Failing to Cover: Point Shaving or Statistical Abnormality?

Abstract

The possibility that coaches, players, or referees might be involved in point shaving has been a subject of debate since Wolfers's (2006) controversial finding that favorites in NCAA college basketball games fail to cover point spreads with disturbing frequency. We reconcile Wolfers's finding with evidence provided by Borghesi (2008), Borghesi and Dare (2009), and others that heavy favorites are not, on average, less likely to cover the point spread. We find that the distribution of game outcomes is bimodal, with one peak on one side of the “no corruption” outcome and one peak on the other side. This finding is consistent with point shaving by favorites, who lose by too little, and underdogs who lose by too much. On average, however, the outcome is consistent with the no-point shaving hypothesis. We compare regular-season and post-season results and find that this effect disappears in the more closely observed NCAA tournament games.