A Power Rankings Formula


I’m taking a short break from writing about the Seahawks and scouting draft prospects to do a little math. The people who know me outside of the blogosphere wont be surprised by this. I have a bit of that “nerd” thing going on, what with my degree in physics and everything. The truth is that I’m fascinated by mathematical models and data.

I’ve also been wanting to start doing power rankings on the blog. Everyone does them, from ESPN to the Seattle Times to all the other local Seahawks blogs, so I shouldn’t be hesitant to join in. But the problem is that everyone does them. Why should I bother re-creating what you can find elsewhere unless I can do it better.

So I’ve decided to combine these two concepts. I’m going to create a mathematical model of the NFL and use it to determine the power rankings, and I want your help. don’t worry, I wont be asking you to do any math, but I want to know what stats I should use within my model.

One thing I know I don’t want to include is Win and Losses. I know that sounds weird, but I want to be able to rank teams from best to worst. As we know, the best team doesn’t always win. There’s a lot of noise within the win-loss data, and I want to avoid that.

I mean, load up your favorite football sim (like Madden) and create a custom league with 32 identical teams and sim a season. You’ll most like end up with team records ranging from 4-12 to 12-4, even though they should all end up 8-8. Such is the nature of football, and I don’t really want to include that noise in my model.

He’s what I want your opinion on:

  1. Yards per game or yards per play as the primary measure of quality for each offense and defense?
  2. Special teams yards? I really am not sure what to do with this one. As we saw both last year vs San Diego, and this year vs San Fransisco, special teams can decide a game. But unless there a couple of big returns for TDs, a lot of kick return yards can simply mean that your defense gave up a lot of scores, and thus the other team kicked off a bunch. Similarly, a lot of punt return yards could simply be the outcome of playing against a bad offense that punts a lot. So while these stats matter, I’m not sure how much to weight to give them. Please give me your thoughts on the matter.
  3. Turnovers – I’m of the belief that at least 80% of the cause of turnovers is on the offense. If that’s the case, should I include defensive takeaways in the model, or just count them against the offense?
  4. Time of possession and run-pass ratio? Both were once stats that could be looked at as predictors of success, but in today’s pass happy league I’m not sure they matter like they once did.
  5. Sacks, or all tackles for loss (including sacks)?
  6. How much do you feel things like penalties, 3rd down efficiency, red-zone efficiency matter? clearly they matter, but are they on par with yards per game, or are they a much smaller factor?

It’s going to be a work in progress, and I’m sure it’ll generate some crazy results that we’ll be able to laugh about for quite some time before we start to get it right. For better or for worse, I’ll come up with my first rankings after this week’s games.