Foul on the Committee? Investigating and Predicting NCAA Tournament Selection Committee Decisions

• Activity Number: 439
• Type: Invited
• Date/Time: Wednesday, August 3, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Sports
• Author(s): Jay Coleman*+ and Mike DeMond and Allen Lynch
• Companies: University of North Florida and Charles River Associates and Mercer University
• Keywords: logit ; probit ; NCAA ; basketball ; group decisions ; bias

Abstract

The NCAA Men's Basketball Committee selects the 34 at-large teams for each season's tournament, and seeds the 65 participating teams. The great debate that follows the Committee's decisions each year often includes charges of bias in favor of major conferences and toward teams with some sort of Committee representation. Another common complaint is that the Committee seems to set limits on the number of teams that can be invited from any given league.

In an extension of research on predicting at-large bids first published in 2001, we investigated the question of bias in the at-large selection of teams and the seeding decisions from 1999-2008, using logit and probit modeling. After controlling for team performance using 41 factors, we found evidence of bias in favor of nearly all major and even mid-major conferences in selection and/or seeding. We also found strong evidence of favoritism toward teams represented on the Committee.

A variation of the at-large selection model predicted 32 of 34 teams selected by the Committee for the 2009 tournament, and 33 of 34 in 2010. Had the model's 2010 last-team-in and first-team-out predictions been reversed, it would have been perfect.