I took some time this summer to go back over the ratings from this past 2011 NFL season, trying to determine what went right and what went wrong with the ratings system I was using. I was unhappy with the results from the previous season, and wanted to totally revamp the system.
I tried out about 25 different iterations to the formulas and components, making changes such as how much weight to apply to margin of victory and performance against strong versus weak teams. I also looked at synthesizing the 10, 8, and 6 game trend ratings along with the normal overall rating.
After I ran all the calculations, I hit upon a new ratings system, which I’m going to roll out this fall. I reran the entire 2011 NFL season against the spread and it performed at a .571 clip, which was higher than the highest computer tracking program as set out on http://www.thepredictiontracker.com, which monitors all the computer rankings programs.
I am now going through the same exercise with the 2011 NCAA Football season and once that is complete, I will provide everyone with more details on what the results from the NCAA tests were, and how I got to the results for both the NFL and NCAA.