Prediction of Premier League Football Matches

The aim of the bachelor's thesis is to determine whether it is possible to predict the outcomes of football matches in the English Premier League. This verification was specifically carried out by creating models using selected methods and subsequently analyzing the results obtained from chosen performance metrics. The data was acquired from the free web portal for football betting, Football-Data.co.uk, and 4180 played football matches from the last 10 completed seasons (2014-2025) were used for the prediction. The methods used in this work were Logistic Regression, Support Vector Machines (SVM), eXtreme Gradient Boosting (XGBoost), and Naive Bayes. The application of the models was preceded by data quality control and Exploratory Data Analysis (EDA), whose task was to obtain basic characteristics of the dataset. Based on the performance metrics, the developed models achieved similar results. A common characteristic of all models was their low success rate in predicting correct draws, which indicates that the problem was not with the chosen methods but with the insufficient ability of the characteristics to predict a draw. The thesis therefore recommends implementing more advanced metrics for future research, which would contribute to improving the performance of the models.