Unveiling the xG metric in soccer: how the adjusted xG improves match predictions

This study aims to expand the analytical value of Expected Goals (xG) by combining Pre-shot Expected Goals (Pre-xG) and Post-shot Expected Goals (PSxG) within a Bayesian probabilistic framework based on Bernoulli trials. This approach made it possible to estimate both partial and final match outcome probabilities throughout different phases of a game. Over the past decade, xG has evolved from a metric that estimates the likelihood of individual shots resulting in goals into a broader performance indicator used to assess teams and players more comprehensively. The model was tested in a high-profile LaLiga match between Real Madrid and FC Barcelona during the 2024–2025 season. Instead of evaluating probabilities across the full match only, the model generated dynamic estimates at various moments of play. Using Pre-xG, the probability of a Real Madrid win was 14.40%, compared to 65.51% for Barcelona and 20.09% for a draw. When PSxG was applied, these probabilities shifted to 1.47%, 91.90% and 6.62%, respectively. The most likely scorelines were 1–2 according to Pre-xG and 0–3 based on PSxG. These findings suggest that integrating Pre-xG and PSxG into a probabilistic framework may provide a deeper understanding of match dynamics, tactical influence, and player contributions in real time.