An explainable machine learning analysis of technical and tactical indicators associated with CSL match outcomes

Area of Science:

• Sports Analytics
• Machine Learning in Sports
• Football Performance Analysis

Background:

• Post-match outcome analysis in football traditionally relies on aggregated statistics.
• The need for explainable models to understand the contribution of various performance indicators is growing.
• Chinese Super League (CSL) data offers a unique context for analyzing technical and tactical influences on match results.

Purpose of the Study:

• To develop and validate an explainable analytical framework for CSL post-match outcome prediction.
• To identify the relative contribution, direction, and patterns of technical, tactical, contextual, and player-attribute indicators.
• To classify match outcomes (home win, draw, home loss) using machine learning and SHAP (Shapley Additive Explanations).

Main Methods:

• Utilized post-match data from 240 Chinese Super League (CSL) matches.
• Constructed an analytical framework integrating machine learning (XGBoost) and SHAP for interpretability.
• Classified variables into seven dimensions: player attributes, match context, attacking, possession, duels, defense, and discipline.

Main Results:

• Expected goals (xG) and team Value demonstrated the highest contributions to outcome classification.
• Indicators like Yellow Cards, Round, Long Passes, Throw-ins, Blocked Shots, Goalkeeper Saves, and Recoveries also significantly informed predictions.
• xG and Value showed opposing effects for home wins versus losses, with draws having less distinct indicator dominance.

Conclusions:

• The developed explainable framework effectively identifies key technical, tactical, and contextual drivers of CSL match outcomes.
• Findings provide data-driven insights for video review, performance diagnosis, and targeted training development.
• The study underscores the importance of chance quality and squad resources in predicting football match results.