NHL Moneyline Model

The NHL Moneyline model calculates win probabilities for each team in a game. It combines team strength differentials (using expected goals), goaltender performance metrics, and situational factors to generate a probability that is then compared against the implied odds from sportsbooks to find edge.

xG (Expected Goals) measures shot quality by calculating the probability each shot becomes a goal based on factors like shot location, angle, and type. We use xGF (expected goals for) and xGA (expected goals against) per game as the foundation for team strength. The model blends season xG (50%), last 10 games (30%), and last 5 games (20%) to capture both baseline ability and recent form.

GSAx (Goals Saved Above Expected) measures goaltender performance relative to expected goals. A positive GSAx means the goalie is stopping more than expected; negative means worse. Starting goalies with strong GSAx provide a significant advantage. The model incorporates confirmed starter GSAx with a confidence factor based on certainty of the start.

Back-to-back games significantly impact NHL performance due to fatigue and often backup goalie usage. The model applies a penalty to teams on the second night of a B2B, with additional adjustments for travel distance and timezone changes. Rest advantages are also captured when one team has more days off.

Corsi (all shot attempts) and Fenwick (unblocked shot attempts) are advanced possession metrics. Teams with high Corsi/Fenwick For % control play and generate more chances. The model uses these as supplementary indicators of team strength, especially useful for evaluating underlying performance vs luck-influenced results.

Home ice provides a measurable advantage in the NHL, typically worth 3-5% win probability. The model applies a home ice adjustment after calculating base team strength, goalie advantage, and situational factors. This captures the benefits of last change, favorable matchups, and crowd support.

Win probabilities are bounded between 20% and 80%. This prevents extreme predictions and accounts for hockey's inherent unpredictability. Even significant mismatches rarely exceed these bounds due to the sport's variance (hot goalies, puck luck, etc.).

Edge = Model Probability - Implied Probability from odds. For example, if our model gives a team 55% win probability but the odds imply only 47%, the edge is +8%. We only surface picks where the edge exceeds our minimum threshold and passes additional filters.

Goalie confirmations typically come 1-3 hours before game time. The model runs with projected starters initially, then updates when confirmations are available. Goalie news can significantly shift projections, so check back closer to game time for the most accurate predictions.

Confidence tiers (MAX, STRONG, STANDARD) reflect our conviction level based on edge size, goalie certainty, and data quality. MAX tier picks have the highest edge and most reliable inputs, warranting larger units. STANDARD picks meet thresholds but have more uncertainty.