A novel way to gamble on the NCAA tournament…
I saw a talk at JSM where I was introduced to a fun new (well, new to me) game to play during the NCAA tournament. First, teams are assigned a price based on their seed. This can be done in many ways, but it was set in the talk that the one seeds cost 25 cents, the two seeds cost 19 cents, all the way down to the 15 and 16 seeds which were a penny each. The goal is to choose a set of teams, that costs, in total, one dollar, that will win the most number of games in the NCAA tournament. So picking all the number one seeds, which will cost exactly one dollar, but the most wins they can earn is 19 (4 each to the final four and then one each for the two semifinals and one for the championship). So, according to the speaker, this usually won’t get you the win. First of all, this game is awesome. Once you can stop thinking about how awesome this game is, the next logical question is: How do you choose the optimal set of teams?
Douglas Noe and his student Geng Chen used an evolutionary algorithm to optimize the selection of teams, and they used Ken Pomeroy’s rankings as a guide to the probability that one team will beat another team in the tournament. Now, I don’t think I ever heard of evolutionary algorithms, and, if I have, I’ve totally forgotten about them. But they are wicked cool. Here is the wikipedia page for evolutionary algorithms, and it’s worth checking out. Does anyone have any suggestions as to a good resource for an introduction to evolutionary algorithms?
Pingback: New England Symposium on Statistics in Sports (NESSIS) 2013 | Stats in the Wild