NHL Engine Highlights
Expected Goals (xG) Modeling
Shot quality is measured using MoneyPuck's expected goals framework. Not all shots are equal — the engine weights each shot by location, type, and game context to produce accurate scoring projections.
Zone-Based Shot Sequencing
The simulation models offensive zone entries, shot generation from different ice zones, and the flow of possession. This captures how teams generate and suppress scoring chances rather than relying on raw shot counts.
Goaltender Form Analysis
Starting goaltender identification and form-weighted save percentages. Recent performance is weighted more heavily than season averages, capturing hot and cold streaks that affect game outcomes.
Bayesian Shrinkage Calibration
Raw model probabilities are calibrated through Bayesian shrinkage to correct for overconfidence. This ensures the edge percentages you see reflect realistic expectations rather than inflated model outputs.