Sunday, February 8, 2009

March Madness Bracket Market Odds, at Last

If the season ended today, using the most recent closing lines, an efficient market would probably price Lunardi's bracket like I've pasted below. I'm not doing Bracketology 101's because their last bracket has some serious errors in it (Northern Illinois is not making the dance, sorry). One thing of note is that Oklahoma is a paper tiger. The market doesn't think they're a top 10 team, and if they manage to game the RPI with small victories over bad teams, there could be some pretty terrific value come March on the other 15 teams in their region.

There are some fairly important caveats for the market bracket I've posted:

1) These are generated using a model that adjusts some high level statistics of tournament teams to derive the past two games of market lines within a half-point to a full point. I use factors in my main model that apply to individual matchups which I can't incorporate feasibly in a giant excel matrix. That said, the differences should be small (~3% or 4% delta in individual game probabilities, which should mostly randomize themselves out of a big bracket. To illustrate, North Carolina's odds to win it all are a factor of hypothetical moneylines with the 32 teams they could face * the odds each of the 32 teams makes the championship game * the odds NC makes the championship game).

2) Bad teams that upset good teams in March will be given some credit by the market. If North Carolina is -18 over some mediocre small conference team, they won't be -18 over them if they meet in the final four. That team will have knocked off some good at teams at long odds. So, the odds of poor teams are probably a little better than posted, and the odds of good teams a little worse than posted. Again, the differences will generally be small.

3) Odds can shift quite a bit depending on the bracket. I can get a 25-1 team to become 35-1 if I move them to the right region.

4) The season doesn't end today, and some weekend outcomes are going to change this around. Small changes in high level metrics can make dramatic changes in tournament odds. If I had to guess, once I see the next Gonzaga line (after being crushed as four point favorites by Memphis), they'll be more like 45-1 than the 34-1 below.

No comments:

Post a Comment