Portfolio analytics project
Did NHL players perform up to their cap hit?
This app estimates 2025–26 performance-based value from public NHL statistics, compares that estimate to actual cap hit, and surfaces performance surplus by player, team, GM, and agent.
Explains roughly 46.954% of cap-hit variation in the original regression dataset.
How to interpret this
Performance Value is the model's estimate of what a player's 2025–26 on-ice production was worth relative to NHL cap hits in this dataset. Performance Surplus is Performance Value minus Actual Cap Hit.
This is not a full contract-negotiation model. It does not include contract term, UFA/RFA leverage, cap inflation, leadership, marketing value, or future expectations. It answers a narrower question: how closely did last season's measured performance align with the player's cap hit?
Methodology summary
The dependent variable is log cap hit. The model uses age, games played, points rate, shot attempt creation, scoring-chance creation, high-danger chance creation, giveaways, takeaways, hits, and blocked shots to estimate expected performance value. Predicted values are converted back from log scale using an exponential transformation.
The dataset toggle refreshes the app after removing entry-level contracts. The model predictions themselves are unchanged; the tables, charts, and rankings are recalculated on the filtered dataset.