Why Offense and Defense Scoring Rankings are Horrible Metrics for the NFL

I am an avid fan of the NFL. I love looking over the stats almost as much as watching games. The numbers tell you a lot about the flow of the game.

The stats, though, are can be misleading. Take scoring offense and scoring defense, for example. Let’s say that a team gives up few points. This can happen because the defense is stout or because the offense is great and controls the ball. With a few less possessions in a game, you’d expect to see fewer points.

From a Lean standpoint, the problem is that if the data is misleading, you may not recognize a problem, or you might see on where there is none.

So, what I propose is a better way of tracking the performance of a team.

I’d like to see an expected points tally that a team could compare to its actual points. To do this, you’d have to look at the historical data. The previous season would probably be enough.

You’d look at all the possessions for every team in every game starting inside the 5 yard line, and tally up the points from all of those possessions. You’d also subtract out the offensive points allowed. With that information, you could then come up with an expected point value for any possession starting in that territory. You’d repeat the process for every 5 yard section of the field and come up with a baseline.

Over the course of a game, you would tally the number of expected points an offense would get and compare it to the actual score. Offensive points allowed-safeties, pick 6’s, and fumbles returned for a touchdown would count against the offense.

The same process would be done in reverse for a defense. You add up the points allowed based on the field possession. Giving up a touchdown when the offense coughed up the ball at their own 5 yard line would hurt the defense’s ranking less than a field goal from a possession that started after a touchback. You’d have to count field goals, and punt returns for touchdowns against the defense, as they are functions of the defense’s actions.

At the end of a game, you might have a defense with a plus 9.3, meaning that they gave up 9.3 more points than would be expected of the average team playing with the same set of defensive possessions.

So, what would this system do for the rankings?

• The offense would not be artificially inflated by extra possessions.
• The defense would not benefit from time consuming offense that reduces the number of possessions.
• A high scoring defense would offset the points it allows.
• The impact of the kicking game would be factored into the rankings. Good field possession would not matter to the ranking, as it would be adjusted based on the average.

If you really wanted to get fancy, you could do the same thing with yards gained and yards allowed. Those stats suffer from the same problems as scoring stats.

The Lean lesson here is that with inaccurate information, it is easy to go down the wrong path. You have to make sure that your metrics are linked to the behaviors you want to track.