I'm still working on this page, but I'll explain what I'm trying to do:
I think NCAA football rankings are crap, especially in the preseason. The polls are dominated by non-tangible, unobjective, emotional factors such as the desire for an underdog (e.g. Boise State) to reach the national championship or the feeling that a team (e.g. Alabama) is number one until beaten. Stover, Hall and Gibson (PSU) published a paper about a fuzzy logic architecture - a way of combining data to provide an output confidence factor, which here will represent a confidence that the team will be good.
Unlike the BCS polls, which are statistically trying to measure how good a team IS, this is more of a measure of how good a team should be and thus is fundamentally different. In addition, the schedule can then be analyzed to determine how the season might play out.
One of the interesting twists is that I am going to allow the user to control the weighting of each factor - YOU will get to decide how much a given piece of information (say, whether the starting QB is returning) affects how good a team will be. Then you can use that weighting scheme to recalculate the top 25 both in terms of how good the teams are and in terms of season expectations.
Sounds awesome? I think so. For now, here are some examples of teams: