Decision models and online betting behavior: Comparing mean-variance, expected utility, and prospect theory

We examine the drivers of decision making under risk by leveraging the odds offered by bookmakers in online sport betting. First, we use a dataset of over 1.6 million odds from ten years of soccer and 25 years of tennis online betting to estimate parameters for Expected Utility Theory (EUT), Cumulative Prospect Theory (CPT), and extended mean-variance models. Subsequently, we compare their out-of-sample predictive performance. We find that moment-based models that do not explicitly include nonlinear probability weighting perform on par with CPT. Our results further indicate that bettors are neither risk-averse (EUT) nor loss-averse (CPT) and exhibit preferences for skewness and kurtosis. Given the large share of adults who gamble regularly, these findings have broad relevance: We show that extended mean-variance frameworks, commonly used in finance, offer a simple alternative to prospect theory with comparable predictive power in online betting. Our study also offers insights into potential behavioral drivers of asset pricing anomalies, such as the underperformance of high-variance and high-skewness stocks. Furthermore, we provide parameter estimates for each framework that can be used in future empirical research, especially regarding the influence of gamblers in financial markets.