“…Concluding that accurate models are not always profitable, tried to improve them following another approach Kyriakides et al (2015) and applying risk management as filters on their preferences. Finally, considering the performance of both machine learning and linear algebra ranking systems on prediction of upcoming matches' outcomes, Kyriakides et al (2017) suggested a hybrid method, combining Colley's, mHITS and the novel AccuRate rank/rating systems with machine learning methods (Artificial Neural Networks, Decision Tables, Naive Bayes, Logistic Model Trees, Bagging, Stacking). They suggested that combination models performed the best and allow greater flexibility in terms of the desired goal (accuracy, profit).…”