Improving expected Goals (xG) models: methods, results, and novel extensions

We present an enhanced expected goals (xG) modeling framework with improved data pipeline, richer features, and interactive deployment. Using StatsBomb event data, we implemented thorough cleaning (merging event and lineup JSON, extracting freeze-frame defense data) and engineered novel features (angular defensive pressure, goalkeeper distance, pre-shot sequence). An XGBoost model was trained and calibrated, achieving strong discrimination (AUC ~ 0.878) and calibration (Brier ~ 0.0686) on held-out shots. Key predictors included game-context and shot geometry (goal difference, shot angle and distance) and defensive metrics, as revealed by SHAP analysis. We summarized recent xG studies, highlighting that our model outperformed prior work (e.g. AUC~0.80) by incorporating these new features. An accompanying Streamlit app demonstrated real-time xG prediction (single-shot sliders, batch CSV upload) and SHAP explanations. Results indicated that the enriched feature set significantly improved predictive accuracy over baseline models, and the deployment prototype facilitated practical analytics for coaches and analysts. Our contributions include (i) a novel “angular pressure” metric, (ii) logic for pre-shot pass sequences, and (iii) an open pipeline and app for xG analysis.