@@ -12,12 +12,7 @@ We used two tree-based ensemble methods: Random Forest and XGBoost. These were c
Random Forest offers robustness and interpretability, while XGBoost provides optimized boosting performance and fine-tuned control over learning.
## Summary & Future
Tree-based models can effectively predict goal events in hockey.
XGBoost provided the best performance.
Key features include xG, shot location, and manpower situation.
These insights can support coaching strategies and live analytics.
Future improvements could include temporal sequence modeling (e.g., using RNNs), incorporating player tracking data, or testing model generalization across seasons.
Adding real-time inference capabilities could also enhance its use during live games.
Tree-based models, especially XGBoost, proved effective for predicting hockey goal events, with key features like xG, shot location, and manpower situation driving performance. These insights can aid coaching and live analytics. Future improvements may include sequence modeling (e.g., RNNs), integrating player tracking data, testing cross-season generalization, and enabling real-time inference for live game use.
