|Mike DeSimone's College Football
Division I-A Ratings
Rating Algorithm Description
I am using a least squares prediction algorithm to determine these ratings. The algorithm uses factors that vary from week to week during the season. For example: Last years final ratings, the number of returning starters, schedule strength, coaching experience, and a key returnee factor, are among a few factors used.
I use statistical analysis to determine if these factors are significant based on previous prediction history. Given the weeks factors along with their values and prediction history I solve for a set of coeeficients. I use these coeeficients in a polynomial equation and provide the updated prediction for a given week.