Convolutional Neural Network and Feature Encoding for Predicting the Outcome of Cricket Matches

This paper proposes two novel approaches for predicting the outcome of cricket matches by modelling the team performance based on the performances of it's players in other matches. Our first approach is based on feature encoding, which assumes that there are different categories of players exist and models each team as a composition of player-category relationships. The second approach is based on a shallow Convolutional Neural Network (CNN) architecture, which contains only four layers to learn an end-to-end mapping between the performance of the players and the outcome of matches. Both of our approaches give considerable improvement over the baseline approaches we consider, and our shallow CNN architecture performs better than our proposed feature encoding-based approach. We show that the outcome of a match can be predicted with over 70% of accuracy.