Abstract
We investigate the phenomenon known as the "middle-seed anomaly", which is the statistical fact that it is better to be a 10, 11 or 12 seed than an 8 or 9 seed in the NCAA Division I Men's and Women's Basketball March Madness Tournament. We use simulations to show that this anomaly persists beyond 16-team tournaments and applies to single-elimination tournaments of any size where the number of players is a power of two. We show which rounds of the tournament the anomaly occurs and estimate that some lower-seeds have over a 10% greater chance of reaching later rounds compared to some higher-seeds. This demonstrates that given certain distributions of player strength, the standard seeding may not provide the most fair tournament.

This publication has 4 references indexed in Scilit: