic.
#12 by NJBammer // May 16, 2018 - 4:35pm
I think teams put too many resources into stopping the run.
Most statistical analysis I have studied shows that running the ball almost always leads to fewer expected points than passing. Furthermore play action is more effective than simply throwing, I assume this is because the linebackers freeze for a split second to stop the run which gives the receivers more time to get open. Studies like this also show that running the ball doesn't prolong drives more effectively than passing and as you note long drives don't make the defense better anyway.
So if all this is true, why don't defenses stop focusing on the run at all? Why not design a defense which focuses almost entirely on stopping the passing game? This will make the other team's passing game less effective and will also encourage them to run more, which reduces their expected points and chance to win.
Nickle has become more and more common in the NFL, why not just run the dime as your base defense package? Train the DL to always go after the passer first and play the run second. All DL/LB are drafted or signed for their ability to get to the passer, not stopping the run. Wouldn't this be a smart strategy?
#14 by Will Allen // May 16, 2018 - 5:43pm
You may have noticed a nontrivial number of defenses doing that against the Patriots. The Patriots frequently respond by running the ball down the opponents' throat, then passing when it becomes more advantageous, and winning very efficiently.
The arguments being put forth here are way, way, too simplistic.
#19 by dank067 // May 17, 2018 - 11:25am
Interestingly, every-down dime defense has emerged in the Big 12 to stop its spread offenses and has found legitimate success. Keep in mind that a lot of spread offenses in college (with maybe the exception of a pure air raid) are actually quite run-heavy and often feature a QB who is a running threat, so no sane defensive coordinator can ignore or disregard the running game.
A big part of making it work (against both the run and pass) is having versatile players, particularly hybrid-style safeties capable of crashing down on run plays—a lot of these packages feature 3 safeties. Against the run in particular, with versatile athletes that can all take on multiple assignments, defenses can manipulate their fronts/looks and run fits to make it extremely difficult for the offense to understand and execute its blocking assignments.
Link below gives a fantastic overview of how this works, for anyone who has any interest. (As more spread offense concepts bleed into the NFL, this certainly won't be far behind.) It's fascinating to watch Memphis go with 2 TEs and attempt to run over Iowa State, with 3 DL and just 5 or 6 guys total in the box, and utterly fail to block anyone.
#22 by gomer_rs // May 22, 2018 - 10:29pm
No NFL team has willingly used any of the running spreads. That's the difference. If Chip Kelly had been allowed to push the tempo instead of the NFL officials winning a succesful revolt to control the place of play, and had Kelly not been given control over player personnel, then it may have been different.
It's the running spreads, particularly the teams like Baylor that merry air raid passing concepts with power running game, that really change the game.
And NOBODY is likely to ever get the idea to money ball the QB position by hiring 3 running, Tebowesque, QBs for 7-10 Mil, and invest the saved money elsewhere.
I remember when they were the Sea-chickens.
#21 by J-Mac // May 19, 2018 - 12:02am
Personnel should also contribute to the decision-making process. E.g., if I had OGs the caliber of Grimm and Keuchenberg and a center like Mike Webster, blocking for, say, Csonka or even Earl Campbell, I'd have used the run a bit more than the percentages stated above.