Informational differences in NFL point spread and moneyline markets

Many past works have examined the efficiency of National Football League (NFL) betting markets (for example Pankoff 1968, Vergin and Scriabin 1978, Gander et al. 1988, Gray and Gray 1997, Levitt 2004, Nichols 2012). Though some papers find chances to profit, very few opportunities to earn statistically significant profits have been documented as such a threshold is quite high. Gandar et al. (1988) describe the scarcity of such findings, and few new examples have emerged in recent years. Burkey (2005) explains how even the findings of statistically insignificant, but profitable, trading rules are often the result of data mining or other concerns. Typical works examining wagering rely on betting rules which expose improperly set betting lines or odds. The finding of such an anomaly suggests bookmakers systematically make mistakes when judging the relative strengths of teams in a contest. In this paper we examine the ability of two NFL sports betting markets (point spread and moneyline) to simultaneously incorporate information. To our knowledge, this is the first paper to examine informational differences between the two NFL betting markets. Rather than attempting to judge the correctness of point spreads or moneylines based on external information, we simply employ the reported moneylines as a means to make betting decisions in the point spread market. While bookmakers may properly evaluate available information when setting point spreads or odds, we show that the nature of betting markets makes full, simultaneous information incorporation difficult.Journal ArticlePre-publicatio