Writers of Pro Football Prospectus 2008

18 Jul 2016

Running by Number of Backs, 2015

by Andrew Potter

In the lead up to the publication of Football Outsiders Almanac 2016, we're running a series of on-site articles discussing some of the charting stats you'll see in the book. Sterling Xie has already covered play-action passing (offense and defense), and Carl Yedor has discussed pass pressure (offense and defense).

Today, it's my turn to look at offensive rushing success based on number of backs in the backfield. For the purposes of this analysis, we're considering rushing attempts by any player who was lined up in the offensive backfield but was not at the quarterback position. Single-back formations mean one player other than the quarterback was lined up in the backfield and ran the ball; two-back or multiple-back formations mean more than one player other than the quarterback was in the backfield, and one of those players ran the ball. The numbers include direct snaps to a player who was lined up as a running back, but do not include scrambles, quarterback draws, wide receiver or tight end end-arounds or reverses, or "Wildcat"-style plays; only runs by a player who was lined up as a running back at the snap.

Without further ado, here's the table:

Team Rushing by Number of Running Backs, 2015
Offense Pct 2+ RB Rk 1 RB Rushes 2+ RB Rushes Difference
STL 61.7% 1 137 5.2 -19.0% 24 221 3.7 -18.9% 16 1.5 0.0% 20
GB 50.7% 2 181 4.1 -19.6% 25 186 3.9 -11.4% 12 0.2 -8.1% 28
SF 46.6% 3 166 3.8 -26.2% 31 145 3.2 -19.2% 17 0.6 -7.0% 26
BUF 46.4% 4 201 4.6 8.0% 2 174 4.9 -7.3% 9 -0.3 15.2% 8
NYJ 45.1% 5 212 4.7 -1.6% 6 174 3.3 -39.4% 30 1.3 37.8% 2
DET 45.0% 6 166 3.9 -22.8% 27 136 3.3 -26.9% 27 0.6 4.1% 15
CAR 42.9% 7 214 4.9 1.3% 4 161 3.1 -22.3% 24 1.8 23.6% 5
ATL 42.0% 8 221 4.0 -17.4% 22 160 4.2 -13.2% 14 -0.2 -4.2% 24
SEA 37.4% 9 241 4.5 1.3% 3 144 4.3 0.7% 4 0.2 0.6% 19
MIN 37.4% 10 260 4.7 -10.4% 17 155 4.3 -1.5% 6 0.4 -8.9% 29
TEN 37.2% 11 201 3.7 -25.2% 30 119 3.6 -21.7% 23 0.1 -3.5% 22
CIN 36.8% 12 241 4.2 -2.2% 8 140 3.7 -5.9% 8 0.5 3.7% 17
NYG 35.7% 13 245 4.5 -2.5% 9 136 3.2 -30.8% 28 1.3 28.3% 3
TB 34.4% 14 263 5.1 -5.5% 11 138 4.5 -21.0% 20 0.6 15.5% 7
PIT 30.5% 15 232 4.0 -2.0% 7 102 5.9 16.5% 2 -1.9 -18.5% 30
KC 29.0% 16 237 4.7 10.2% 1 97 3.8 -2.6% 7 0.9 12.8% 10
Offense Pct 2+ RB Rk 1 RB Rushes 2+ RB Rushes Difference
CLE 27.7% 17 222 3.9 -15.5% 21 85 3.2 -25.8% 26 0.7 10.2% 12
BAL 25.9% 18 260 4.3 -8.0% 14 91 3.2 -16.9% 15 1.1 8.8% 13
IND 25.8% 19 245 3.8 -24.1% 29 85 2.6 -48.6% 31 1.2 24.5% 4
HOU 25.3% 20 307 3.9 -18.0% 23 104 3.5 -25.2% 25 0.4 7.2% 14
OAK 25.3% 21 248 4.2 -10.9% 18 84 3.4 -21.3% 21 0.8 10.3% 11
DAL 24.9% 22 277 4.7 -7.0% 13 92 4.0 -8.1% 11 0.7 1.1% 18
NO 22.5% 23 283 4.4 -4.1% 10 82 2.6 -19.3% 18 1.8 15.1% 9
SD 18.6% 24 303 3.6 -23.1% 28 69 3.0 -21.5% 22 0.6 -1.6% 21
WAS 16.1% 25 334 4.0 -27.4% 32 64 3.1 -19.6% 19 0.9 -7.8% 27
JAC 13.2% 26 256 4.1 -22.8% 26 39 2.1 -62.3% 32 2.0 39.6% 1
NE 12.0% 27 293 4.0 -11.3% 19 40 2.7 -7.6% 10 1.3 -3.7% 23
CHI 11.7% 28 369 4.0 -0.9% 5 49 3.0 18.5% 1 1.0 -19.4% 31
ARI 8.5% 29 375 4.8 -8.2% 15 35 2.7 -12.2% 13 2.1 4.0% 16
DEN 5.8% 30 356 4.4 -10.3% 16 22 3.2 -32.7% 29 1.2 22.4% 6
MIA 3.8% 31 281 4.2 -15.5% 20 11 5.8 14.7% 3 -1.6 -30.2% 32
PHI 2.2% 32 403 4.1 -6.5% 12 9 4.0 0.3% 5 0.1 -6.9% 25
NFL 29.0%
8230 4.3 -10.6%
3349 3.7 -16.8%
0.7 5.2%

General Trends

Strategy in the NFL is a constantly evolving, fluid concept and so, contrary to many analysts' complaints, is conventional wisdom. In 2007 and 2008, the average NFL team ran out of a two-back formation 56 percent of the time (the median in each year was very slightly higher at 57 percent and 61 percent due to a low-end outlier, discussed below). The league combined for more than 7,000 such attempts in both years, and more than 12,000 total rushing attempts from the backfield. The league leaders in each season ran more than 75 percent of their rushing attempts from multiple-back formations, and in 2007 three teams were over that 75 percent mark.

Norv Turner's San Diego Chargers, featuring the legendary combination of LaDainian Tomlinson and Lorenzo Neal, set the highest mark in the league in 2007 with 76.5 percent of their attempts coming from multiple-back formations. Neal left after that season for Baltimore, where he found an even higher workload as San Diego's raw number of multiple-back attempts -- 342 -- was eclipsed by the Ravens and their new head coach John Harbaugh. First-team All-Pro Le'Ron McClain, Willis McGahee, and a rookie Ray Rice combined for 376 multiple-back rushing attempts, more than 2,000 rushing yards, and a conference championship defeat to the eventual Super Bowl champion Steelers.

Despite the raw numbers, the Ravens ran so much out of both one- and multiple-back sets that they were only fifth in percentage of multiple-back runs at 72.3 percent, with Mike McCarthy's newly Favre-less Green Bay Packers the league leaders at 76.3 percent. The only team in either season to run from multiple-back formations less than 25 percent of the time was Peyton Manning's Indianapolis Colts, and they remained unique as the only team lower than 20 percent each year from 2007 right through Manning's neck injury that cost him the entire 2011 season.

Fast-forward to 2015, and for the first time in our records not one team rushed from two-back sets more than two-thirds of the time or had at least 230 such attempts. Only three teams have been over 70 percent or 300 attempts since 2011, and all of those were coached by Harbaughs: John Harbaugh's 2011 Ravens had the highest mark on record at 80.4 percent, and Jim Harbaugh's 2012 and 2013 49ers were only slightly tamer at 74 and 78 percent, respectively. By contrast, from 2007 to 2010 there were at least three teams higher than 70 percent every single year.

The top and bottom trends are reflected in the league-wide numbers. Sterling noted in last year's article that "the league-wide percentage of running plays from sets with two or more backs has declined every single season since 2008, and is now down to 33 percent after sitting at 40 percent in 2013." That trend continued through 2015, with a new low of just 29 percent of league-wide rushing attempts coming out of two-back sets. Only two teams ran more than half of their rushing attempts from two-back sets: St. Louis (predictably) and Green Bay (less so). Nine teams ran from two-back sets less than 20 percent of the time, and four of those were below 10 percent. Thirteen different franchises have had at least one season below 20 percent since 2012, after the Colts were the only franchise to achieve that even once from 2007 to 2010.

The total number of rushing attempts out of the backfield has also been falling fairly consistently, though less drastically, and in 2015 hit a nine-year low of 11,579. Teams are running less (though not by much; the actual raw attempt total including all runs is only down 498 attempts from 2007 to 2015, or about one rush per team per game), they're running out of passing sets instead of running sets, and many are running in historically unconventional ways: read-option, designed quarterback runs, Tavon Austin.

Volume Numbers

Speaking of Austin, for the second straight season Jeff Fisher's 2005 2015 St. Louis Rams led the NFL both in number and percentage of rushing attempts out of two-back formations. Fisher's offense barely saw a rushing attempt it didn't like: it ranked third by run percentage in the first half (45 percent), second on first down (61 percent), first in power situations (73 percent), third when trailing in the second half (31 percent), and third when leading in the second half (60 percent). No other offense ranked in the top three of more than two of those categories, though Chicago was top-five in all except power situations. Finally, as Sterling noted in the 2015 play-action offense article, the Rams also ranked fourth in the percentage of pass attempts which featured play-action, meaning approximately six of every ten Rams plays featured either a run or the threat of a run as its basis.

Of course, that run emphasis is easy to justify when you look at their passing game. The Rams offense ranked 14th in rushing DVOA but 31st in passing -- and that 31st-ranked passing offense was actually 32nd on straight dropback pass plays, but ranked ninth with the aid of play-action. We hear during just about every game broadcast that teams should run until the other team proves they can stop it; St. Louis seems to have taken the more negative angle of "don't pass until we absolutely have to." Whatever the reasoning, the net result is another year with more than a 10 percent lead on the rest of the league... or another year with more than a 10 percent deficit behind the rest of the league, depending on your perspective.

At the opposite end of the table, Philadelphia yet again ran from two-back sets less than 5 percent of the time, though the nine two-back runs they had in 2015 was still more than four times the total they had in 2014. Two of those nine runs came in Week 17, after Chip Kelly was removed as head coach; from Weeks 1 to 16 the Eagles yet again averaged less than one two-back rush every two games. If Kelly continues to rely on one-back sets with his new team in San Francisco, 2016 is set to be an extended offseason for Bruce Miller. Miami was the only other team to average less than one two-back rushing attempt per game, and Denver the third to average less than two.

Efficiency Numbers

This is where the strategy question really comes into play. Every team wants to be able to run the ball at least a token amount. The question isn't "whether," but "how" -- how the plays are drawn up and executed, how often, how effectively. The offense has proactive control over their formation and play calling, and should in theory be able to dictate running sets to at least some degree. If a team is good at one thing and bad at another, it's reasonable to expect them to know and adjust accordingly.

Despite this, there appears to be basically no correlation whatsoever between how often a team ran from multiple-back sets and how good the team was at doing so. The team that ran the most from two-back sets (St. Louis) was mediocre in efficiency (-18.9% DVOA, 16th in the league). The team most efficient at it (Chicago, 18.5%) had only ten more attempts than the team that was by far the least efficient (Jacksonville, -62.3%) (!??!), and 36 fewer than the second-least-efficient team (Indianapolis, -48.6%). The team that ran from two-back sets the least (Philadelphia) was slightly above average, and the team that ran the second-least (Miami) was well above average (14.7%), but the team that ran two-back the third-least (Denver) was significantly below average (-32.7%). Denver was much more efficient (well, less inefficient) out of one-back sets, Miami much more efficient out of two-back sets, and Philadelphia and St. Louis not much different whichever set they chose.

Overall league-wide efficiency was greater, as we'd expect, from one-back sets. DVOA for two-back sets hit a nine-year low at -16.8%, and has now declined every year since 2011. The Jacksonville Jaguars set a new futility record for two-back rushes (minimum 16 attempts*) at -62.3% DVOA, and though the Chicago Bears matched the fourth-highest mark on record at 18.5% (a number shared by the 2008 Panthers and the 2012 Giants), the overall trend is definitely downward. It's not just a question of more efficient passing skewing the running numbers either, as two-back yards per attempt has likewise declined every year since 2012. The single-back number has remained constant at around 4.3 yards per attempt over that period, though single-back DVOA has also declined considerably over the last two seasons.

* I'm applying a minimum threshold of 16 attempts here, or one per game. Philadelphia hit the lowest-ever single-season two-back DVOA mark of -66.6% on a paltry six attempts in 2013, then set the highest-ever mark of 56.3% on two attempts in 2014. Their overall running game was far more efficient in 2013 than 2014. Small samples, y'all.

We've long known that passing is much more efficient than running from a pure "progress toward first down" perspective. What is becoming more and more apparent is that increasing the number of backs, far from aiding the running game, is actually more of a hindrance for most teams. As we'll see on Wednesday, almost every run defense in the league is better against two-back sets than single-back sets. Only three offenses were better by at least 10.0% in DVOA with two backs than one, and two of those were on a sample of less than one in eight of their rushes. (The third, Pittsburgh, perhaps merits more investigation.)

A third of the league was appreciably worse with two backs than with one, yet nine of those teams ran more than 25 percent of their running plays from multiple-back sets. Even with their relatively small total of 13 percent of plays, the Jacksonville Jaguars may have been the worst offenders, as this sequence will attest:

  • 1-1-BUF 1 (7:29) JAC 93 reports eligible 21-T.Gerhart right guard to BUF 1 for no gain (53-N.Bradham).
  • 2-1-BUF 1 (6:51) 21-T.Gerhart right guard to BUF 1 for no gain (52-P.Brown).
  • 3-1-BUF 1 (6:17) 21-T.Gerhart left guard to BUF 1 for no gain (98-A.Carrington).
  • 4-1-BUF 1( 5:37) 21-T.Gerhart right tackle to BUF 1 for no gain (30-B.Rambo).

Every one of those plays was run from a goal-line formation, no wide receivers, and defensive lineman Tyson Alualu at fullback. The Bills run defense was a basically-average -22.3% against two-back formations, but a league-worst 3.4% against single-back formations. The numbers are clear: this is the worst two-back rushing offense in the league, facing a team that is much better against two-back formations than single-back sets. So of course, the Jaguars went heavy and were stuffed on basically the exact same play four times. That sequence is as clear a demonstration as I can find for how you go from being merely bad at running with one back to the worst team on record running with two. (Oh, and they didn't learn: the Jaguars almost repeated the sequence later in the season against Tennessee, but Blake Bortles eventually scored on a quarterback sneak on fourth down.)

Fortunately, fewer and fewer teams are making the same mistake. Conventional wisdom is evolving, even it not as quickly as we'd prefer. Offenses have been moving further away from two-back formations every year since 2008, and there's very little reason to think that will change in 2016.

Posted by: Andrew Potter on 18 Jul 2016

17 comments, Last at 20 Jul 2016, 7:15pm by Sherman


by Bright Blue Shorts :: Mon, 07/18/2016 - 1:51pm

Don't know how many seasons back these numbers go back ... but wonder where Jeff Fisher ranked against the rest of the league when he had Steve McNair in Tennessee?

by Andrew Potter :: Mon, 07/18/2016 - 2:20pm

I believe the earliest numbers we have are from the 2007 season, which is why I only went back that far.

As for Fisher, his latter-day Tennessee teams were fairly middle-of-the-pack.

2007: 22nd at 52.3%
2008: 20th at 53.7%
2009: 12th at 61.9%
2010: 19th at 49.9%

His first year in St. Louis they ranked 26th at 33%. It's likely that his offense isn't a philosophical commitment, and would be drastically different with a better quarterback.

by tuluse :: Mon, 07/18/2016 - 2:55pm

How do you deal with end-arounds? Do Tavon Austin's 52 rushing attempts affect STL's numbers?

Edit: I should read more closely. I see end-around are removed.

by Andrew Potter :: Mon, 07/18/2016 - 3:22pm

Paragraph two of the article introduction details what's counted and what isn't, but the basic rule is this: if the runner was lined up in the backfield, it's counted here. If he wasn't, it isn't. Quarterback runs and "wildcat" plays are the exceptions, as those aren't counted.

Of Austin's 52 rushing attempts, nine were counted here. Three in Week 1 versus Seattle, one each against Washington and Minnesota, two against Detroit and two in Week 17 at San Francisco. They went for a total of 45 yards, two first downs and a touchdown. The first five, totalling 21 yards and a touchdown, were from multi-back sets; the latter four for 24 yards and two first downs came from single-back sets.

by Bright Blue Shorts :: Tue, 07/19/2016 - 4:00am

Thanks Andrew.

With Goff being a rookie, I'd expect the Rams to continue to emphasise the run this coming year but if Fisher lasts beyond next year then perhaps we'll get to see.

by theslothook :: Mon, 07/18/2016 - 3:49pm

This sounds like a crazy statement, but at this point, every team would be better off(crappy qb or not) from transitioning to a ne style shotgun quick pass spread team with a shifty receiver style back. It may not lead to a good offense, but it will be better than the alternative.

by Bright Blue Shorts :: Tue, 07/19/2016 - 3:58am

If everybody started doing it, wouldn't defenses find it easier to defend?

Isn't a part of New England's success in doing what other teams aren't doing. Firstly because it's harder to defend and secondly because the players to run the scheme are undervalued and therefore leave salary cap for other positions.

by theslothook :: Tue, 07/19/2016 - 1:29pm

One of the big reasons that the passing explosion has occurred is because of YAC and the short passing game. I don't think its a new England specific strategy - as you see the depth of target shrink every year while the offensive payoff has shown no commensurate decrease. At this point - I think its simply a case that defenses lag behind in their ability to stop the short passing game. More teams should do it until defenses learn to stop it. But so far they have no been. This year was once again the highest anya of any year in history - coming on the heels of last year's record breaking anya.

by theslothook :: Tue, 07/19/2016 - 1:29pm

One of the big reasons that the passing explosion has occurred is because of YAC and the short passing game. I don't think its a new England specific strategy - as you see the depth of target shrink every year while the offensive payoff has shown no commensurate decrease. At this point - I think its simply a case that defenses lag behind in their ability to stop the short passing game. More teams should do it until defenses learn to stop it. But so far they have no been. This year was once again the highest anya of any year in history - coming on the heels of last year's record breaking anya.

by zenbitz :: Mon, 07/18/2016 - 7:15pm

Not much of a comment, but word out of camp in SF is that Bruce Miller is converting to TE.

by Sherman :: Tue, 07/19/2016 - 8:09am

It seems obvious to me that the Y/A stats are a very questionable measure for this purpose. Multi-back sets are often used in short yardage situations, and if you score on forth and goal from the one, that's a huge success, but it will decrease your Y/A. I'm surprised this isn't mentioned.

Does DVOA take into account this down and distance factor?

by Andrew Potter :: Tue, 07/19/2016 - 8:55am

Down-and-distance is actually one of the key components of DVOA, as explained here.

Also, the short-yardage effect is actually something of a myth. Only around 650 of the 3349 multiple-back plays came with under three yards to go. By far the most common down-and-distance was first-and-10, with around 1750 plays, and the average yards to gain was 7.6.

I'm being slightly vague in my numbers here because I forgot to exclude some QB runs and aborted snaps when filtering the play-by-play, but those only account for about 25 extra plays.

by lightsout85 :: Wed, 07/20/2016 - 4:47pm

I love moments like this, where the numbers show that commonly held beliefs are completely wrong. Just like when people claim that a TD-heavy RB would have a better YPC if they didn't have so many TDs.

by Sherman :: Wed, 07/20/2016 - 7:15pm

"Also, the short-yardage effect is actually something of a myth. Only around 650 of the 3349 multiple-back plays came with under three yards to go"

Well how about that? Around 19%. I would have guessed at least 50%

So if the additional backs are not in the backfield to act as a potential lead blocker on a short yardage focused power running play, what ARE they doing there? Something I'll have to watch this coming season.

by brian30tw :: Wed, 07/20/2016 - 9:14am

Looks like the (positive) 22.8% for JAC 1-RB DVOA should actually be (negative) 22.8%.

by Andrew Potter :: Wed, 07/20/2016 - 11:13am

Fixed. Thanks.

by lightsout85 :: Wed, 07/20/2016 - 7:26pm

PIT's case is an interesting one, IMO. Not that 1/2+ RB is directly analogous to nickel/base defenses faced, but PFF's article (earlier on the off-season) about the types of defenses backs faced also highlighted LeVeon Bell performing much better in a traditional running environment (but, ran well vs. nickel in the past). He's such a talented back, the difference is unusual (that he's worse under conditions that on-average produce better runs).

Though, the same could be said for Adrian Peterson - his talent is unquestioned but he's always been considerably better with a lead blocker (/vs base defenses, as he started in the NFL running mostly single-back). It'd be interesting to see a write-up on what traits are shared by backs who are very productive/talented, but generally perform worse under conditions that on average backs perform better under (1 RB / vs 5+ DBs).