Expected goals estimates the probability that a scoring opportunity becomes a goal. Because football is low-scoring and individual results are noisy, xG can reveal performance trends that raw scores often hide.
This paper evaluates classification approaches over event data from the 2018 and 2022 FIFA World Cups and the 2018–2022 UEFA Champions League seasons. It studies how feature selection, dataset composition, and model choice affect calibration and predictive performance, then uses visualizations to make team and player comparisons easier to interpret.