The ratings are calculated in a two part process combining a team's results for the year and the team's seeding for the year. I'm not going to get into mathematical specifics here, but here's an overview:
A team's raw power rating is calculated by combining their winning percentage with an estimate of their winning percentage using an offshoot of Bill James' Pythagorean theorem. Fudge factors are included in order to at least somewhat be able to handle teams that are unbeaten, untied, and unscored upon. The raw power rating is modified according to the average power rating of the opposition, then by the opposition's opposition's average power rating, etc. until the (opposition)^n's power rating converges for all teams being rated.
In order to be able to have reasonable ratings during the beginnning of the season, the system calculates pre-season power ratings or seedings at the beginning of each year. For more detailed information, click here.
The pre-season seeding is combined with the modified power rating in a function that decreases the impact of the seeding as the season goes by. The pre-season seeding never goes away completely, but has very minor impact on the power ratings at the end of the year.
The power rating is defined as the chance the team has to beat an average Division 1-A team with no effect from location. This, a team with a .700 power rating has a 70% chance to beat an average Division 1-A team at a neutral site.
The ratings are calculated using a spreadsheet. The ratings do not use results on a game by game basis; rather, they use seasonal totals and schedules. The chance a team wins can be calculated based on the power ratings of the teams in the game and the home field advantage, and an expected pointspread can be calculate if you know the amount of points expected to be scored in the game.
Back to ARGH Power Ratings home page