Author: Michael Canmann (michaelcanmann2024@u.northwestern.edu)

Many clichés get thrown around in football.

“Defense wins championships.”

“Establish the run.”

“The game is won in the trenches.”

At their core, each of these statements attempt to point out a causal relationship between a specific aspect of football that has an outsized effect on outcomes. As the football analytics movement grows, many of these common theories have come into question. Ask anyone in the football analytics community about running backs, and they’ll tell you that the position doesn’t matter. A divide certainly exists amongst NFL teams’ decision makers, with some franchises cycling through running backs on a yearly basis, while others hand out massive contracts to their tailbacks. The answer to this debate has massive significance, as the hard salary cap in the NFL presents teams with an annual constrained optimization problem, as they must seek to maximize team performance with a (largely) fixed amount of money. Thus, teams have a strong incentive to better understand the returns they can generate by investing in various aspects of the game. In this piece, I analyze the relative value of rushing and passing, but also consider the importance of offense compared to defense. 

To assess the degree to which each component of the game influences winning, I broke the analysis down into the four separate areas: passing offense, rushing offense, passing defense, and running defense. Teams’ performance in each was quantified using expected points added (EPA), which measures the change in expected points on a given play based on down, distance, and location on the field. For a more in-depth description of EPA, read here. The general idea behind EPA is to account for context in a way that yards gained does not to provide a better description of the result of a given play. I then regressed the number of wins a team had on their average EPA in each component of the game. The results of the regression are shown below.

Note that the coefficients on defense are negative because EPA allowed is being measured rather than EPA gained, so allowing a higher EPA unsurprisingly has a negative relationship with winning. A few details from this model stand out. First, the returns for a better passing offense and defense are both higher than rushing offense and rushing defense, respectively. This suggests that teams should invest more heavily in their passing game on both sides of the ball than in the running game as an equal gain in passing EPA performance will yield more expected wins. Perhaps teams have started to notice this value difference, which would explain the large appreciation in the price of receivers, such as Davante Adams’s trade and Tyreek Hill’s extension values. 

It is also worth noting that the coefficients on both parts of offensive production have a larger magnitude than their defensive counterparts. While it is possible there is significance here, a restricted model shows that it cannot be ruled out that this difference is just noise. Overall, the key insight from this model is that teams should invest much more heavily in the passing game than the rushing game. 

While the model shows that the passing game contributes more to winning than the rushing game, it looks at each year in isolation. However, front offices must keep an eye on the future and try to establish year-over-year success. Therefore, consistency across years is another important trait when considering the relative value of the four facets of football. The year-over-year correlation of each portion of football was used to analyze this. The correlations were as follows:

Offensive passing: 0.400

Offensive rushing: 0.260

Defensive passing: 0.352

Defensive rushing: 0.185

These correlations reinforce the conclusion that investing in passing success is more critical than rushing success on both offense and defense. Additionally, the correlations for both types of plays on offense are stronger than defense. Though again, they are not necessarily significant. 

While it has not been proven in this analysis, preliminary evidence and other recent football analytics research has suggested that offensive success is indeed more important than defensive success. To dive deeper into this claim, I isolated especially good teams (teams with 12 or more wins) to see if they excel offensively more than defensively. The 12-win threshold was selected arbitrarily, but I determined it was optimal to ensure both a sufficient sample size and a meaningful separation in performance from that of average teams. I used the average ranking, in terms of EPA per play, of teams with at least 12 wins across each of the same four groups to see if these good teams performed especially well in each of them. The rankings are as follows:

Offensive passing: Average ranking of 6.91 out of 32 teams

Offensive rushing: Average ranking of 10.76 out of 32 teams

Defensive passing: Average ranking of 10.12 out of 32 teams

Defensive rushing: Average ranking of 13.39 out of 32 teams

Again, this has reinforced the argument that offensive production is more significant than defensive production as good teams generally rank higher in offensive than defensive performance. While the evidence in this article may not be overwhelmingly in favor of prioritizing offense over defense, the three pieces of analysis all suggest this conclusion in conjunction with the prevailing sentiments of the football analytics community. 

While these results are meaningful, there are two caveats worth noting. First, on each side of the ball, passing and rushing performance are not independent of each other. The opponent must always be prepared for both types of plays, and players at every position impact every play, to varying degrees. Second, while the returns for better performance are linear, this does not mean the cost of talent is. Determining the exact relationship between performance and cost requires further research, but there naturally are far more average players than elite ones. Despite this, this research has yielded two central conclusions — passing performance is more important than rushing on both sides of the ball, and offense is more important than defense. 

Taking this all back to the central purpose of this article — to provide insight into optimal roster construction for front offices — these conclusions suggest that teams should be allocating far more resources into improving their passing performance while also maintaining a preference towards offense when spending. These results are not surprising, as elite passing teams like the Chiefs have dominated the NFL recently, but actually proving the existence of this relationship is nonetheless meaningful and should provide lesser teams with a blueprint for improvement.