Wednesday, January 14, 2009

Bracket Beta

So I took a few hours to build up a matrix for tournament odds which can easily have teams switched in or out. For daily games I have a couple models that interact to make individual adjustments for each team. At the base of all this are KenPom's efficiency and tempo statistics. It's going to take me a lot more work to build a fair bracket, or in other words, to derive fair market odds. So for now, what's displayed is just what KenPom would say fair futures would be if a certain incarnation of Lunardi's bracket was locked in stone for March. Again, Lunardi's constantly updated bracket prediction is here: http://sports.espn.go.com/ncb/bracketology

VERY IMPORTANT CAVEATS
1) So far this is just KenPom. KenPom has all sorts of bias in it, and does not adjust for any kind of luck in efficiency stats. KenPom often violently disagrees with market lines. For instance, KenPom evaluates Duke as essentially as good as UNC was supposed to be.
2) Many of the teams in Lunardi's bracket won't even make it in March, and many of the higher seeds may fall to worse seeds. #1 or #2 seeding is hugely valuable over a #3 or #4.
3) Who you have in your region impacts your odds greatly. Not all #1 and #2 seeds are created equal. In the bracket below Duke is severely depressing fair odds for every other team in their region because KenPom thinks they're so fabulous.

The numbers in blue at the top are the rounds, and the percentages below them are a team's odds to make that round. So KenPom would say Pittsburgh is 56.5% to make the round of 8 given how all teams have played to date and given Lunardi's bracket structure.

FWIW I'm sitting on some California 400-1, ASU 70-1, Gonzaga 75-1, Georgetown 45-1, Oklahoma ~50-1 (WSEX 4-Pay), and Marquette ~75-1 (WSEX 4-Pay)

Off the top of my head, KenPom loves Georgetown, West Virginia, and Duke a good bit more than the market does on a daily game basis.



2 comments: