I forget the true number on the stat, but generally you lose, on average, 30-45% of any given class by their senior year. If you ask me, that is too significant of a number to ignore. I know you were trying to make a meaningful analysis here and it may look nice, but looks might be deceiving. I honestly don't see how one can take any meaningful insight from it. If you wanted to make a more impactful analysis, why not consider focusing on 2013 and 2014, revising your "class rankings" number based on a 'revisionary ranking' that discludes, namely, all players lost to attrition, then from there take into account how many fr, so, jr, & sr's there are as projected starters for the '13 season and then use your 'class weighting percentage' based on that. Then you still have to ask the question, if there are an equal amount of sophomores to seniors and more juniors than any other class, should my number not reflect that statistic? Secondly, wouldn't the two deep be a better indicator used to better gauge the talent level of the teams being compared?
Just a thought ... I know you worked hard on it and I'm trashing it a little by saying it's entirely insignificant, but in my opinion it has real potential if you were to use the model to focus on current talent level rather than inflated rankings and vague class percentage values.