Football Analytics with Python & R

Chapter 4. Multiple Regression: Rushing Yards Over Expected

In the previous chapter, you looked at rushing yards through a different lens, by controlling for the number of yards needed to gain for a first down or a touchdown. You found that common sense prevails; the more yards a team needs to earn for a first down or a touchdown, the easier it is to gain yards on the ground. This tells you that adjusting for such things is going to be an important part of understanding running back play.

A limitation of simple linear regression is that you will have more than one important variable to adjust for in the running game. While distance to first down or touchdown matters a great deal, perhaps the down is important as well. A team is more likely to run the ball on first down and 10 yards to go than it is to run the ball on third down and 10 yards to go, and hence the defense is more likely to be more geared up to stop the run on the former than the latter.

Another example is point differential. The score of the game influences the expectation in multiple ways, as a team that is ahead by a lot of points will not crowd the line of scrimmage as much as a team that is locked into a close game with its opponent. In general, a plethora of variables need to be controlled for when evaluating a football play. The way we do this is through multiple linear regression.

Definition of Multiple Linear Regression

We know that running the football isn’t affected by just one thing, so we need to build ...