My system was developed as a hobby and out of the love of the game. The formulation of the model is complex and based on the concept that winning / losing is important, but scoring margins should also be considered. Since the ratings are based on a blend of "predictive" and "retrodictive" algorithms, they have a tendency to show a high correlation to the consensus of a group of models.
The ratings are unbiased and the basic components include:
The programming is advanced and repeatedly recalculates the ratings until they stabilize, and automatically considers the strength of opposition to an infinite depth.
Since the model does not use information from previous seasons, the rankings become more meaningful and less variable as the season progresses. The ratings represent a team's scoring potential at a neutral site. For college football, predictions based on the power ratings can be obtained by comparing two team's ratings and applying a home field advantage. I believe that 3.5 points is a good general estimate for home field advantage for college football.
For the NFL ratings, a more involved prediction model is used to provide the actual score predictions. This prediction model includes calculations for home field advantage trends, away field disadvantage trends, and momentum (time sensitive weighting).
Questions or comments may be directed to firstname.lastname@example.org