## Passing is more important than running

Since I had created a model of NFL team performance based only on offensive and defensive passing and running efficiency, plus turnovers, I could then predict the number of wins a team would be expected to have in a season based on the team's efficiency to date.

By using regression, the relative importance of each efficiency stat is weighed in order to best fit the data. I used team wins from the 2004 and 2005 seasons to derive the model's coefficients. The model would look something like this:

c + x*OffPassEff + y*OffRunEff + z*DefPassEff + w*DefRunEff + v*Turnovers = Team Wins

It's a straightforward linear model, where c is a constant, and v, w, x, y, and z are the relative "weights" or coefficients of the respective stats.

The reason I wanted to analyze these stats was to learn the relative weights of each efficiency stat. Is passing more important than running? Was defense more important than offense? Did turnovers matter and by how much? Plus, the explanatory power of these stats could be determined.

To cut to the chase, all 5 variables were significant at the p=.01 level or better. The coefficients are divided by their variable's standard deviation to determine the standardized coefficient, which lets us clearly see the relative importance of each variable.

`Variable      Std. Coeff.-------------------------Turnovers       .388OffPassEff      .488OffRunEff       .215DefRunEff      -.174DefPassEff     -.286r-squared = .79`
The r-squared number means that almost 80% of the variance in team wins is accounted for by the model's variables. Together with the significance of each variable, this tells us the model is valid. The rest of the variance would be accounteded for by factors not in the model (penalties or special teams for example) and by random luck.

Bearing in mind this is based on only 2 seasons worth of data (64 teams), we already learned a lot. Offensive passing appears to be the most important factor in producing a winning team. Defending the run appears to be the least important.

The conventional wisdom around the NFL, judging by the endless chatter of TV analysts, is that the key to winning is to be able to run the ball and play good defense. What this model suggests is just the opposite. If you want to win in the NFL, passing is more important than running, and offense is more important than defense.

### 1 Responses to “Passing is more important than running”

1. Nick says:

I realize I'm commenting on a post over 3 years ago, but I'm just starting to read through your archive- very interesting stuff. I'd like to comment now, here, because I might forget later...

Anyways, when predicting some sort of outcome that depends on other people (i.e. games, fights, etc...), are you worried that the win data isn't *really* independent? For exampe, if team A wins more, then other teams have to lose more. I realize this is a nitpicky thing. I guess one way to get around this is look at only one team for eah year there are statistics for and do the regression.