How does scheme similarity from one season to the next translate to defensive performance?

Earlier this week, we ranked the best play-callers on the defensive side of the ball in the NFL. This ranking takes into consideration our play-by-play grades, along with contextual variables like down and distance.

There’s an element to any evaluation system, even a grainy one that takes into account play-by-play data, that is part “resulting” as well as part process. To dive more deeply into the process aspect of coaching defense, we sought to quantify schemes on both sides of the ball.

We talked about this process a bit when we looked at our ability to predict hot starts and the improvement seen in the Kansas City Chiefs defense. Our approach is to take our data and look at how often teams are in an even front or an odd front, base, nickel or dime, man coverage, etc., adjust for league-level averages in a given year and use principal component analysis to deal with colinearity in the data and reduce the dimension of the problem.