NCAA Game Theory

Last year I downloaded a huge set of data to build a model that would estimate the bracket with the highest probability of winning a March Madness pool. I did OK in the three pools I entered, winning one of them. But I've had some time to rethink the strategy, and incorporate another cross-section of data. Now that the betting is closed for this year's tournament, I'll share a few of my thoughts.

(from Flickr user SD Dirk)

First, picking March Madness winners is at least as much about game theory as it is about raw mathematics. The problem with the bracket that my model tells me has the highest probability of winning is that it picks the same winner that most of the other participants are picking. This year, that's Kansas. When you have the 'favorite' team as the winner in your bracket, that means you need to stay ahead by picking a lot of early games correctly - that's tough because those games have the least predictable outcomes. The better strategy is to start with the team that has the most plausible chance of winning, but which the fewest others have picked. For me, this year, that's Syracuse.

Second, picking upsets early is a sucker's game, even in pools that give bonus points for correct upset picks. Undoubtedly the first weekend of the tournament is the most exciting, because everyone still has a fighting chance in their pools. At the end of the first weekend, scores are tabulated and people go into work on Monday morning either happy or upset with their placement. But it really doesn't matter, because picking upsets early doesn't guarantee you'll pick the right Final Four or the right champion. If you go overboard with the upsets, you'll probably "upset out" the teams that have the best change of winning the thing!

Third, it's probably true that it's better not to know anything about college basketball before Selection Sunday. That doesn't mean you should be completely ignorant - it means you should base picks on the statistics, not some narrative you've built up in your head. There is a lot of risk for psychological bias in bracket picking. It's extremely tempting to fall into this trap.

Best of luck to everyone this year.


    On March 19, 2010 Nate said...

    my bracket is terrible and entirely based upon narrative reasons in my head