Sunday, March 29, 2009

April 3, 2009; Homewood, Clark 110
-- Raymond Cheong

Title: "How to use bioinformatics and other methods to win sports pools without knowing anything (much) about sports"

Abstract: In sports pools, such as those related to the ongoing men's college basketball tournament, it is often observed that those who know the least about sports end up making more accurate predictions than those who are more "knowledgeable" about sports. As a hobby project over the past ~10 years, I have been putting this idea to the test by investigating the performance of unsupervised algorithms to predict the winner of team sports. In this informal talk, I will highlight a variety of creative algorithms and discuss associated methodology and strategy. These algorithms include simple coin-flipping and Google-based methods, as well as bioinformatics-inspired linear models, network-oriented ratings, and clustering.