Market Efficiency and Censoring Bias in College Football Gambling

Abstract

The team totals and point spread betting lines jointly predict team scores in college football games. The gambling market’s score predictions are shown to be biased due to censoring at zero points. This presents arbitrage opportunities for gamblers who place wagers on the team totals line in cases for which expected bias is high. A naïve betting strategy that exploits this bias using only information contained within the two betting lines themselves yields a win percentage of over 55 percent over the past two decades. This significantly exceeds the typical transaction costs associated with wagers, indicating the market is semi-strong inefficient.