Sports betting is a rapidly growing industry, with annual revenue rising from about half a billion in 2018 to over 13 billion in 2024. This growth is complemented by the advent of sports betting apps such as DraftKings and Fanatics Sportsbook. This raises the question: is it possible to develop a neural network that places more accurate and profitable bets than the average bettor?
I aim to train neural networks (NNs) to predict the winning side of bets offered by popular sports betting apps and sportsbooks. The NN will be trained on relevant data, including team performance from prior seasons, individual player statistics, and, if time permits, dynamic data such as recent player injuries or other major news updates.
There are potential applications to major sports leagues, such as the National Football League (NFL), National Basketball Association (NBA), and Major League Baseball (MLB). Datasets containing historical seasonal data are freely available on Kaggle for the NBA, NFL, and MLB.
Once trained, the neural network will be deployed by providing it with bets as input and recording the success rate of its output.
The primary objective is to determine whether a neural network can generate a hypothetical profit based on real-world sports bets. Further, the project may also touch on risk management strategies to maintain profitability. A potential avenue for further exploration includes training and comparing multiple models—one trained on high-risk, high-reward bets, while another trained on lower-risk, low return bets. It would also be interesting to see if this NN model can be modified to formulate parlay bets.
- Train NNs on existing datasets and add data to these datasets if certain key information is missing
- Develop a NN that picks a side for a sports bet
- Determine whether a NN can be developed to return a hypothetical profit in sports betting