Machine Learning in Sports: A Case Study on Using Explainable Models for Predicting Outcomes of Volleyball Matches

Transparency matters in sports AI: The paper demonstrates that while deep neural networks achieve high accuracy in predicting volleyball match outcomes, they function as "black boxes" that cannot explain their predictions. This highlights a key principle for responsible AI: accuracy alone is insufficient—stakeholders deserve to understand why a system makes decisions that affect teams, players, and fans.

Interpretable models can be practical alternatives: The study shows that simpler, rule-based models (Boolean Rule Column Generation and Logistic Regression) provide clear, understandable explanations of match predictions while maintaining acceptable performance levels. This suggests that organizations don't always need to sacrifice interpretability for functionality when deploying AI in high-stakes domains.

"Post-hoc" explanations add accountability after the fact: The researchers used methods like SHAP and ProtoDash to explain predictions from complex models after they were built. While useful, this approach suggests a responsible AI best practice: building interpretability into systems from the start may be preferable to retrofitting explanations onto opaque models.

Domain-specific AI requires domain expertise: Applying XAI techniques to volleyball specifically—rather than using generic sports prediction models—underscores that responsible AI implementation requires understanding the particular context, stakeholders, and decision-making needs of each application.