The Application Of The Machine Learning Principles In The Sports Betting Systems

Anotace: Use of machine learning for football match result prediction is still a difficult task. Betting companies use predictions for betting odds estimation and profit generation. Chosing the correct data for training models is thus crucial. In this paper the data model is selected and multiple machine learning algorithms are compared. Along with data model selection, impact of number of game seasons on success rate was analyzed. Last but not least, simple three class prediction was compared to separate binary predictions of win, lose and draw. All models were tested by one whole game season from last 380 games of English football championship Premier League. The best results acieved passed 60% accuracy, which outperforms all the analyzed works.