Modeling March Madness Picks

Presented by:
Simon-Peter Nyamoko-Agata (Elon University)
Abstract:

Millions of college basketball fans across America participate in the March Madness Bracket Challenge each year. People utilize different strategies, ranging from simple tactics to complex computational strategies. Due to the overwhelming number of possible outcomes, no one has been able to create a perfect bracket. My research studies how people choose which teams will advance in the tournament. To do this I collected a variety of data including traditional ranking information, advanced team performance data, and previous NCAA Tournament success. My project models the percentage of people who will pick a team to advance to the following round. Specifically, I produced a model for each of the six rounds that reflects the odds ratio of a team being picked to advance. My research selects the statistically significant variables for predicting the odds ratio of a team advancing in that particular round and compares it to those in the other rounds. Via this method, I was able to find that seed and previous tournament wins were significant predictors of the odds ratio for all six rounds and rank was significant for the first three rounds only.