Enhancing Sports Team Management Through Machine Learning

Sports like basketball and baseball have seen significant improvements through the effective use of sports analytics. However, research on machine learning in football has primarily focused on prediction. There is a need to explore whether machine learning can contribute to generating more in-depth and accurate insights into football analysis. This essay presents a descriptive analysis of player performance in football matches. Typically, player performance in football is evaluated by consulting a panel of experts, though the criteria they use for evaluation remain undisclosed. By applying machine learning algorithms and pruning techniques, several classifications were developed to identify the most critical functional attributes of players. The functional metrics that delivered the best results in rating predictions had an average error of 0.17. The most significant performance data for each position on the field were closely aligned with the factors experts consider when grading players. The 34 performance characteristics identified achieved a prediction accuracy of 63.4% for match outcomes.