I've been thinking about trying to do an analysis of tournament results based on 'probabilities' and actual scores. MD for example won, but the MOV makes it 'anomalous' in that it should have been much greater in a 2 v 15 match-up. It is in fact in the same range of shocks as a 5 losing to a twelve seed. I may get around to it.
RPI does have an effect on both bubble teams and seeding. And seeding has a huge effect on results -
Temple with their wins and an RPI 50 would have gotten in, and change two of their wins by 50 RPI points at the back end and two of their losses by 10 points at the front end and they move from high sixties to 50. And if you seed a team that should be getting a 7 seed to a 6 seed, then the opponent in the first round is probably a worse team and a win is more likely. Conversely move a five seed to a six seed line and their opponent is likely a better team. And a team on the 4/5 bubble is changed from a host to an away site.
If you consider that RPI which generally favors P5 teams might result in a one seed line bump for half of the P5 teams that get in - it isn't really going to effect the teams getting the top three seeds, but it might, by giving the other teams a slightly easier first round - it would help perpetuate the myth of their strength. The other issue with RPI is because it doesn't reference point spread in any way, it can penalize competitive losing teams and reward scheduling quirks for less competitive winning teams.
And by using RPI to rank good wins and bad losses it pretends that RPI 80 and RPI 120 really means anything significant to a team ranked RPI 50 Or that a win against RPI 20 BYU means more than a win against 69 Temple
In the 'big' ESPN take down of RPI, they reference Uconn being ranked #3 as being ridiculous, but you are really talking about 2 positions amongst four very strong teams - any statistical analysis could be off by that much. I think the real questions are when you look further down - Duquesne at #17 and BYU at #20 are more difficult to believe.
The reality in the women's game is by the time you get beyond the 5 seeds which represent the top 20 teams, things get pretty shaky in any evaluation system and beyond the six seeds you are really just guessing. There is too much inconsistency in the performance of the teams to rely on the few distinctions that any system will use. When you get a top 15 team like Stanford failing to score 40 points in one game and 50 points in another against similarly ranked teams, what can you rely on. TN can easily beat Oregon State one week and lose to LSU and Alabama in another week.
Just as a interesting comparison - RPI 20 BYU against RPI 69 Temple on Massey results in Temple having a 52% chance of winning - exactly the same percentage as against #47 S. Dakota State - and #45 Princeton is favored in the match-up by 56% and #50 St. John's is favored by 53%
