Vytvoření a predikce ratingu hráče

The thesis focuses on the creation of an overall player rating based solely on data. This evaluation is then compared to reality and two sports simulators. Data from the FootyStats website related to players from the English Premier League in the 2021/22 season were used for the custom models. Three models were developed for defenders, midfielders, and forwards. Subsequently, a simple predictive model was created to forecast ratings for young players. This comparison revealed that players who achieved the highest ratings in the thesis were also the best in reality and in both simulators. This confirms that both simulators to some extent utilize real data for their evaluations. The discrepancy occurred only for players in the defender position. The defender model differed the most from reality and both gaming simulators because it disproportionately favored players in central defense positions, while full-backs were not as appreciated. Due to their focus on defensive attributes, players on the wing could not compete with those in the center, as wingers also have offensive duties and do not defend as much as those in the center of the field. The model for midfielders and forwards was very accurate in all respects. It would be appropriate to split defenders into models for central and wing defenders. For full-backs, exchanging some defensive rating attributes for offensive ones would be appropriate. Overall, the models are more accurate than anticipated before the start of the thesis. The work can contribute to understanding how ratings are created in those two sports simulators. Subsequently, the thesis demonstrates that handling data and its subsequent evaluation can greatly assist anyone in a professional football environment.