ABSTRACT
This paper quantifies bettor biases by developing a model of sportsbook pricing, and backing out consumer beliefs from observed betting lines. I model the sportsbook as a monopolist facing consumers with different beliefs about the outcome of a game. They set optimal odds considering their own beliefs and the distribution of consumer beliefs. Using NFL moneyline odds from U.S. sportsbooks (2023-2025) and forecast probabilities from independent sources, I back out the distribution of bettor beliefs and quantify the gap between consumer and sportsbook expectations. The model predicts that sportsbooks expected profit is minimized when bettor beliefs align with true probabilities, and increases with disagreement. Empirically, the recovered belief distributions indicate that the average bias across games, sportsbooks and states is 2.5pp. On average 15% of sportsbooks expected profit can be attirubted to taking advantage of biases in consumer beliefs about the outcome of games, while game contexts that increase viewership such as having a primetime kickoff or being a playoff matchup are associated with a 0.76pp and 1.13pp increase in biased beliefs respectively. Counterfactuals suggest that sportsbooks can expected profits by 12% if they influence public beliefs by increasing bias by 2pp., raising policy concerns about media partnerships.