## NFL Research: Rushing Production and Margin Of Victory

One of the heuristics I use when picking RBs: Load up on big favorites. The reason? Teams that have a lead will take a more conservative approach, aka run the ball more, while teams that are trailing will have to air it out to catch up. After looking through NFL data using the programming language R, I gathered some evidence to support that idea. While the research below simply confirms what I already believed to be true, I think its always useful to research any theory you have with statistics.

Because the NFL is a fast changing game, I looked at only data going back to 2012. While I had data going back all the way to 1985, I felt the increase in my sample wouldn’t be worth the noise from old data. The first thing I did was look at the correlation between margin of victory and fantasy points from rushing alone, 1 point per 10 rushing yards, 6 points for a rushing TD. The correlation coefficient for margin of victory and rushing fantasy points was ~.44, which as a statistical rule of thumb shows a strong positive relationship between the factors. Further illustrating the relationship between victory margin and fantasy points from rushing: the mean FP rushing is **only 12.68** when a team loses, but **a remarkable 19.14** for a winning team.

The problem with this approach is that margin of victory is also highly correlated with points scored, which is another factor with a strong positive correlation with fantasy points rushing. So a different approach is needed.

I decided to look at the relationship between rushing percentage and margin of victory. If rushing percentage is strongly correlated with margin of victory, and more rushing attempts does not decrease rushing efficiency, than margin of victory should have a strong impact on expected fantasy points from running backs.

**cor(gameData$MOV, gameData$RushPercent)**

** [1] 0.5277942**

This R code simply shows me the correlation between margin of victory and rushing percentage, which is quite high, ~.53. While the graph looks scattered, the positive relationship is quite clear. In the graph to the left, the x axis is margin of victory, while the y axis is rush play percentage.

Applying linear regression to this data illuminates the relationship more.

*Call:*

* lm(formula = gameData$RushPercent ~ gameData$MOV)*

*Coefficients:*

* (Intercept) gameData$MOV*

**0.433274 0.003904**

The last two numbers are the y intercept and the slope. 43.3% is the amount we expect a team to run the ball, with ~.4% per point of victory. So this means if a team is a 10 point favorite, we should expect them to run the ball .433 + .004*10 = 47.3% of the time. This is the formula used in our NFL Point Share Tool to help predict pass and run percentages.

But when teams run the ball more do they run less effectively? Signs clearly point to no. The correlation between fantasy points rushing and rush percentage is the strongest of all the relationships we’ve looked at with a **.635 correlation coefficient**. The correlation between yards per attempt rushing and rush percentage is much smaller, ~.21, but the positive value indicates when teams rush the ball more, they actually do so more efficiently.

**Conclusions**

The idea that a team rushes more when they have the lead is absolutely true. As a team runs the ball more, the fantasy points we expect from rushing increases. Therefore, projected margin of victory should play a big part in who we choose at RB each week. This analysis did not include fantasy points for running backs in the passing game, which I suspect also increases as margin of victory increases. Look for an article next week examining the relationship between fantasy points in the passing game and victory margin.

View all posts by Daniel Steinberg