Using Convolutional Neural Network to Predict EURO Games

This paper presents the application of convolutional neural networks to the domain of soccer game predictions. The convolutional neural networks were trained on past soccer games. The statistics for any two teams that were playing in a game were organized in a 6x6 grid representation. This representation was shifted by one column to the right once and twice to augment the training set. This approach resulted in a 69.8% accuracy for predicting the 2020 men’s euro and an 80% accuracy for predicting the 2022 women’s euro. Compared to an ensemble technique approach, these results represent a 2% improvement in predicting the 2020 men’s euro and a 10% improvement in predicting the 2022 women’s euro.