Before the Super Bowl, Bill Belichick told his Giants defense to let Thurman Thomas rush for 100 yards.
As David Halberstam writes in Education of a Coach, it was a tough sell before the 1991 Super Bowl against Buffalo. The New York Giants played a physical defense that prided itself on not allowing 100 yard rushers.
No matter, the short, stout coach looked straight into the eyes of Lawrence Taylor and Pepper Johnson and said, “You guys have to believe me. If Thomas runs for a hundred yards, we win this game.”
Just in case his players didn’t listen, Belichick took it upon himself to ensure Thomas got his yards. He took out a defensive lineman and linebacker and replaced these large bodies with two defensive backs. In football lingo, the Giants played a 2-3-6 defense designed to struggle against the run.
Did Bill Belichick go insane? I certainly thought so when I first read this story years ago.
However, analytics is on Belichick’s side. Let me explain.
Visual shows the importance of passing over rushing
When it comes to winning in the NFL, passing is king. Rushing hardly matters.
To quantify this, our football obsessed culture must look past misleading statistics such as rush yards per game. Teams with the lead run the ball to take time off the clock. Any team can rush for 100 yards if they run it 50 times.
To measure true skill, it is better to look at efficiency metrics like yards per attempt. A team can’t fake their way to 5 yards per carry by running the ball more.
Here, efficiency for passing and rushing is defined as yards gained per attempt on offense minus yards allowed per attempt on defense. Higher values indicate more team strength. Sacks count as pass attempts, and these negative yards lower pass efficiency on offense.
The visual shows the pass and rush efficiency during the regular season for all NFL playoff teams from 2003 through 2012.
From the left panel, playoff teams excel in passing, both throwing the ball on offense and preventing the pass on defense. Only 15 of 120 playoff teams in this era allowed more yards per pass attempt than they gained.
The visual also highlights teams that played in the Super Bowl. Eight of the ten Super Bowl champions were among the NFL’s elite in pass efficiency. However, excellence in the air does not guarantee playoff success. The New York Giants in 2007 and Baltimore in 2012 won the Super Bowl despite subpar pass efficiency.
Rushing hardly matters in the NFL
While the importance of passing in the NFL will not surprise anyone, the insignificance of rushing might. The visual for rush efficiency shows playoff teams as a random scatter of positive and negative values for their regular season statistics. A strong run game on offense and defense does not help a team make the playoffs.
Moreover, teams with a high rush efficiency do not suddenly become clutch in the playoffs. Almost half of the teams that played in the Super Bowl allowed more yards per carry than they gained. In 2006, Indianapolis won the Super Bowl while having the worst rush efficiency in the NFL. Green Bay in 2010 and the New York Giants in 2011 weren’t much better.
A guessing game of a team’s wins
Running the ball does not affect winning as much as you think. To illustrate this point, consider this guessing game. Suppose you want to guess how many games a team will win during the regular season. Without any other data, it makes sense to guess 8, the average number of wins in a 16 game season.
From 2003 through 2012, this estimate would be wrong by 3.1 wins. In technical jargon, 3.1 is the standard deviation of actual wins from the guess of 8. In normal people language, it says 2 of 3 teams will be within 3.1 wins of the guess. About two thirds of NFL teams won between 5 and 11 games between 2003 and 2012.
With the rush efficiency for each team, how much better does your guess get? The right panel of the visual below shows how rush efficiency relates to wins for every NFL team from 2003 through 2012. Simple linear regression gives the best fit line through the data.
The regression line gives a new guess about the number of games a team will win. For example, suppose a team has a rush efficiency of 0.6 yards per carry. Instead of guessing 8 wins for this team, the line gives 8.7 wins for this team.
How much better are these new guesses? Not much. The error only drops from 3.1 wins to 3.03 wins. In technical jargon, rush efficiency explains only 4.4% of the variance in wins. You might as well guess randomly.
The results get better using pass efficiency, as shown in the left panel. The error in estimating wins drops from 3.1 to 1.96. Pass efficiency explains 62% of the variance in wins in the NFL. The strong relationship is clear from the visual.
In college football, rush efficiency correlates more strongly with wins than in the NFL. Teams like Alabama, Stanford and Wisconsin have won with a power running game and a physical front seven on defense. The insignificance of running the ball is unique to the NFL.
Analytics gives a broad view of how passing and rushing affect winning. But to dig deeper, let’s look at specific teams and their strengths in these areas.
Indianapolis Colts
Under the leadership of GM Bill Polian and QB Peyton Manning, the Colts had a remarkable run from 2003 through 2010. They won at least 12 games each year before slacking off with 10 wins in 2010.
They achieved success through the air, ranking in the top 8 in pass efficiency each year. Peyton Manning and his offense played the bigger role, but the pass defense helped out some years. The Colts ranked in the top 10 in pass defense (yards allowed per attempt) from 2007 through 2009.
However, Indianapolis was really bad in the run game. Only once in this era (2007) did they gain more yards per carry than they allowed. As mentioned before, they were dead last in the NFL in rush efficiency in 2006 when they beat Chicago in the Super Bowl.
New England Patriots
New England won 125 games, 2 Super Bowls and played in 2 others during the 10 seasons covered by the visual. They followed the same script as Indianapolis: strong in passing, weak in rushing.
From 2003 through 2012, New England ranked in the top 10 in pass efficiency in each year except 2008 and 2012. In 2008, QB Tom Brady got hurt in the first game of the season. New England ended the season 13th in yards gained per pass attempt and did not make the playoffs, the only time this happened during these 10 years.
However, New England has never cracked the top 10 in rush efficiency. Coach Bill Belichick might not have seen the data presented here, but he gets the futility of rushing in the NFL. This understanding extends as far back as his days as defensive coordinator for the Giants.
Indianapolis and New England have built their teams around passing at the expense of rushing. They, along with New Orleans of recent seasons, have had success in winning games and Super Bowls. Now let’s look at teams that excel at rushing.
Minnesota Vikings
More than any other team, the Vikings dominate the ground game. They feature RB Adrian Peterson on offense and have tackles Pat and Kevin Williams clogging up the middle on defense. For the 6 years between 2007 and 2012, Minnesota has finished 1st in rush efficiency 4 of those years.
However, this strength has led to ups and downs in wins. Minnesota went 3-13 in 2011 despite leading the NFL in rush efficiency. The next season, they led the NFL again behind a monster season from Peterson, who made a remarkable return from knee surgery. The Vikings had 10-6 record that season.
The Viking’s best season over this stretch came in 2009. They finished 12th in rush efficiency that season. The difference? A QB named Brett Farve came out of retirement to play for Minnesota. The Vikings finished 7th in yards gained per pass attempt. They went 12-4 and came within a late turnover against New Orleans of playing in the Super Bowl.
San Francisco
The Niners started winning games when coach Jim Harbaugh became coach in 2011. However, they had their strengths before he arrived. Behind DE Justin Smith and LB Patrick Willis, San Francisco had an elite run defense. From 2007 through 2012, they never finished worse than 8th in yards allowed per carry.
This run defense didn’t help them win much the first 4 seasons, as the Niners won only 26 games. The pass defense never finished better than 15th during this time.
When Harbaugh arrived in 2011, San Francisco drafted LB Aldon Smith, a pass rush monster out of Missouri. They also signed CB Carlos Rogers, who had his first Pro Bowl season in 2011. The Niners have finished 9th and 3rd in pass defense in 2011 and 2012 respectively. This resulted in 24 wins during these two seasons.
How to evaluate NFL statistics
In Super Bowl XXV, Bill Belichick’s plan to let Thurman Thomas rush for 100 yards worked, maybe too well. Against a small defense designed to slow down the pass, Thomas ran for 135 yards on 15 carries, a staggering 9 yards per carry. In the second half, he broke off a 31 yard run for a touchdown.
The game ended when Bills kicker Scott Norwood sent a field goal attempt wide right. The Giants won the Super Bowl 20-19.
The Giants did not win the game solely because of Belichick’s defensive plan. The offense generated two long scoring drives in the second half that took time off the clock. And I would bet my life savings Belichick did not want his defense to allow that 31 yard touchdown run to Thomas.
But, as Halberstam discusses in Education of a Coach, Belichick did want the Bills to pick up small gains on the ground if it meant keeping Jim Kelly from throwing the ball. He understood that rushing means almost nothing to winning in the NFL.
If you’re going to remember anything from this article, it should be this: look at a team’s passing instead of rushing numbers to determine whether they will win games.
Some concerns about the rushing data analysis. It may be distorted by late fourth quarter plays. For example, the team with the lead may have taken a knee the last three plays, thereby reducing their rushing efficiency. On defense, the team with the lead may have dropped back 8 players, happy to give up 6-8 yards per rushing attempt because they wanted the c;lock to keep running.
Ken, thanks. Kneel downs should be eliminated from the rushing data. However, I don’t think it will change the conclusions.
As for dropping 8 players, I think that’s a key point. Some defense will do that, and it doesn’t seem to matter in the big picture. Belichick did it this season against Denver when they didn’t have a lead. I think we’ll see it again in the AFC Championship game.
You completely misunderstood why that strategy worked. It didn’t work because it was better, the strategy worked because it was different and the Bills hadn’t seen it. Had the Bills been told “Guess what, the Giants are going to play 5 in the box!” they would’ve scripted every possible double team, kick out, power run and gashed them for 8 yards a play. They were caught completely off guard and still almost won. Just like Bill’s patriots gave up 5 TD’s on 6 plays to the wildcat. As the league gets to know a strategy there is diminishing results. The 2-3-6 is nowhere to be found.
In a regular season or postseason game the team who rushes for the most yards has a better than 60% chance of winning. Team that passes for the most in a game wins less than 50%.
Running efficiency is not real. Every down you can run the ball for positive yardage is a help, unless you are down by 7+. Commitment to the run makes the passing game much easier. As a former D-linemen I can tell you I lived for 3rd and long. I also teach d-linemen the difference between run stopping and pass rushing. Rushing the passer is much easier.
Thanks for the comment. My thinking about run/pass balance is evolving. I do think teams need to run to set up the pass, they just don’t need to be that good at it.
Check out item #3 in this resource guide.
https://thepowerrank.com/top-analytics-articles/
I’m confused as to whether or not you even read the article. You say that if the Bills knew they were going to run dime every play they would have gotten 8 yards per play running… he explicitly says they got 9 yards per rush with their main running back and still lost, so what’s your point.
One of my readers, Jason, pointed out that the 2006 Colts had a much better rush defense in the playoffs than the regular season.
The Colts allowed an NFL worst 5.33 yards per carry during the regular season. However, they improved to 4.13 yards per carry in 4 playoff games. While better, that’s still about the NFL average over the last 10 years.
Nice work!
It would be interesting to see the net rush and pass efficiency numbers of playoff qualifiers over this period when they are in stall situations (as determined by some function of time and score). In these situations we might expect them to run more often and with less success (in terms of yards per rush), while their opponent might run less often and with more success. And they might pass less often and with more success while their opponent passes more often and with less success.
Playoff qualifiers should be in stall situations much more frequently than non-qualifiers, so if any of this is going on it would explain some of the difference in the relationships of rush and pass efficiencies to winning.
Great point. I plan to look at how efficiency changes as a team’s situation changes in the game.
But don’t assume that pass efficiency goes down just because a team passes more (or vice versa). In fact, history suggests the opposite. Check out the charts in this post from Brian Burke at Advanced NFL Stats.
http://www.advancednflstats.com/2010/01/run-pass-balance-historical-analysis.html
Teams ran less as a fraction of plays in the 80’s, but rush efficiency didn’t budge. The percentage of pass plays has been the same the past decade but pass efficiency has gotten better.
An even better way to examine this is % differential. You can get the YPA (Yards per pass) % differential. This shows how team A has performed against its opponents in a stat vs how all other teams have performed against team A’s opponents. This would take strength of schedule into account and normalize things. If not, you are using flat stats and teams have different strength of schedules so not always true enough. Also, with YPA , you can take it further and use Net adjusted passing yards per attempt and then add in a Home bonus or night bonus for the home team. The net YPA removes sack yds lost and the Net adjusted YPA penalizes INTs a certain amount that you apply. I think Net adjusted YPA is pretty good predictor, but in H2H it is only one piece of the puzzle. Home vs Away , Conference, Night vs Day, % differences, elite qb or not,etc. are all important as well. Recent performance is also meaningful along with injuries,etc. ,but yes, Net adjusted YPA is prob one of the best indicators of winning. I like to use that and add in some other important tweaks. Of course rushing is not as important because the yards rushing is much lower than yards passing, but a good run game is nice to have though to compliment the passing game. Also, an even better predictor of winning is points and I would use % difference point differential. I also look at % T/O differential. It never ends.
Your whole post seems to be based on the idea that while winning causes more rushing attempts winning does not cause rushing efficiency to change at all.
I don’t believe this is the case at all and I think it is significantly affecting your results. For the same reason that teams rush more when they have a lead they also rush less efficiently. The defense knows they’re going to run the ball and they play accordingly. The same thing happens with losing teams and passing the ball.
As a preliminary example of the problem look at these two querries on pro-football-reference.com
http://pfref.com/tiny/ZhRWH
http://pfref.com/tiny/KzivF
The first is every rushing attempt in 2013 in the 4th quarter while leading by a TD or more. The second is the same but trailing by a TD or more. The Y/A while leading is 3.12 and the Y/A while losing is 5.04. That is a huge and significant difference.
Thanks for the awesome comment.
Yes, I should really dig deeper into the data. I will when I start looking at this again, hopefully this summer.
The 4th quarter data points out the importance of deception in football. Teams in the lead are expected to rush, hence the low ypc. Teams behind on the scoreboard are expected to throw, hence the large ypc.
However, I don’t think this will affect the primary result that passing is MUCH more important than rushing in the NFL. We could look at 1st half yards per carry and still see little correlation with winning.
Greetings Ed,
Thanks for the article.
What do you make of the fact that Ottis Anderson ran for 102 yards for the Giants and won the Super Bowl MVP in this game?
The Bellichek quote is interesting, but there are far many more coaches who pride themselves defensively in making a team one-dimensional and aim first and foremost to stop the run. Another famous coach (Woody Hayes?) once said something to the effect that when the ball is thrown only three things will happen – and two of those things are bad for the offense.
We’ve bent the rules in an attempt to highlight the QB and make it “a QB’s league”, but for all of the hype, teams with mega passers are often only able to cite stats and not rings at season’s end. Meanwhile, teams with a more balanced attack, effective special teams and solid defense tend to do well.
This is the reason they don’t hire statisticians and actuaries to coach NFL teams. Although stats certainly have their place in the scheme of things, NFL and college football is mostly about television drama.
I am sorry, but passing yards do not win football games. The reason this statistic appears to be exclusive to the NFL is because the NFL pushes the passing game for its dramatic entertainment value. Reliance on passing is simply more prevalent in the NFL than in college. If passing were as important to winning as these stats presume, one would expect the leading passing team in the league each year to win most Super Bowls, or at least a respectable number of them.
Let’s put the theory to the test. The truth is that only five times has the top passing team in the league even made it to the Super Bowl. They are 0-5, losing by an average score of 33-15. The losers? Dan Marino, Kurt Warner, Rich Gannon, Tom Brady, and most recently, Peyton Manning, in that order. No team has ever led the league in passing and won a Super Bowl in the same season. Don’t place all you bets on passing teams and discount the run just yet.
Passing games result in more turnovers per yard gained than the running game, so in that respect the passing game benefits your opponent. It is also more vulnerable to a “prevent” defense. Winning depends upon scoring, not passing yards. Kickers, not receivers or running backs, are usually the leading scorers on most NFL teams and are virtually automatic every time. If a team scores 13 points, 7 of those points are usually made by the kicker.
The main flaw here is rationalizing that yards gained is the most important factor in winning a football game, when in reality, it is yards gained per point scored. From my standpoint, it takes fewer rushing yards to score a point in the NFL than passing yards.
Football is more complex than chess. Yards per point scored automatically blend in stats for defensive efficiency, special teams effectiveness, field possession, turnovers, scoring by the defense and special teams, and many other aspects of the game. I think if you looked at the stats from this perspective, you would discover that gaining offensive yards does not, by itself, win football games. There is a diminishing rate of return in passing yards per point scored that should be considered. This season alone, teams with a huge advantage in offensive passing yards often lost to teams who ran the ball for fewer yards, but scored more points.
Yes, rushing games do take up more time, but “running out the clock” is not the main reason these yards are important. It is because rushing games weaken the defense of the opponent and rests your own defense, both of which pay dividends as the game progresses. The opponent has fewer opportunities to score points and when they do have the opportunity, find it more difficult against a fresh defense, presuming it is a good defense. On the other hand, your offense scores more easily as the opponent’s defense wears down. Offenses can substitute more freely than defenses, so they are able to remain fresh against a tiring defense. Relatively balanced offenses that are more run-oriented are more effective at winning games, especially against strong defenses. Rework your stats using yards per point scored and look at the regression as a curve (maybe using a fifth or sixth degree polynomial) rather than linear, and you’ll see that the win – loss ratio isn’t just about offense and isn’t just about passing yards. I’m not sure I trust the linear regression results as having quite the predictive capability you suggest. What I see is a lot more uniformity with regard to passing than running in the league, that’s all.
Give me a team with a great defense, great Special Teams, and a great running game any day over a team whose only goals seems to be to pad the QB’s passing stats. Look how many NFL teams have started using that formula since the Seahawks found success with it, and how difficult it has been to predict game outcomes this year. Complete teams have a better chance of winning than a one dimensional team who puts up a ton of passing yards.
The ideal algorithm will never exist, and even looking at yards per point scored won’t produce results that are that much different from your stats. It will account for some predictions that turn out better, while others turn out to be less accurate, because some things simply cannot be foreseen or factored into the equation, and that creates the drama and excitement we all love.
Maybe you missed the part where he said that being excellent at passing is no guarantee to win the Super Bowl. Ironic that a reader on this site doesn’t understand the effects of randomness and thinks that for a stat to be most important it needs to be 100% correlated with winning the SB.
I agree passing yards have no correlation with winning any individual game. Indeed, the two conditions that connect, in any individual game, to a win or loss are a) running yards (whether cause or effect) and b) INTs. Overcoming both of these, in any individual contest (whether cause or effect) is unlikely, my guess is in the 5% range (assuming a team “loses” in both comparisons).
But the interesting thing is that you can sort team quality, year after year, by net YPPA, that is, looking to the team’s offensive YPPA and defensive YPPA and ordering the deltas.
I think Pete Palmer stumbled onto this first (as a footnote re: YPPA) and Paul Zimmerman noted the INT (correlates with wins)/Fumbles lost (not much, if any, correlation with wins)—-but Bud Goode studied it for a number of years before his demise.
In the NFL over the past few decades, the quality of the quarterback has determined a team’s ability to win more than any other factor. The influence of the QB had become overwhelming. Then we had last season, when the QB matchup was practically irrelevant to the Super Bowl outcome.
Pro football works that way. Popular trends (mobile QB, West Coast, possession pass, whatever) become widely adopted over many seasons, then somebody exploits the league-wide bias with a starkly different strategy. Last season’s SB teams (Denver, Carolina) overinvested in defense – from a personnel mgmt perspective – while most of the league was obsessed with the quest for the franchise QB/WR tandem.
That is the thing that sticks out to me most. Many NFL teams would rather overinvest in a questionable QB than optimize ROI and build a strength. The popular view was that winning without a franchise QB was impossible. That apparently justified the approach of playing QB roulette until the longshot hits.
It’s as though NFL teams are run by PR firms. GMs don’t want to face marketing the team without having a “name” at QB, even that name is a rookie who never even won a major bowl game in college.
Sometimes, good rushing teams in the NFL get there by philosophy. Sometimes, they get there because the QB situation is poor, while good runners are plentiful. Rushing stats also can pile up if a team starts a lot of possessions on the 25 rather than in good field position. High-possession games, long field, a lot of rushing yards.
Yards per carry = a stat influenced heavily by two factors. One is down/distance situation. Teams that get a lot of their rushes in the red zone, or in other short-yardage situations, naturally average fewer yards per rush. I like to look at yards gained as a % of yards to first down or TD. Does not strip out all the difficulty factor of running into a prepared defense vs. rushing primarily as a changeup, but it helps.
The other factor is offensive design. An offense that uses the rush to keep the defense honest naturally averages higher yards per carry. Fewer running attempts, defense spread out not stacked up, and the rusher is going to be quick/fast, plays designed to enable the back to beat a couple of defenders instead of trying to find a seam in a pile.
One thing you’ll probably find is that overinvestment in a talent-rich position (RB for example) is likely to have a poor ROI. How often does it pay to draft RBs in the first round? Does it make sense to carry the salary of a superstar rusher? This factor might help to explain why the team with the greatest RB never wins it all anymore (if ever). Paying too much to one player shrinks the blanket that’s supposed to cover the rest of the bed.
Is it possible to write algorithms that capture all the important differences between NFL teams?
It’s not possible to capture everything about NFL teams quantitatively. That’s what makes football great!
While I enjoyed reading this piece if only now in 2017 I feel the need to point out the Giants rushed 39 times for 172 yards and held the ball for over 40 minutes in the Super Bowl anecdote you used. Running the football is what arguably won the game for the Giants as the respective yards per pass attempt was nearly identical