Profitability in Sports Betting: A Case Study of Men's Tennis

Thumbnail Image
Degree type
Graduate group
Sports betting
machine learning
neural networks
predictive modeling
risk management
Kelly Criterion
quantitative finance
Data Science
Finance and Financial Management
Grant number
Copyright date
Related resources

This paper evaluates the feasibility of profit generation through sports betting. While sports gambling represents a large and rapidly growing economic sector, few bettors are actually profitable and there is limited evidence of successful publicly available strategies. We investigate how such a strategy can be built for the game of men’s tennis. Our methodology for creating a strategy consists of two components. First, it includes a predictive analytics component, in which we combine a large number of observable player, match, and tournament characteristics in order to estimate the probability of either player winning the match. We study both linear and non-linear multivariate combination approaches. Second, our methodology contains a financial strategy component, in which we focus on using money allocation techniques to achieve optimal returns. Through statistical simulations and back-testing, we find that it is possible to generate positive expected profits at sustainable levels of risk, with both formal and informal strategies. Interestingly, we also establish that building a successful strategy does not necessarily require the bettor’s model to have higher predictive accuracy than the betting markets. Instead, bettors can focus on a narrow segment of matches (for example “upsets” – i.e., matches in which the lower-ranked player wins) and outperform the market in that segment alone. We conclude that sports betting can be used as a profitable investment vehicle. Beyond tennis, these techniques can be applied to most other sports, especially those for which large volumes of historical data are publicly available.

Eric Bradlow
Date of degree
Date Range for Data Collection (Start Date)
Date Range for Data Collection (End Date)
Digital Object Identifier
Series name and number
Volume number
Issue number
Publisher DOI
Journal Issue
Recommended citation