Use of Performance Metrics to Forecast Success in the National Hockey League

Predicting success in hockey is an important area of research which has received little attention in the sports data mining community. We are the first to propose a machine learning approach to forecast success in the National Hockey League. Our approach combines traditional statistics, such as goals for and against, and performance metrics such as possession and luck, in order to build a classification model. We construct several classification models with novel features such as possession and luck in order to build a classification model. Our results indicate that Neural Networks construct the most robust classification models. This confirms the work of earlier researchers, who have also employed Neural Networks in other sports data mining domains. Our results also show the statistics of PDO (which shows, in the short term, the teams playing better or worse than the expected variance) does not aid the prediction.