1. Adjust the average recruiting class for juniors/seniors rather than weighting each equally. Having the #1 overall incoming freshman class is unlikely to help you in the current year.
2. Boise State has been an abberation for awhile. Whether it's an easy schedule, luck regarding the QB position, or whatever, I'd think of them as an anomaly rather than proof that rankings dont matter. We'll see how they do in Big East play.
3. Using "Average Class" and "Difference" doesn't make sense. Alabama is an 8.4 average ranking, but in that time period that may be the 3rd best average. Nobody is #1 4 straight years, so your formula forces there to be significant differences between place of finish in classes and 2011 season. Instead, consider calculating Average Class (maybe with just seniors/juniors) then assigning the ranking in order of who has the best average (i.e. USC's 5.8 gets reordered to #1, Texas's 6.4 gets reordered to #2, Alabama's 8.4 gets reordered to #3, etc.) that way it is actually possible to see if "best recruiting class maturing to juniors/seniors equals or doesn't equal on field success."
4. I dont look at recruiting class rank any differently than I look at a prospect's individual star ranking.
As in, having a better recruiting class doesn't guarantee success, it just makes the odds better that your team will end up high up. If you lined up the top 20 "average classes" top to bottom, I bet it wouldn't be random but would rather be about a 50/50 split between top-20 recruiting classes and non-top-20 classes. Therefore, if you bring in a top-20 class every year, you would be 50% likely to end up in the top-20 (factors including conference, coaching, schedule not withstanding) compared to the other 100 schools with a 10% chance of ending up in the top 20. Thus, instead of saying "its an equal mix, recruiting classes don't matter" the conclusion would be "recruiting classes give you a 5 times better shot of finishing the season in the top 20." This is an example but you can understand the line of thinking hopefully.