Wednesday, March 24, 2010

Similarity Predictions: Championship Odds Table

I realized just now that I shouldn’t have called this a log5 table in last night’s post, as there are no log5 antics involved. I’m not transforming a rating into a probability; I’m pulling the probability directly from the similarity prediction system.  Whatever you want to call it, here are the odds – according to the similarity predictions – on each team advancing to future rounds of the tournament:

image

I’ve highlighted a few points of interest.

Yellow – This is just to remind you that Robbie Hummel, Kalin Lucas, and Darryl Bryant are injured, so these teams will be harder to predict.  However, the WVY and MSU predictions don’t even take the injuries into account – the system just doesn’t like the matchups at all for these teams.  For example, I have Kentucky as the worst #1 seed for West Virginia to face, and Washington is actually a tougher Sweet 16 opponent for them than Tennessee or Michigan State would be.

Blue – These are the Final Four odds for the four teams in the East.  I just wanted to point out how much of a 4-way toss up this region is, with even Cornell sitting at an 18.4% chance.  While Kentucky is the favorite head to head against all 3 other teams, none of them are huge edges.  Note that Washington’s Final Four odds are higher than West Virginia, despite WVU being favored in their head-to-head game.  That’s because Washington matches up better with both Kentucky and Cornell.

Green – These are a few odds where my system gives a team a much higher chance than the straight Pomeroy log5 table does.  Note that Northern Iowa actually is listed with better Final Four odds than Ohio State, due to them being favored 58%-42% in a head-to-head Elite 8 tilt.  The similarity system also thinks that Syracuse and Kansas State are dominant teams – both teams are 2 to 1 favorites against Ohio State in a potential Final Four game.

Red – Obviously, these are teams that my system doesn’t like so much.  Against the other #1/#2/#3 seeds, Kentucky is only favored head-to-head against West Virginia, and are only a toss up against St. Mary’s.  I don’t believe this for a second, and one of the things I’ll need to work on during the offseason is not letting a couple of outlier results skew the prediction, which I think is happening in that match up.

To try to figure out which team was hurt the most by their remaining draw and matchups, here’s same list, in the same order, but this time I’m showing the team’s expected winning percentage if the 16 of them competed in a  round-robin neutral-court tournament.  The teams highlighted in green have a higher expected winning percentage than at least one team ahead of them, but lower real-life championship odds – they’re the biggest victims of match ups and seeding.

image

Xavier and Butler are probably on there because the West has the strongest overall remaining teams.  Baylor’s there because Duke is a juggernaut.  And WVU is listed mostly because Northern Iowa has an easy remaining path for such a low seed (of course, they can thank themselves for that).

One last note - I’ve got the odds of a 9/10/11/12 seed reaching the Final Four at 66%, which would be pretty awesome.

No comments:

Post a Comment