The Curious Case of Correlation Between Pass and Rush Offense

Kansas City Chiefs OL Mitchell Schwartz
Kansas City Chiefs OL Mitchell Schwartz
Photo: USA Today Sports Images

Guest column by Lau Sze Yui

We all know the football analytics talking points: "Running backs don't matter," or "establishing the run is meaningless." But the football analytics community has tended to overlook questions about the correlation of passing and rushing offense that might change the way we see the running game. A better understanding of why pass offense and run offense correlate would lead us to a better understanding of NFL analytics in general.

To demonstrate the correlation, here is a plot of passing vs. rushing DVOA from 2015-2019. There is a pretty significant trend.

Chart 1

What if there is bias in DVOA adjusting for opponent strength and situation? A similar (but slightly weaker) trend can be observed using EPA (from nflfastR) as well.

Chart 2

Here are a few hypotheses as to why this correlation is happening:

Hypothesis 1: Play-action usage leads to better passing efficiency while running the ball more.

For the correlation between passing and rushing offense, the first thing that naturally comes to mind is that teams with the better run game would "sell" the play-action better, and hence better play-action efficiency would lead to an overall increase in passing efficiency.

However, in this piece on play-action passing, Ben Baldwin shows that play-action and non-play-action passing efficiency gaps don't change with either the frequency or the efficiency of running the ball.

Chart 3

Therefore, rushing has no relationship with "selling" the play fake.

Hypothesis 2: Defenses will have more defenders in the box against a good run offense and hence will give up more yards to passing.

One of the best ways to stop a running game is to send more defenders into the box, but that may hamper pass defense performance. In 2019, for example, the Kansas City Chiefs led by Patrick Mahomes faced the fewest average defenders in the box while having the best passing efficiency. The graph below also shows that when there are more defenders in the box, passing efficiency increases and rushing efficiency decreases as expected.

Chart 4Chart 5

However, on a team level, this trend just vanishes:

Chart 6Chart 7

As a result, the hypothesis that it is the number of defenders in the box that leads to correlation between passing and rushing offense is not supported.

Hypothesis 3: Offensive linemen who are good at pass-blocking are also better at run-blocking and boost both passing and rushing offenses.

It is straightforward thinking that offensive lineman can affect both passing and rushing. However, it is hard to prove with evidence, since the charting of offensive linemen by various companies does not go very far and may not be adjusting correctly for team-level performance.

By using historical sack rate and DVOA data, it is possible that we can build a proxy of offensive line strength and compare it to passing and rushing efficiency.

Chart 8Chart 9

Unsurprisingly, sack rate is more strongly correlated to passing offense than to the running game. To give a comparison, passing offense efficiency can be roughly divided into sack rate, interception rate, completion rate, and yards per completion. Here is the correlation of those other three metrics to team rushing DVOA:

Chart 10Chart 11Chart 12

As shown above, a positive passing metric (i.e., higher yards per completion, higher completion percentage, and lower interception rate) leads to a better rushing offense. However, the data span across more than three decades and the variance of sack rate is larger due to generational differences. Therefore an alternative approach is to calculate the single-season correlation between sack rate and rushing DVOA so we can see how it evolves over time.

Chart 13

The R^2 coefficient is about 0.05, which is pretty mild. As a comparison, here is the plot with other passing metrics:

Chart 14

Looking at all four metrics, a perhaps surprising result is that interception rate actually has the highest correlation to rushing DVOA, while sack rate has the lowest. Still, a low correlation of sack rate to rushing DVOA does not imply that offensive linemen don't matter since a good blocking lineman can decrease interceptions by allowing less pressure to the quarterback.

To conclude all the points above, neither play-action nor defender count seems to be the culprit explaining why passing offense is correlated to rushing offense. Rushing offense seems to affect all facets of the passing game and sacks actually have the lowest correlation to rushing DVOA. Therefore, the best conclusion so far seems to be "passing and rushing are correlated because both are based on the performance of offensive linemen," but it's still not exactly clear and needs further work in the future.

Lau Sze Yui studies sports analytics in his spare time and can be found on Twitter @903124S.


41 comments, Last at 31 Aug 2020, 10:07am

#1 by Aaron Brooks G… // Aug 21, 2020 - 11:45am

Sacks are a QB stat more than a line stat. I'm inclined to think pressure rate is a better line stat, and probably better something like pressure rate/mean time to pass -- which better handles the Manning types on one end who throw instantly to bail out a bad line and the Wilson types who scramble forever until pressure eventually arrives.

What I think you're seeing is a reversible function -- better rushing helps the passing game, and vice-versa. If the defense has to account for one part of the game, it eases the job the other has to do, because the opponent isn't prioritizing it. But also, one good component puts the other in better down-and-distance situations -- 2nd-1 is easier to exploit than 3rd-17.

But some of this is just good teams do well and poor teams do not.

Points: 0

#2 by theslothook // Aug 21, 2020 - 1:29pm

I would be curious to see what the correlation is between pass dvoa and run dvoa over a large span of time than the last 3 years. 

There's also a lot of endogeneity baked into DVOA that teasing out cause and effect is going to be very hard to do. 

Rush DVOA is going to reward teams that convert a lot of 3rd downs and goal line plays. Good offenses will naturally select run plays based on certain down and distances and defensive looks while poor offenses are forced to run the ball in any and all circumstances. That might suggest that Team  A's run offense is better than Team B's run offense, but that's not exactly what it will be showing. Instead it will say Team A's offense had better opportunities than Team B's and the run metrics are measuring net of opportunity, not raw quality. 

Secondly, what do you do with this information? If the answer is to build a good pass offense, well that was a given anyways. If you are going to try and quantify how much your run offense will improve with a good pass offense, that's hard still because you don't have an objective measure of your run offense as is net of opportunities. 

Points: 0

#3 by QuiGonJim // Aug 21, 2020 - 1:30pm

My instinct makes me think if you take the average distance between starting position of defenders you would see a correlation with Rushing/Passing DVOA (the farther apart the defenders, the higher the rush/pass DVOA). This is perhaps the opposite of the play action hypothesis. I would think a good Passing Offense spreads the defense and creates a good rushing offense more often than a Rush Offense creating a good Passing Offense.

My hypothesis would be that if a defense is forced to play a capable QB, they spread themselves out a little more to not get beat deep, which creates room for the Rush Offense. I think there are multiple pieces to what I'm saying, so you'd first need to see if there were any correlation between spread of defense and rush/pass DVOA. If there is, my you could then look at correlation between QB DVOA and spread of defense. Perhaps look at formations vs spread of d and line movement vs spread of defense as well. 


This may all be wrong, I'd just be curious to see. 


Points: 0

#5 by Aaron Brooks G… // Aug 21, 2020 - 2:05pm

This is perhaps the opposite of the play action hypothesis. I would think a good Passing Offense spreads the defense and creates a good rushing offense more often than a Rush Offense creating a good Passing Offense.

I think it cuts both ways.

Flexbone and wishbone teams pass very little, but are pretty efficient when they do. They are kind of helpless on 3rd-10, though.

Points: 0

#6 by QuiGonJim // Aug 21, 2020 - 2:50pm

My theory for why it doesn't cut both ways is that from a defensive standpoint stacking the box to handle a below replacement level QB/above replacement level RB has a higher rate of success than spreading the field on defense to contain an above replacement level QB/below replacement level RB.


What comes to mind most recently is the Kansas City Chiefs with the Mahomes/Damien Williams, again this is just based on instinct/watching the game as opposed to having actual Data to back it up...

Points: 0

#8 by Aaron Brooks G… // Aug 21, 2020 - 4:21pm

Do you think 2012 Joe Webb was actually a 0 DVOA QB, or do you think he just looked that way because Adrian Peterson was able to run iso plays into 10-man boxes for 2000 yards?

Do you think Tim Tebow throws for 316 yards against the #3 DVOA pass defense if they are playing a nickel rather than stacking the box?

Points: 0

#11 by QuiGonJim // Aug 21, 2020 - 5:23pm

I guess I’m not sure what you’re asking here. In terms of Joe Webb/AP/2012 Vikings, their overall offensive DVOA was 0.3% and ranked 15th in the league according to FO. That puts them right in the middle of the road, and their 10-6 record makes sense? Plus they lost in the wildcard. In terms of Tebow, I suggested based on probability, which takes into account the law of large game where my theory wouldn’t hold true certainly doesn’t disprove my theory... 


I think I was trying to go one step beyond correlation to find some sort of causation for the correlation between pass DVOA and rush DVOA. I think my original theory is simply that the correlation between average distance between defensive player and offensive DVOA.  At that point it becomes about finding a solution on how to spread out the defense...




Points: 0

#38 by Noahrk // Aug 25, 2020 - 12:23pm

I think you are 100% correct. I mean, as a Dolphin fan I know without a doubt Damien W. is not a good RB running the ball (though he is a good receiving back). Yet the KC passing game gives him enough space that he can perform very well.

The hypothesis that linemen who are good at pass blocking are good at run blocking and viceversa is patently false. There are plenty of counter examples.


Points: 0

#39 by Noahrk // Aug 25, 2020 - 12:34pm

Maybe hypothesis three can be restated so that it makes more sense: offensive linemen who are good at pass blocking find it easier to run block when defenders approach the play as a passing down, giving the OL better run blocking angles.

Points: 0

#4 by Paydro // Aug 21, 2020 - 1:56pm

I think a thorough breakdown of the relationship between passing and rushing DVOA is needed before attempts to explain it with correlations to individual statistics. Intuitively, we would expect the relationship to have multiple causes: offensive playcalling, defensive playcalling, individual player responses, overlapping player skills, etc. 

Those causes might not have the same effect at different levels of performance, which muddies the waters.  If I'm a defensive coordinator playing a team that's atrocious at passing but merely "bad" running, I may not react the same way (rationally or no) when faced with a similar gap between an "average" passing offense and a "very good" rushing offense.

Really I'm much more interested in the final graph, as it suggests the strength of the relationship is increasing on a variety of metrics.

Points: 0

#7 by Lebo // Aug 21, 2020 - 3:14pm

As one of the other posters alluded to, play calling matters. Perhaps rushing and passing DVOA are positively correlated because good play callers call plays that are more likely to be successful, whether they are runs or passes, and bad play callers call plays that are less likely to be successful, whether they are runs or passes.

Points: 0

#10 by Joseph // Aug 21, 2020 - 5:05pm

Playcalling matters very little. How many highlight clips of Barry Sanders show him making a great play out of a bad situation? (Of course, give the ball to Barry was always a good playcall!) But honestly, if the coach knew it was a bad playcall, why did he call that play? Why didn't the QB audible when he saw it wasn't going to work?

My hypothesis: talent matters. (Of course, this is high-level insight LOL.) If you have a good passing QB, esp. in the last few years, you run 21 personnel, and teams cannot stack the box--lest you carve them to shreds. This allows your good QB to audible to a run, and you are more likely to have a successful run play, thus higher rushing DVOA. If you have a talented RB (or Lamar Jackson), the opposite is true--you can carve the D up with play-action, or even a non-PA pass out of a more traditional run formation. (Think 2019 Titans with Tannehill and Henry.)

However, this is definitely a feedback loop--good offenses are by definition more successful, and put themselves in favorable down & distance. Favorable down & distance situations on 2nd and 3rd down begets higher DVOA/success rate/etc. Favorable down & distance also creates a better situation for the offense, as they can be less predictable, making their chance for success higher--the defense can't stop everything.

They also have talented players that are able to make something out of nothing, make contested catches, break tackles, outrun defenders, etc. So talent gives them a better chance of success, which puts them in favorable circumstances more often, which begets more first downs, giving them more chances to succeed, put points on the board, and place themselves in even MORE favorable circumstances.

Overall, good article. Nice use of graphs, data, and potential hypotheses to explain why. But in the end, talent is probably responsible for 80-90% of the difference.


Points: 0

#13 by theslothook // Aug 21, 2020 - 6:50pm

Playcalling is an extension of coaching and that does matter. Brees is successful in this offense, not necessarily in one that attempts to jam him into a style he isn't suited for.


Playcalling is a function of talent on hand but also the coach's principles. The two go hand in hand, 

Points: 0

#25 by Joseph // Aug 22, 2020 - 5:21pm

I don't disagree--but, most successful QB's would be similarly successful in other types of offenses. Obviously, a good coach/playcaller will do his best to choose plays that take advantage of his players' talents, should work against the defense that week, etc.

One name that comes to mind is Brandon Marshall--he put up solid numbers with a number of different QBs/teams. Similarly, Ryan Fitzpatrick seems to have a reasonable level of play, and has played for numerous teams. That is a reflection of their talents, independent of their coaches.

Good coaching can enhance a player's performance by giving them opportunities to succeed based upon the schools they have; bad coaching can do the opposite. It is much easier to quantify players' successes and failures versus the coaches. For example, the OC calls a play, the QB throws a perfect pass, and the WR drops it. Even the most sure-handed WR's drop one occasionally. But a handful of blown blocks, dropped passes, bad passes, stumbles, etc. can make an entire offense look bad, and the OC gets fired. If we assume 95% success of every player on every snap=33 blown assigments per game (60 offensive plays * 11 players * . 05 error rate=33 errors) The playcaller's success is basically judged on whether those errors committed by players will sabotage a play, or whether another player will be able to overcome that error by his teammate (break a tackle, make a man miss, catch a wayward pass).

[Obviously, I presume that there is a pretty high level of success; some teams have more plays per game; changing the error rate affects the total. But to achieve an error on 50% of plays=95.45 success rate on every assignment.] If your team messes up 1/2 of its plays, that's bad. If teammates overcome 1/2 of those errors, or some plays have multiple errors (meaning more plays have none), you might execute the play exactly right 75% of the time. But that still doesn't mean it is a successful play (as defined by FO's success rate/DVOA). That also doesn't take into account how bad those errors are--is the mistake an incomplete pass, or a turnover? A sack/tackle for loss, or a 2 yard gain that could easily have been 5-7 yards if the block was executed? Nor does it take into account when the defense makes a great individual play (like batting down a pass at the line, even though the O-line gave the QB a clean pocket). 

In other words, all of these variables affect how we evaluate coaches--and really, they have nothing to do with the coach; they have to do with the talents and execution of the players on the field. Yes coaching matters--but talent greatly affects how we evaluate those coaches.

Points: 0

#32 by theslothook // Aug 23, 2020 - 12:47pm

For every Ryan Fitzpatrick story, there is a Mike Singletary to Jim Harbaugh story. Ben Muth described the Jets o line debacle as a coaching debacle because everyone was making communication mistakes and this was a group of all veteran lineman.

For every one perfect pass threaded to a receiver who drops it, there are 10 other routine plays that require basic competence from all the other non stars to do their job to their best abilities. That is also down to coaching.

I'm also not convinced Drew Brees or Tom Brady become great quarterbacks on every team. Or even most teams. I'm not convinced Aaron Rodgers doesn't become a bust if he's drafted by the 49ers.

Points: 0

#37 by Joseph // Aug 24, 2020 - 1:33pm

For sure--and there are other examples that you could cite as well.

I guess we could probably agree on this--any player who even makes a training camp in the NFL has some baseline of talent. There are some whose talent is greater, which allows them to do things that other players cannot (because they are faster/bigger/etc). The coaches' job is to get 11 (at a time) guys' individual talents to work together as a team, while using the unique talents by each guy to the maximum benefit of the team as a whole. Some coaches are better/more imaginative/etc. than others, and some get to use more talented players than others.

However, in my opinion, for what that's worth, there are some talented individuals whose talent can overcome bad coaching, and there are some good coaches that can maximize the team output from guys whose talent is not as great or obvious. Where those lines are is not readily obvious/measureable.

Points: 0

#16 by 903124 // Aug 21, 2020 - 9:28pm

If you have a talented RB (or Lamar Jackson), the opposite is true--you can carve the D up with play-action, or even a non-PA pass out of a more traditional run formation. (Think 2019 Titans with Tannehill and Henry.)

We now know that play-action minus non play-action efficiency is highly dependent on shotgun vs under center. Lamar actually perform slightly worse in play-action plays.

However, this is definitely a feedback loop--good offenses are by definition more successful, and put themselves in favorable down & distance. Favorable down & distance situations on 2nd and 3rd down begets higher DVOA/success rate/etc. Favorable down & distance also creates a better situation for the offense, as they can be less predictable, making their chance for success higher--the defense can't stop everything.

 Here I limit to 1st&10 and the same issue still exist. Obviously playing ahead vs behind still matter though but there is zero correlation between offensive and defensive efficiency

Points: 0

#27 by Joseph // Aug 22, 2020 - 6:14pm

Lau, only the graphs in the 2nd hypothesis say that they are limited to 1st and 10. While I think that limiting your analyses to 1st and 10 is wise to control for a lot of variables regarding the number of men in the box, it doesn't matter HOW you achieve success on 1st down.

IMO, that's why on the team level, that trend vanishes. KC with Mahomes can be successful with a pass no matter how many men are in the box, and the Ravens with Jackson can be successful with a run regardless of the number of men in the box too. They succeed in general, no matter what the defense does. Conversely, bad offenses fail in general, regardless of the number of players in the box or what the defense does.

Why? Because the talent on those offenses matters--they tend to succeed overall, pass or run; whatever the defense tries over the course of the game will not be good enough to stop them in the long term. The Texans, Titans, and 49ers limited the Chiefs for parts of their postseason games--enough to get a double digit lead! But it wasn't good enough in the end. 

I want to reiterate that your article seems well-researched, well-reasoned, and well presented. I think the obvious conclusion is: talent matters a lot, and coaching matters some too. We can debate how much each one matters, and luck sometimes plays a part in a play and even the result of the game. But in the end, the more talented team wins the game most of the time.

Points: 0

#21 by Lebo // Aug 22, 2020 - 4:11am

DVOA controls for down and distance. So, favourable (or unfavourable) down and distance won't affect DVOA.

Points: 0

#26 by Joseph // Aug 22, 2020 - 5:44pm

Correct. My point is that successful plays (good DVOA) make it easier to continue to achieve more successful plays (good DVOA). Negative yardage plays (bad DVOA) make it harder to get good DVOA on subsequent snaps.

For example, 1st and 10 turns to 2nd and 6 after a 4 yard gain. This is slightly negative in DVOA, but still gives the offense a reasonable chance at a 1st down. A 4 yard gain on 2nd down would be positive DVOA, and a very good chance at a 1st down (positive DVOA again.) However, and incomplete pass on 1st down means that the offense needs 6 yards on 2nd down to get positive DVOA, and the offense has less than a 50% chance of picking up the first down. A negative yardage play on 1st or 2nd down makes it even tougher to achieve a positive DVOA total for those 3 plays (I am ignoring 4th down here).

So, while each play's DVOA is controlled for down and distance, positive or even slightly negative DVOA on the previous play(s) creates easier opportunity to achieve more positive DVOA on the following snap--and negative DVOA makes it harder. While better team offensive DVOA is quite obviously indicative of better offense in general, my point overall is that most of the time, good offenses are good at both running and passing in DVOA, because the good passing game gives the run game opportunities for short yardage success (it's easier to get positive DVOA on a 3rd and 1 run versus 3rd and 4). Similarly, a 6 yard run on 1st and 10 makes it easier to succeed on that 2nd & 4 pass. Also, the defense who has to commit that extra man in the box to stop the run game is more susceptible to giving up a successful pass, and vice-versa.

Points: 0

#29 by David // Aug 23, 2020 - 4:36am

In reply to by Joseph

You're confusing success rate with DVOA.

positive or even slightly negative DVOA on the previous play(s) creates easier opportunity to achieve more positive DVOA on the following snap

No - because the DVOA of the subsequent snap is measured against what other teams did in the same situation. If it was true that positive DVOA on snap n led to an easier opportunity to achieve more positive DVOA on snap n+1 then this would be true for all teams, raising the benchmark, and nullifying the increase.

What you have expressed MAY be true for success rate (a successful play on snap makes it easier to have a successful play on snap n+1) ​​but is not true for DVOA by design

Points: 0

#36 by Joseph // Aug 24, 2020 - 1:23pm

Yes I am, and no I am not. For one, success rate and DVOA are similar, in that they measure how well a team is successful moving down the field and stopping their opponent from doing the same. On the other hand, success rate is a yes/no question, whereas DVOA measures how much success was attained/prevented.

You are correct in that a successful play on snap n makes it easier to have a successful play on n + 1; but when I say that it creates an easier opportunity to get successful DVOA, maybe I should have clarified. Since there are bonuses for first downs & TD's, those bonuses are easier to obtain. Also, a successful play (or nearly successful) as measured by DVOA or success rate allows the coach/playcaller to have more freedom of play choice/formation/etc. For example, in a neutral game setting, 2nd & 12 should always be a pass play of some type--11 personnel would be quite probable. Either the TE or RB might even line up in a slot position versus in the backfield/next to the tackle. Conversely, 2nd and 5 could be any type of pass or run play--even a 3 yard run would be "successful" by both metrics by FO's standards. This is where I am saying that a successful play (by either measure) makes it easier to attain success by either measure on the subsequent down. That 3 yard run is quite unsuccessful on 2nd and 12, but a success/positive play on 2nd and 5. Considering that the average rush in the NFL is around 4.25 yds, 3 yds is below average overall, but positive/successful in all situations considered "short yardage," and still somewhat successful in 1st or 2nd down with 5-6 yards for a first down or touchdown (1st and goal/1st and 5 after a penalty, not 1st and 10).

So, my point is that the baseline yardage for success/positive DVOA is lower after a successful play on 1st and 10--which makes it easier to achieve. Now, I also realize that some of the charts in the article only cover 1st and 10 situations, so my point does not apply there. But let's be honest--over the course of a game, the winning team is almost always the team who is successful on more of their plays than their opponent. Most often, these are small advantages per play that add up throughout the game. DVOA measures this better than success rate generally also.

As I have stated in other comments, talent and coaching play a part in those successes, and the percentages each one plays can be debated and are hard to measure precisely.


Points: 0

#9 by Pat // Aug 21, 2020 - 4:40pm

Why wouldn't you *expect* EPA on a team level to lose its correlation with number of guys in the box? Teams aren't going to choose a dumb tactic on purpose. I mean, in week 1, a team's not going to come out with 8 guys in the box and say "let's go ahead and *make* Patrick Mahomes force us out of the box."

Likewise, when the Jets were trotting out Random QB of the Week with Darnold out, teams aren't gonna drop everyone into coverage until that QB forces them to, and, uh, he never did.



Points: 0

#15 by 903124 // Aug 21, 2020 - 9:21pm

There is some relationship here as Mahomes Chiefs face by some margin least number of defender in box. But at the same time Kirk Cousins's Vikings and Lamar's Ravens face lots of defender in box and they are good passing team as well.

Points: 0

#23 by Pat // Aug 22, 2020 - 9:21am

I don't think those are necessarily counterexamples: the question isn't how many defenders they face, it's the *range*. In order to see a correlation, you need a span of values in the independent variable. Without it, the correlation goes away.

In other words, the question isn't "if the Ravens are a good passing team why are there so many defenders in the box?" It's "how good *would* the Ravens passing game be if they *didn't* have so many in the box?"

Sadly, defenses don't just randomly try bad ideas to see what will happen, so we can't know.

That's why I said you'd expect it to go away: defenses shift strategy against different teams, so a basic correlation shows up, but against single teams I think they tend to stick to one, although there are probably exceptions. In those cases though the defense is probably floundering and sucking in general.

Points: 0

#12 by andrewmilne // Aug 21, 2020 - 5:26pm

(1) Is there a (long-term) correlation between being good at offense and the percentage of the cap spent on offense? (Maybe teams that invest in offense are better at both run and pass.)

(2) I am guessing that the correlation is between season running DVOA and season passing DVOA? I am wondering if the answer could be in the game-to-game variability. Good offensive teams are good at beating up on the other defense's weaknesses, and rack up good ratings whatever. Bad teams can't beat anyone no matter what they try...

Points: 0

#14 by Dan // Aug 21, 2020 - 8:17pm

I'd expect the correlation to primarily come from how defenses try to stop the offense.

If an offense is much better at running than at passing, then the defense is going to put bigger guys on the field, line up with more men in the box, call plays that are sounder against the run, instruct LBs to focus more on the run, and so on, compared to what they'd do against a more balanced offense in the same situation (same down, distance, field position, offensive personnel grouping, offensive formation). Whereas if the offense is much better at passing than running, the defense will focus more on pass rush & coverage rather than stopping the run, field more DBs & pass rushers (rather than LBs & run stuffers), put fewer guys in the box, make more exotic play calls to generate pressure or disguise coverage (even at the expense of soundness against the run), etc.

So if a team has average skills at running the ball and an awful passing offense, they'll wind up with below average rushing efficiency because defenses can focus on stopping the run. Whereas if a team has average rushing skills and a great passing offense, they'll wind up with better than average rushing efficiency because defenses have to focus on stopping the pass. And that's a positive correlation between passing efficiency & rushing efficiency.

The raw correlation with # of defenders in the box might not capture this that well because 1) a lot of the variation in the # of men in the box is due to offensive personnel & formation & game situation rather than the defense responding to the relative strength of the offense's passing & rushing games, and 2) the defense's response involves a bunch of other changes besides # of men in the box.

Points: 0

#19 by 903124 // Aug 22, 2020 - 1:42am

For instance, the correlation between the number of linebackers and rushing EPA is pretty weak

Points: 0

#17 by zenbitz // Aug 21, 2020 - 10:25pm

The answer is more straight forward, although I am not sure how to test it.

If you are good at both running and passing, then you have more tactical options.  That means that the defense (no matter how many defenders in the box) will have to guess more, making both options slightly better.

Isnt there an autocorrelation with offense?  Like if you fail on 1st and 2nd down you are in a tough spot on third.  This you could maybe detect by looking at Rush vs Pass EPA/P only on 1st and 10

I am a little disappointed that the initial graphs don't have R-sq and p-value though ;)

Points: 0

#20 by 903124 // Aug 22, 2020 - 1:47am

The analysis is done on 1st and 10 only, also the R-sqaure value for passing and rushing offense DVOA is 0.136 and p-value is <2.2e-16 but it's not very important if you can see the trend clearly on graph

Points: 0

#24 by zenbitz // Aug 22, 2020 - 4:15pm

Thanks! Not sure how I missed it was 1st and 10 only.  Does the correlation get stronger or weaker if you include all downs?

Points: 0

#18 by Pen // Aug 21, 2020 - 10:37pm

So very tiny sample size, but here's what we see with Russell Wilson at QB.

Lynch years

2012 536 rushes 2579 yds.  Wilson passer rating 100

2013 509 rushes 2188 yds.  Wilson passer rating 101.2

2014 525 rushes 2762 yds.  Wilson passer rating 95.0

2015 500 rushes 2268 yds.  Wilson passer rating 110.1

Lynch gone

2016 403 rushes 1591 yds.  Wilson passer rating 92.6

2017 409 rushes 1629 yds.  Wilson passer rating 95.4

Chris Carson years

2018 534 rushes 2560 yds.  Wilson passer rating 110.9

2019 481 rushes 2200 yds.  Wilson passer rating 106.3


So is it the skill of the RB lifting the passing game?  If they fear the RB, perhaps that opens things up a bit?



Points: 0

#22 by guga31bb // Aug 22, 2020 - 8:35am

And notably missed the entirety of Wilson's hot stretch in the second half of the season. He appeared in 7 games and had about 400 rushing yards. Attributing Wilson's performance this year to Lynch seems like a stretch.

Points: 0

#33 by Pen // Aug 23, 2020 - 10:54pm

I specifically bolded 2016 and 2017.  When there was neither Lynch nor Carson.    You seemed to have missed my point entirely.

Points: 0

#28 by MJK // Aug 22, 2020 - 8:17pm

It seems that no one is considering a very obvious answer why run and pass offense should be correlated: coaching and front office.

Good front offices will be better at assembling talented offenses, which will be better at running and at passing because they're more talented, while bad front offices will, generally, put less talent on the field and therefore be worse overall, at both running and passing.

Good coaches will have their players better prepared, come up with better gameplans to exploit opposing teams weaknesses, and make better in-game adjustments, and therefore will do better at both running and passing.  Bad coaching staffs will allow their teams to play sloppy, be poorly prepared, and stick to rigid gameplans even when they're not working... leading to worse running and worse passing.

In general, when two variables are correlated, it's either because A causes B, B causes A, or some other factor C causes both A and B.  Coaching and overall talent assembled seem like a pretty likely C.

Points: 0

#31 by 903124 // Aug 23, 2020 - 8:28am

I would not assume that front office would play a huge role since there is virtually zero correlation between offense and defense unless you assume that ability to build an offense is completely unrelated on the defensive side

Points: 0

#34 by takeleavebelieve // Aug 24, 2020 - 11:16am

Most teams’ offenses and defenses have separate coordinators, separate assistant coaches, and potentially even separate scouting departments. It seems reasonable that performance wouldn’t correlate.

Points: 0

#30 by David // Aug 23, 2020 - 4:45am

Hypothesis 2 is still unknown. The graphs for the per team numbers are calculated differently to the league wide numbers (actual men in the box versus average men in the box )

The original hypothesis (that whether run or pass is more efficient is dictated by the defense) is strongly supported (a reference to the original article would be good) and stating that it's unsupported by slicing the data and misrepresenting the result is weak. 

Points: 0

#35 by takeleavebelieve // Aug 24, 2020 - 11:21am

Of the 3 hypotheses, I think the offensive line has merit. Many previous studies have proven that QBs perform significantly better when kept clean, and that YBC is strongly correlated to rushing performance. 

I also think there’s merit to the suggestion that coaching/playcalling plays a role. For instance, motion correlates to rushing success, and that’s entirely a coaching decision.  

Points: 0

#40 by Sugarrush // Aug 30, 2020 - 10:29pm

Heres an excerpt from an article on sharpfootballanalysis regarding modern defenses:

Last year, the most advantageous matchup for an offense to run the ball came from 11 personnel against dime defenses, per data from Sports Info Solutions. There were just below 600 such plays last season (598) which accounted for 10% of runs from 11 personnel. But on those plays, runs had an EPA per carry of 0.18 with a positive play rate of 51%. Runs from 11 against nickel (five defensive backs) averaged 0.00 EPA per carry with a 49% positive play rate. The average run across the NFL had -0.03 EPA per carry and 47% positive play rate in 2019. Offenses gained the biggest advantage running when the defense sold out for the pass. 

The opposite holds true for the pass. Some of the most efficient passing plays last season came from 12 personnel (two tight ends) when the defense played base with just four defensive backs (0.10 EPA per attempt. 51% positive play rate)

So maybe it is depending on how a defense reacts to a good pass offense

Points: 0

#41 by ImNewAroundThe… // Aug 31, 2020 - 10:07am

Get a good passing offense. Then worried about running, if you want/can

Points: 0

Save 10%
& Support the Writers
Support Football Outsiders' independent media. Use promo code WRITERS to save 10% on any FO+ membership and give half the cost of your membership to tip the team of writers.