Neural networks mine for gold at the greyhound racetrack

This paper contains a case study where neural networks are used for data mining in the gambling domain. The proposed method uses only publicly available data to train neural networks for predicting the outcome of greyhound racing. Several different betting formats are evaluated, including Win, Quinella and Exacta. The betting strategy based on the trained neural networks is as simple as possible, but still the suggested approach constantly beats the market (i.e. returns a positive result) for the harder formats. The presented technique could be used as a base for a more refined prediction tool for greyhound racing and similar domains. More generally, the paper serves as a demonstration of the power of neural networks when applied to hard and unusual data mining tasks.