The tournament is over, and Kentucky are the champs. Who predicted that? Well, lots of people. Among the rankings I tracked, however (including my game simulations), only the tournament committee got it right by making them the overall number one seed. Here’s how the initial brackets fared for each ranking:
My simulations (final row) got off to a strong start through the first four rounds, as did the other quantitative approaches (Pomeroy and Sagarin rankings). However, seeding finished strong. By percent correct, the quant methods were slightly better (65 to 66% correct), but choosing based on seeding would have attained the top score by usual bracket scoring methods (1 for round 1 wins, 2 for round 2 wins, 4 for round 3 wins, etc.). The usual scoring method gives 32 points for predicting the champion, which is a big reward for seeding this year.
If I forgive mistakes in the brackets and redo them after each round, here are the performance numbers:
These numbers are pretty similar to historical numbers for all five methods and about even with prediction using the Vegas line. All in all, I’m calling this year a success. I can make some improvements to these simulations for next year (allow for endgame strategy decisions, for example). Still, my simulations this year were correct on 58.7% of over/under bets and 54.7% of spread bets, which is enough to be profitable. The simulations tended to prefer underdogs against the spread and the money line, so money line bets based on the simulations would have been extremely profitable this year (correct bets on Norfolk State and Lehigh had very high payouts).
If I can find the time and the data, I’ll have NBA playoff simulations coming up later this month. Thanks again to TeamRankings.com for providing the data for my simulations this year!