New Changes to Selection Criteria

CHN (well, Adam) reports on a bunch of changes to the NCAA selection criteria (http://www.collegehockeynews.com/news/2006/09/04_committee.php). Among the highlights:
- RPI criteria (win %, opp win %, opp opp win %) now 25/21/54 instead of 25/50/25.
- Only top 25 in RPI will be TUC (not teams with RPI over .500)
- Winning conference tourney doesn't make you a TUC
- TUC comparison only kicks in when teams being compared have at least 10 games vs. TUC's.
- RPI bonus only for road wins.
- Wins that lower your RPI won't lower your RPI (same rule that was applied to beating weak teams in the playoffs, now applied all season long)

The stated goal is to minimize the impact of playing weak teams. At first blush, this seems to help Cornell and its weak OOC schedule. Though it will be interesting to see how many teams (particularly in the ecac) meet the 10 game TUC minimum.

Edited to clarify RPI change.

It will help our RPI, but if they are correct that it won't change the order much (JTW, any thoughts on that mathematical claim?), then that's not really a help.

This year, in particular, the top 25 only and no-conference-winner-TUCs will hurt us, since we won't be playing many actual good teams, and with TUC not counting unless you have 10... will we even play that many (depends on how many ECAC teams are TUCs, I guess).

I think I like the changes in general though, once they have an ability to impact scheduling (i.e. not this year). But eliminating the RPI penalty, making RPI bonus for road only, and making it depend less on opponent's win percentage, should encourage more traveling for strong teams and more inter-skill-level leaving.

Heck, the opportunity for a road bonus may even get a western team to our barn (west = WCHA, I know the CCHA comes occassional). If not, at least we can prove that it basically is only a superiority complex

Has anyone implemented the new criteria in some script and run last year's results to see how it changes the NCAA's?

Seems like the new formula (25/21/54) will push everyone's RPI much closer to .500. Every team has only 30-40 opponents, but more than 1000 opponents' opponents. Since the overall winning percentage is .500, a (random) sample of 1000 teams will have a distribution much more tightly centered around .500 than a sample of 30 teams. That fourth decimal place in the calcuations could become mighty important.