Responsible Gambling: Using Statistical Models to Beat a Sports Betting Market

Abstract

This paper is concerned with establishing a framework for making consistent returns sports betting. Specifically, this study examines moneyline bets for the National Basketball Association (NBA), the premier basketball league in the world. The existing literature is mostly focused on improving the accuracy of machine learning models in predicting the winner of NBA games. Therefore, one outcome of interest for this paper is whether adding advanced analytics can improve the accuracy of these models. Advanced analytics are statistics that are not included in a sport’s traditional box score and have been largely ignored by academics as inputs to these models. It was hypothesized that adding these advanced statistics will improve a model’s ability to predict games since it provides more meaningful information about a game.

There is significantly less research into applying these models to the growing sports betting industry, which amassed a market size of 83.6 billion dollars in 2022. The betting industry in general has a serious problem with gambling debt and addiction. Thus, the goal of this paper was to find a better way of betting - based on statistics and not pure chance. The hypothesis was that betting (using one of the wagering methods described later in the paper) on games with an expected return greater than 0 will result in statistically significant positive returns in the long run.