The continuous development of big data technology makes data acquisition and organization easier and faster than before, and predicting the results of sports games through data mining algorithms has become a hot research direction. In this paper, crawler technology is used to collect data on players’ performance, game scores, wins, and losses in sports soccer games. Once the data has been pre-processed, the key player performance features are identified through a stepwise regression method, and the gray correlation coefficient and weighted gray correlation degree of dimensionless data are calculated. Finally, based on the gray correlation coefficient between player performance and match score and the SHAP model, a prediction model for the win and loss of soccer matches was constructed. The study shows that there is a strong correlation between the performance of players and the winners and losers in soccer matches, and the SHAP model has good applicability to the complex non-linear relationship between the performance statistics of players and the winners and losers in soccer matches, and the prediction accuracy can reach 91.6%. This paper provides certain decision support for the tactical formulation and execution scenarios of teams in soccer matches and lays the foundation for the research direction in the field of competitive sports in the prediction of match results.