NFL Power Rankings

Last week I said that I was going to start working on a Power Ranking formula, and let it evolve and improve over time. Well,  it’s time to debut my mathematical power rankings. The results are a bit on the weird side, but that’s expected.

This week I took into account yards per game, both on offense and defense, turnovers, and point differential. Turnovers and yards per game seem like obvious choices to me. I chose to use points because it take into account other variables that will be hard to include for a while, like special teams, third down and red zone efficiencies, and such things that tend to have an effect on the final score. Turnovers and yards also effect scoring, so think of it as counting those variables twice, since they are really important.

Lets looks the results; then we can analyze the results.

I think the most surprising thing you’ll see here is that Green Bay is all the way down at #6. I actually don’t have a problem with that. While their offense has been very good, their defense has not been. If it wasn’t for a couple very timely interceptions, Green Bay wouldn’t have the record that they have. Having the Saints, Houston and Baltimore at the top seems about right. As for Dallas being that high, well, I said the model was still extremely flawed.

It’s also interesting to see the Seahawks down at #30, right in the middle of all the winless teams. I don’t think the ‘Hawks really belong down there at this point, but I know why they are. The first couple weeks the offense was really bad. Same is true for last week. This team has been good at times and really bad at others, and it averages out to where they currently are. I don’t believe that the Seahawks will be down there later in the year.

Mathematical explanation:

Here’s the formula I decided to go with this week: Power = ((OYd – DYd) + 2*TO + .5(PD)+300)/5

OYD = offensive yards per game

DYd = deefenive yards allowed per game

TO = turn over differential

PD = point differential

Why this particular formula? Simply: it was an easy starting point. This week was more about finding a good source for the data I required that could easily be pulled into a spreadsheet. What I wanted was a formula for this week that put most of the emphasis on offensive and defensive yards per game, and so that was a basis. The turnovers and point differential are meant to be small adjustments to the yardage component.

The problem with the level of the “adjustment” I used was that each turnover counts the same as just a 2 yard difference on either offense or defense. Clearly turnovers are worth more than that in determining which team wins. Obviously, this model has a very very long ways to go.

Adding 300 and then dividing the entire thing by 5 was a quick and dirty way to make the results end up between 0 and 100 for all teams without changing the relative distance between each team. There are better way to do that mathematically, but since this model will chance each week as I improve it, I didn’t want to spend the time required to derive the proper adjustment at this time. Don’t worry, I will once we get further along in the development process of our formula.

I can already see the problem with using point differential and not doing it using points per game. The totals will keep getting further and further zero, and thus will lead to more extreme Power levels. I’ll adjust that for next week. While turn overs have the same problem, those tend to remain closer to zero, and are much more random than you might think. Still, I’ll likely switch that to a per-game version as well.