# Sports Betting: an application of neural networks and modern portfolio theory to the English Premier League

**Authors:** Vélez Jiménez, Román Alberto, Lecuanda Ontiveros, José Manuel, Edgar Possani  
**arXiv:** [2307.13807](https://arxiv.org/abs/2307.13807)  
**Date:** 2023-07-11  
**Summary page:** https://www.summarizepaper.com/en/paper/2307.13807v1/

## Key points

- Novel approach for optimizing betting strategies in sports gambling
- Integrates Von Neumann-Morgenstern Expected Utility Theory, deep learning techniques, and advanced formulations of the Kelly Criterion
- Achieved remarkable profits of 135.8% relative to initial wealth during latter half of 20/21 season of English Premier League
- Explores complete and restricted strategies, evaluating performance, risk management, and diversification
- Findings open avenues for future research in financial methodologies and predictive modeling in sports betting
- Examining long-term behavior of returns with arbitrage opportunities
- Exploring L2 regularization for betting portfolios
- Investigating divergences in portfolio composition and returns between restricted and full strategies through simulation studies
- Adopting a Bayesian perspective for prediction problems and incorporating player-level data alongside team data to improve model accuracy
- Including data from other soccer leagues for more comprehensive modeling approach

## Summary

This study presents a novel approach for optimizing betting strategies in sports gambling by integrating Von Neumann-Morgenstern Expected Utility Theory, deep learning techniques, and advanced formulations of the Kelly Criterion. The method combines neural network models with portfolio optimization and achieves remarkable profits of 135.8% relative to the initial wealth during the latter half of the 20/21 season of the English Premier League. The research explores complete and restricted strategies, evaluating their performance, risk management, and diversification. The findings open up various avenues for future research in both financial methodologies and predictive modeling in the context of sports betting. From a financial perspective, examining the long-term behavior of returns when encountering arbitrage opportunities and exploring regularization techniques such as L2 regularization for betting portfolios could provide valuable insights. Investigating divergences in portfolio composition and returns between restricted and full strategies through simulation studies would also enhance understanding. In terms of predictive modeling, adopting a Bayesian perspective to analyze prediction problems and incorporating player-level data in addition to aggregate team data could improve model accuracy. Including data from other soccer leagues could contribute to a more comprehensive modeling approach. It is important to acknowledge that while this study is grounded in Von Neumann and Morgenstein's Axioms of Preference, discrepancies between theoretical preferences and empirical observations should be further explored. The appendix provides details on the databases utilized for estimating probabilities and obtaining odds from bookmakers. The variables used in predicting match outcomes are specified along with descriptions of each variable's significance. Overall, future research directions outlined have the potential to advance understanding of financial strategies, predictive modeling approaches, and decision-making dynamics in sports betting beyond the scope of this paper.