Analyzing NFL running games through 10 weeks

NFL teams are passing more frequently and more effectively than ever before. Given enough opportunities, most teams will eventually connect on big plays through the air. But while running backs have taken a backseat in most offenses, a successful rushing attack is still a significant component in most effective offenses.

Success

As teams — and by extension, their opponents — become more prolific at passing, the opportunity cost of not passing increases. That makes an unsuccessful run particularly damaging. A run on third and short that forces a punt, or a run on 1st or 2nd down that makes it harder for his team to move the chains, hurts a team more significantly than ever before. In the ’70s, the running game was supposed to win games for teams, as running was a more effective optionthan passing. In some ways, the goal of the running game now is to not mess things up for the passing game, by forcing a punt or an unfavorable third down situation.

About 25 years ago, Bob Carroll, Pete Palmer and John Thorn wrote the Hidden Game of Football, a fascinating book on football theory and win probability. They went through and graded each play as a success or failure based on how many yards were gained as a percentage of how many yards were needed to pick up a first down or touchdown.

When I wrote a series on the most dominant running backs of all-time, I noted that yards per carry was a misleading statistic for running backs. Rushing is more about consistent success than passing, and rushing has a positive feedback loop in place that might lower yards per carry averages. Yards per carry is highly sensitive to large runs, decreasing the correlation it would have with the overall strength of a running game. I had a discussion with Brian Burke about this a couple of years ago, and he now uses rush success rate in his team efficiency models.

So to analyze NFL running games so far this season, I decided to use my own version of rush success rate. Here’s exactly what I did:

  • I looked at every play from scrimmage where the running back was credited with a carry. Note that all carries by non-running backs were specifically excluded. I then removed all instances of 3rd or 4th down carries where the player needed to gain more than 5 yards for a first down. It doesn’t feel right blaming the runner for not gaining 7 yards on 3rd and 7, so I have simply eliminated all failures from the data. However, on the rare occasions where a running back did convert for the first down, those plays were not removed. This has only happened 21 times this season, highlighted by an Arian Foster 16-yard run on 3rd and 15 against the Ravens.
  • For all 3rd or 4th downs, a success is defined as gaining the first down. A failure is every rush that does not gain a first down. On all downs, a first down (which includes a touchdown) is a success.
  • On all second downs, a success is achieved when the player gains at least 50% of the yards needed for the first down. This means that 2nd-and-10 runs are failures unless they pick up 5 yards; on 2nd-and-7, the running back must gain at least 4 yards. A rush for one yard on 2nd-and-3 is a failure, and so on.
  • On first down, a running back is credited with a successful play if he gains at least 40% of the yards needed; therefore, four yards are required on 1st-and-10 before the running back is given credit. On 1st and goal from the 5, a two-yard gain would be considered a success.

Using this criteria, 49% of all running plays by running backs have been successful this season. The average gain on all carries was 4.34 yards.

It’s worth remembering that 9 or 10 games it not a very large sample size, especially when the success rate among running backs is generally in the tight window between 45 and 55 percent. On 120 carries, we’re only talking about 12 more successful plays, which may only amount to 12 more yards. Over time, and on the team level, I have a bit more faith in the data. That said, the table below lists the running backs with the largest number of successful runs this season:

Player                  Tm    Car   Succ Succ%  RshYd YPC
Maurice Jones-Drew      JAX   188   94   0.50   845   4.5
Fred Jackson            BUF   161   91   0.57   920   5.7
LeSean McCoy            PHI   163   89   0.55   891   5.5
Adrian Peterson         MIN   178   88   0.49   837   4.7
Matt Forte              CHI   158   80   0.51   853   5.4
Cedric Benson           CIN   149   79   0.53   593   4.0
Chris Wells             ARI   146   78   0.53   588   4.0
Michael Turner          ATL   178   78   0.44   787   4.4
Arian Foster            HOU   165   77   0.47   711   4.3
Shonn Greene            NYJ   144   74   0.51   557   3.9
Frank Gore              SFO   158   74   0.47   754   4.8
Ben Tate                HOU   118   68   0.58   673   5.7
Rashard Mendenhall      PIT   134   66   0.49   515   3.8
BenJarvus Green-Ellis   NWE   115   65   0.57   455   4.0
Steven Jackson          STL   133   63   0.47   650   4.9
Willis McGahee          DEN   126   62   0.49   637   5.1
Ray Rice                BAL   134   62   0.46   544   4.1
Ryan Mathews            SDG   116   60   0.52   531   4.6
DeMarco Murray          DAL    97   59   0.61   646   6.7
Chris Johnson           TEN   146   59   0.40   496   3.4
Marshawn Lynch          SEA   126   58   0.46   505   4.0
James Starks            GNB   108   57   0.53   500   4.6
Ahmad Bradshaw          NYG   111   54   0.49   440   4.0
Michael Bush            OAK   105   53   0.50   450   4.3
Mark Ingram             NOR    90   52   0.58   338   3.8
Darren McFadden         OAK   110   50   0.45   613   5.6
Reggie Bush             MIA   101   48   0.48   469   4.6
LeGarrette Blount       TAM   100   47   0.47   434   4.3
Daniel Thomas           MIA    97   45   0.46   352   3.6
Jackie Battle           KAN    86   44   0.51   402   4.7
Tim Hightower           WAS    83   41   0.49   299   3.6
Mike Tolbert            SDG    73   41   0.56   283   3.9
Pierre Thomas           NOR    72   40   0.56   340   4.7
Delone Carter           IND    83   39   0.47   325   3.9
DeAngelo Williams       CAR    83   37   0.45   391   4.7
Jonathan Stewart        CAR    73   37   0.51   346   4.7
Isaac Redman            PIT    64   37   0.58   245   3.8
Peyton Hillis           CLE    59   35   0.59   216   3.7
Jahvid Best             DET    82   34   0.41   388   4.7
Brandon Jacobs          NYG    78   34   0.44   253   3.2
Felix Jones             DAL    65   33   0.51   256   3.9
Joseph Addai            IND    58   32   0.55   257   4.4
Ryan Grant              GNB    68   30   0.44   250   3.7
Montario Hardesty       CLE    75   29   0.39   245   3.3
Cadillac Williams       STL    60   29   0.48   230   3.8
Dexter McCluster        KAN    55   29   0.53   265   4.8
Darren Sproles          NOR    50   29   0.58   346   6.9
Donald Brown            IND    61   28   0.46   267   4.4
Bernard Scott           CIN    65   27   0.42   223   3.4
Kendall Hunter          SFO    50   27   0.54   253   5.1
Lance Ball              DEN    50   25   0.50   225   4.5
Thomas Jones            KAN    67   24   0.36   199   3.0
Chris Ogbonnaya         CLE    47   24   0.51   170   3.6
Ryan Torain             WAS    50   23   0.46   186   3.7
Roy Helu                WAS    43   23   0.53   213   5.0
Danny Woodhead          NWE    43   23   0.53   180   4.2
Ricky Williams          BAL    50   21   0.42   209   4.2
Deji Karim              JAX    54   20   0.37   131   2.4
Jacquizz Rodgers        ATL    32   20   0.63   137   4.3
LaDainian Tomlinson     NYJ    43   19   0.44   152   3.5
Marion Barber           CHI    43   18   0.42   143   3.3
Earnest Graham          TAM    37   18   0.49   206   5.6
Danny Ware              NYG    28   18   0.64    91   3.3
Maurice Morris          DET    41   17   0.41   172   4.2
Stevan Ridley           NWE    35   17   0.49   195   5.6
Derrick Ward            HOU    29   17   0.59    88   3.0
Javon Ringer            TEN    44   16   0.36   147   3.3

As painful as it is to watch the Jacksonville offense, it’s scary to fathom how bad it would look without Maurice Jones-Drew. He leads the league with 94 successful carries. If his 50% success rate does not impress you, consider that backup Deji Karim has been successful on only 37% of his runs.

DeMarco Murray has an incredible 6.7 yards per carry average, in part thanks to a 91-yard touchdown run on his first carry against the Rams. But Murray has been more than just a home run hitter, succeeding on an incredible 61% of his limited carries so far this season.

Meanwhile, Jahvid Best is drawing comparisons — both good and bad — to another former Detroit Lion. Like Barry Sanders, Best is a big-play machine, and his 4.7 yards per carry average looks impressive. But Best gained 131 yards on two carries against the Bears and has averaged just 3.2 yards per carry on his other 82 runs. He also averaged only 3.2 yards per carry as a rookie, and his inability to consistently gain yards has left the Lions ranked 29th in rushing first downs. Being only a big-play threat isn’t bad or even necessarily an insult; but to truly understand how an offense is operating, knowing that Best averages 4.7 yards per carry will not tell you much. And Best is only a small part of the problem: the Detroit offensive line is a horrible run-blocking unit as a whole.

For players, I thought it fairer to grade them by total number of successful plays. But on the team level, let’s sort by success rate:

Team	Car	Succ	Succ%	RshYd	YPC
NOR	233	132	56.7%	1109	4.8
BUF	185	102	55.1%	1050	5.7
NWE	203	110	54.2%	 861	4.2
SDG	213	115	54.0%	 904	4.2
DAL	207	111	53.6%	1039	5.0
PIT	226	120	53.1%	 967	4.3
PHI	190	100	52.6%	 983	5.2
ARI	196	102	52.0%	 745	3.8
HOU	323	167	51.7%	1498	4.6
CIN	224	114	50.9%	 876	3.9
MIN	199	101	50.8%	 971	4.9
GNB	187	93	49.7%	 774	4.1
WAS	177	88	49.7%	 705	4.0
NYG	218	107	49.1%	 789	3.6
SFO	214	105	49.1%	1027	4.8
IND	202	99	49.0%	 849	4.2
OAK	233	114	48.9%	1190	5.1
NYJ	211	103	48.8%	 780	3.7
MIA	214	104	48.6%	 858	4.0
CAR	159	77	48.4%	 747	4.7
CHI	212	102	48.1%	1019	4.8
DEN	217	104	47.9%	1040	4.8
JAX	247	118	47.8%	1003	4.1
CLE	192	91	47.4%	 643	3.3
STL	202	95	47.0%	 906	4.5
TAM	160	75	46.9%	 726	4.5
KAN	228	105	46.1%	 965	4.2
ATL	227	103	45.4%	 964	4.2
BAL	190	84	44.2%	 767	4.0
SEA	175	77	44.0%	 686	3.9
DET	182	74	40.7%	 750	4.1
TEN	197	78	39.6%	 659	3.3

The top three teams may surprise you. The most publicized portion of the Saints’ running game has been the failure of Mark Ingram to live up to high expectations. And while there’s nothing impressive about his 340 rushing yards, 3 touchdowns or 3.7 yards per carry average, his 58% success rate is something to be lauded. That high rate has been equaled by the other newcomer to the Saints backfield, Darren Sproles. To be clear, the differences end there: not only does Sproles have an incredible 6.8 yards per carry average, his 60 catches rank third in the league and his 1,604 all-purpose yards lead the NFL. As Chris described, Sproles is the top space player in the game today. Throw in Pierre Thomas’ also outstanding success rate (56%), and one could fairly attribute much of the success of the Saints rushing game to the offensive line. New Orleans a fantastic pair of run-blocking guards in Jahri Evans and Carl Nicks, while fullback Jed Collins has been masterful filling the void left by the retired Heath Evans.

In Buffalo, it’s all about Fred Jackson. His 57% success rate trails only Ben Tate among running backs with at least 100 carries. The Bills have looked terrible the last two weeks, but Jackson’s fantastic season continues. He leads the NFL in rushing yards, yards per game and yards from scrimmage, while averaging a healthy 5.6 yards per carry and recording an incredibly high success rate. Jackson’s ability is finally being recognized, but did you know that BenJarvus Green-Ellis’ 57% success rate ranks third in that category? And that’s after a couple of miserable games (perhaps due to injury) against the Steelers and Jets. Danny Woodhead (53%) and Stevan Ridley (49%) have been slightly less successful, but the Patriots have been a very effective team on the ground this season. It’s tempting to give Brady or Brees credit for that because of the respect defenses must pay them, but don’t forget that many Peyton Manning and Philip Rivers teams in years past struggled to run the ball despite having fantastic passing attacks. The Jets running attack has declined each season as Mark Sanchez progresses from a rookie to third year quarterback. There isn’t anywhere near the correlation between passing and rushing success in the NFL as there it as the lower levels of play. If anything, I’d credit Brady because of his ability to check to and from running plays based on how the defense is lined up (but again, see Manning, Peyton).

In fact, while Rivers is having his worst season in years, the Chargers running game has never been better post-Tomlinson. Mike Tolbert has the lower yards per carry average but the higher success rate, while Ryan Mathews has been the big play running back for the Chargers. Together, the duo give San Diego a fantastic running attack that is shockingly the strength of the team. Just last season, the Chargers led the league in net yards per pass and net yards per pass allowed while being stout against the run; San Diego has regressed everywhere but the running game (and special teams) this season.

Houston’s running attack has received praise throughout the football world, and I’m not going to tell you that Arian Foster, Ben Tate and the Texans’ offensive line aren’t fantastic. The Texans are 3rd in rushing yards and 2nd in rushing touchdowns (on a per game basis), so the stats back up the theory. But those statistics also reflect the fact that the Texans have had the highest points differential during their average play of any team this season. The Texans have led after the first quarter in nine of ten games thus far; on average, Houston is leading by 11.6 points at halftime. That’s played a key role in the high raw rushing totals, as Houston has at least 32 carries in each of their seven wins but fewer than 25 carries in each of their three losses. Lest one be led to believe that the Texans’ problem in their losses is that they are simply not running frequently enough, the issue is that the Houston running backs have recorded a 56% success rate in wins but just a 37% success rate in losses. There’s no doubt that Houston has a great running game, but I’m curious to see how their rushing production changes once they play in more competitive games. Unfortunately, that’s almost certainly bound to happen now.

On the other end of the spectrum, there’s no way to put Chris Johnson and the Titans in a positive light. Even with Johnson gaining over 100 yards against the Panthers this weekend, it was more methodical than explosive: Tennessee running backs had 33 carries for 149 yards. Tennessee still ranks 32nd in rushing yards even after their best rushing performance of the season.

I wrote about the Baltimore Ravens incredibly high adjusted pass-to-rush ratio for the New York Times’ blog last week. While we all love Ray Rice, his low success rate is likely playing a part in the offensive philosophy. Rice (46%) and Ricky Williams (42%) have not done enough to dissuade Cam Cameron from shifting the team’s offensive philosophy, even if the passing attack is led by Joe Flacco. On 70 first-and-ten rushes, Rice is averaging 4.0 yards per carry but only succeeding on 44% of his carries; Williams has been even worse, with just a 36% success rate on 1st-and-10 and a 3.5 yards per carry average. That doesn’t make life easy on an offensive coordinator. But then again, calling 52 pass plays and only five runs for Rice is shifting things way too far in the other direction, and won’t exactly appease many fans.

Stopping the run

When the 49ers were 4-1, I wrote that Jim Harbaugh was doing another masterful job in the Bay Area. In that article, I noted that the 49ers were the only team that had not allowed a rushing touchdown. At 8-1, both of those statements remain true. The 49ers rank first in the league in rushing yards allowed and yards per rush allowed, as well. They’ll have to settle for #2 in defensive rush success rate: here, of course, a lower success rate is better:

Team	Car	Succ	Succ%	RshYd	YPC
SEA	220	87	39.5%	771	3.5
SFO	157	65	41.4%	516	3.3
ATL	176	73	41.5%	676	3.8
DET	199	84	42.2%	944	4.7
JAX	212	92	43.4%	835	3.9
DEN	215	95	44.2%	908	4.2
CHI	159	72	45.3%	832	5.2
BAL	219	100	45.7%	750	3.4
CIN	199	93	46.7%	741	3.7
PIT	209	99	47.4%	868	4.2
OAK	194	92	47.4%	986	5.1
HOU	189	90	47.6%	823	4.4
MIN	193	92	47.7%	706	3.7
WAS	209	100	47.8%	905	4.3
PHI	197	95	48.2%	980	5.0
NYJ	235	114	48.5%	963	4.1
NYG	195	96	49.2%	879	4.5
TEN	225	111	49.3%	920	4.1
KAN	237	120	50.6%	975	4.1
TAM	224	115	51.3%	1104	4.9
DAL	181	93	51.4%	825	4.6
NOR	195	101	51.8%	1051	5.4
MIA	208	108	51.9%	813	3.9
STL	215	112	52.1%	1132	5.3
SDG	211	112	53.1%	957	4.5
CAR	232	125	53.9%	1164	5.0
CLE	257	139	54.1%	1168	4.5
ARI	220	119	54.1%	941	4.3
GNB	157	85	54.1%	728	4.6
IND	311	170	54.7%	1292	4.2
BUF	210	118	56.2%	950	4.5
NWE	183	103	56.3%	747	4.1

For all the criticism Pete Carroll’s Seahawks have endured, let’s give them credit for having a fantastic run defense. Pro Football Focus ranks the Seattle run defense as third best in the league, in large part thanks to the spectacular work by Alan Branch, Red Bryant, Brandon Mebane and K.J. Wright. Branch ranks as PFF’s top defensive tackle in the league against the run while Mebane is tied for second among defensive tackles (to Cleveland’s Ahtyba Rubin) in stops.

The Falcons and Broncos have been surprisingly difficult to run on in 2011. In Denver, rookie of the year candidate Von Miller has been an absolute monster not just at getting to the quarterback but at frustrating opposing running games. In Atlanta, John Abraham and Ray Edwards have been more notable for their pass rushing prowess than anything else in prior years, but both are holding up very well against the run this season.

At the bottom? The Bills ranked last in rushing yards and yards per carry allowed in 2010 after ranking 30th in those categories in 2009. And while Buffalo got off to a hot start this season, their run defense looks to be a culprit again in 2011. I can’t even imagine how ugly they look against Fred Jackson in practice.

Everyone knows the Patriots rank dead last in passing yards allowed, but the run defense hasn’t been any better. Vince Wilfork is having a miserable season and New England seems incapable of doing anything well on defense against any team besides the Jets.

No surprise seeing the Colts near the bottom of the list. For the first time in years, the Indianapolis run defense hasn’t been bailed out by Peyton Manning in 2011, who has at times covered the spotlight on the awful run defense.

The most surprising name in the bottom five belongs to the defending Super Bowl champs. It’s tempting to think that perhaps Green Bay just allows some runs in garbage time as the Packers always seem to be leading, but that’s not the case. The rush defense struggled mightily against the Saints (70% success rate), Chargers (72%) and Panthers (69%) in games that never truly entered garbage time. The pass defense has been spotty this season, ranking below average in every category except interceptions, but the run defense has been even worse. With New Orleans’ strong running attack, they present as tough a matchup as anyone to stop the Packers from repeating as Super Bowl champs. I put almost no stock in the theory that you beat great offenses by keeping them off the field — in fact, I devoted an entire post to dispelling that theory in my attempt to correct the narrative from the Bills-Giants Super Bowl — but the Saints certainly seem capable of doing just that against the Packers. New Orleans’ three-headed rushing attack and consistent ability to succeed on the ground could prevent significant issues for the Packers in a playoff rematch. And their quarterback is pretty good, too.

  • Vincent Arel-Bundock

    If the ding against yards per carry is that it’s sensitive to huge runs, why not simply use one of the other usual central statistics that are less sensitive to outliers (e.g. geometric mean or median). It seems like the concept of yards per carry is intrinsically interesting and meaningful as a measure of success. A simple adjustment can take care of its sensibility to extreme events. 

  • Chase Stuart

    The median for nearly all running backs is 3.  The geometric mean would probably be an improvement on the arithmetic mean, but is not as easy to understand as a success/failure model.  It’s a good suggestion, because it would remove the extreme events issue; on the other hand, it would be less useful than a success/failure model when discussing short-yardage situations.  There’s no one perfect statistic for running backs, of course, but it’s possible that a combination of the geometric mean and a S/F model would be useful for predictive purposes.

  • Bo Jackson

    Chase, I wonder if you can clear something up for me regarding your opinion that keeping great offenses off the field doesn’t help win games.

    I read your article on the subject and I agree that that in and of itself it doesn’t work. However, wouldn’t the combination of a ball control/clock killing offense and a novel defensive gameplan be effective? By taking up a lot of time on each offensive drive, the total number of drives for both teams is decreased, meaning the opposing offense has fewer opportunities to probe the defense and find a solution to whatever it is they’re doing.

    Thoughts?

  • Chase Stuart

    A couple of things to point out:

    1) It’s important not to conflate “great passing team” with “favorite to win the game.”  There’s a high correlation there — one that is growing every year, it seems — but it’s worthwhile to separate the two.  The 2008 Saints had a great passing attack but were 8-8.  When you’re a favorite playing against a team with a great passing attack, I would argue against any method that both deviates from your optimal strategy on offense and reduces the number of possessions for both teams in a game.

    2) Yes, a ball control/clock-killing offense and a novel defensive gameplan, on the margins, would likely increase your chances of winning as an underdog.  Chris actually ran a post based off one of my comments in an earlier thread on how limiting possessions is a great underdog strategy: http://smartfootball.com/gameplanning/limiting-possessions-key-to-victory.  The Jets did “the stupidest thing ever” against the Pats on SNF by calling a timeout with 1:24 to go near the Pats goal line.  The Jets should have drained the clock, because the Pats ended up scoring a TD before the half. End of half possession optimization is a legitimate underdog strategy.

    3) But if you’re going to change *your* offense because of the other team’s offense, that seems silly to me. By definition, you’re reducing your chances of being successful on offense, because we can assume your normal offense is the offense you believe gives you the best chance to win.  Draining the clock might only mean each team gets 9 possessions instead of 10 or 11, but it doesn’t actually “keep the other team off the field” in any valuable sense because you will still both have the same number of possessions. TOP is highly overrated, because it’s an explanatory stat that’s often connected with winning (like running the ball 30 times). 

    If the goal is to minimize possessions for both teams to increase variance and/or prevent the other team from finding a solution to your innovative defensive strategy, then sure, that’s a sound theory (although you are still likely hurting your offense by going away from your ideal offense).  But I don’t think the media is being as insightful as you’re given them credit for when they say “Run the ball and keep them off the field!”

  • Bo Jackson

    I assure you I am giving no credit to the media when they say “Run the ball and keep them off the field”. That’s not the point I made at all.

  • Chase Stuart

    Understood. I do agree with your general point.

  • Tom A. Hawk

    Great info on the misleading yards per carry stat.  Do you have a similar metric for the QBs/Passing games?

  • Ben_fitzgerald_

    Air yards per attempt?

  • Chase Stuart

    Tom

    For quarterbacks, you don’t have the same issues, IMO.  The sample size is a lot bigger with pass attempts, and the nature of high risk/high reward is pretty well captured in per attempt numbers.

    I like to use net yards per attempt as my predictive stat (which is passing yards minus sack yards lost, divided by pass attempts plus sacks) and adjusted net yards per attempt as my explanatory stat (Passing Yards + 20*TD – 45*INT – SackYardsLost)/(Sacks + Attempts).

  • Tom A. Hawk

    I was thinking more along the lines of an S/F stat for pass attempts.  For instance, if it’s 2nd &10 and a 3 yard pass is completed.  Would that be considered a success or a fail?  That would leave a 3rd & 7 so my guess is fail.  But if the same play gained 8 yards on 2nd down, that would be a pass.  At first, I thought this may highly correlate to completion percentage, but that’s not necessarily so (from initial look at some data I have on hand).

    Also – it seems to me that there are certain receivers that seem to get first downs (the dreaded “possession receiver” label) on many of their receptions and others that may have more receptions and yards, but fewer 1st downs as a % of receptions.

    Could you apply a similar (or adjusted for position) S/F model for these positions?