Chalk, Seeds, and Entropy in the NCAA Men's Basketball Tournament

Abstract

ABSTRACT Based on the results of the 63 games played in each of the 32 NCAA Men’s Basketball Tournaments held from 1985 through 2016 it is shown that the information content of the seeding of the 64 teams invited to participate as to the seed of the eventual winner, varies from year to year, but not in any consistent fashion. The paper thus concludes that the Selection Committee’s seeding process has not improved over time, notwithstanding the availability of more sophisticated metrics for evaluating the teams’ regular-season performance. The fact that a 1-seed wins some 60 percent of the time and a top-three-seed wins 88 percent of the time only reflects the fact that the committee is not seeding the teams at random, but rather is exercising a modicum of judgment, aided and abetted by the tournament’s design.