I have begun doing a poll that's entirely based on stats. The guidelines of the BlogPoll say that ballots should be based solely on the on-field performance of teams by now, and this is my way of both getting in compliance with that and taking pressure off of myself with trying to assemble a top 25 that makes sense.
Here's the top 25, the explanation of how I got here is after the jump. There is some weirdness now, but that's because we have no more than six games per team to go off of. They'll work themselves out as the season goes along.
- Virginia Tech
- Ohio State
- Boise State
- Notre Dame
- Miami (FL)
- Penn State
- Central Michigan
- Georgia Tech
Every piece of my rankings has games against I-AA teams thrown out. Even if a team lost to a I-AA team, it's excised from its record. The stats accumulated in such games make no sense when compared to those from I-A versus I-A games, so for my purposes, I-AA games are just exhibitions.
My rankings have four components. First is national rank in a strength of schedule measure of my own devising. Hit that link for how it works.
Next is national rank in yards per point differential, another stat of my own devising. It takes the yards per point that a team allows (yards allowed/points allowed) and subtracts yards per point that a team gains (yards for/points for). The higher the differential, the more efficient your team is.
Next is national rank in Pythagorean win percentage. It is a more elegant way to include margin of victory than actually using margin of victory, and I use the exponent 2.37 which is about right for football.
Finally, I use national rank in real winning percentage (with results against I-AA teams thrown out, of course). Right now there's a 10-way tie for first among the undefeated teams, so they all get a rank of 1. Next are the 5-1 teams, and they all get a rank of 11. It goes on from there. Real win percentage has a double the weight of anything else because in the end, it's the most important thing.
I take all the ranks in the four categories, double the weight of real win percentage, and then average them all together. Using national ranks smooths out outliers at the top and bottom of everything, so that's why I use it. The final result is a number somewhere between 1 and 120, and the lowest (i.e. closest to 1) is the best.
I will also note that I will curate the top 25 as needed. For instance, Northern Illinois clocked in at No. 24 this week, but as a 3-2 MAC team, they don't belong. I dropped them and brought up No. 26 Utah. No other adjustments have been made.
Here's a table with the top 25 so you can see where everyone ranks in everything and why, for instance, Alabama is five spots below Virginia Tech.
|Rank||Team||SSOS||YPP Diff||PyWin Pct||Real Win Pct||Rank Score