ok, I'm working on a project I mentioned earlier
I'm going to put together a comparison of how the typical #5, #10, #25 team would do facing the different schedules teams have played thus far
I'm going to simply use sagarin predictor and base home fields to come up with a spread
I'm then going to convert that spread to a win pct by using a smoothed out curve of the straight-up winning pct for all games involving a closing spread at that value and immediately surrounding it for every game played from '93 to 2006 in NCAAF
I see a few problems
1) Home field varies from team to team and situation to situation. To be perfect, I'd use something like Phil Steele's HFA point rankings and then add for expected close game and nationally televised game while subtracting for games of a wide margin where homefield has been shown to have less effect (after all, the fans aren't a big deal when a team is up 21 with 6 minutes left).
2) I'm using the current sagarin predictor rating for the teams on the schedule.
--predictor is the best at predicting although some would like the combined score including the elo_chess.
--I'm not using the opponents' ranking at the time of the game. Oftentimes, early assumptions and small sample sizes make these pretty bad in ncaaf. I feel the current ranking is a much better judge of the team as it really is.
I'll put it up on my blog,
TheBCSsolution, once I get it all worked out.