The NHL Model
A transparent breakdown of how we generate NHL player prop analytics. No black boxes. Every number is traceable back to source data.
Data Sources
DraftKings Odds
Real-time player prop lines and odds from DraftKings. We pull every available skater and goalie prop including goals, assists, shots on goal, saves, and more. Market odds serve as the benchmark for identifying value.
NHL API
Official NHL game logs, player statistics, and roster data. We track every game to build comprehensive player profiles across the full season.
Historical Game Logs
Detailed box score data for every game. We maintain rolling windows of 5, 10, 15, and 20 game lookback periods to capture both short-term streaks and longer-term performance trends.
Matchup Context
Opposing goalie stats, team defensive rankings, home/away performance, and power play efficiency. Context that turns raw stat lines into actionable prop predictions.
Hit Rate Windows
Last 5 Games
The most recent 5 games. Captures hot streaks and cold spells.
Last 10 Games
A balanced window that smooths out single-game variance. Our primary hit rate metric.
Last 15 Games
A broader view that reduces noise further. Useful for identifying sustained trends.
Last 20 Games
The widest lookback window. Provides the most stable estimate of a player's true rate.
The Process
Fetch Today's Props & Odds
We pull every NHL player prop from DraftKings, covering skater categories (Goals, Assists, Points, Shots on Goal, Blocked Shots, Hits, Power Play Points) and goalie categories (Saves, Goals Against).
Compute Hit Rates from Game Logs
For each prop, we check how often the player has cleared that exact line in their last 5, 10, 15, and 20 games. We also compute home/away splits.
Calculate Expected Value (EV)
EV compares the hit rate probability to the DraftKings implied probability. Positive EV means the prop is priced better than the historical rate suggests.
Run ML Model Predictions
Our machine learning model analyzes each prop using historical performance, matchup data, and contextual features. It outputs a probability for each outcome.
Rank, Grade & Present
Props are ranked by EV and model confidence. The 100% Club highlights perfect L5 hit rates. Cheat sheets organize by category. Everything is transparent.
Confidence Levels
High Confidence
Strong convergence of signals: high hit rate across multiple windows, positive EV of 15%+, and ML model agreement.
Medium Confidence
Good signal strength with positive EV of 8-15%. Hit rates are solid in at least two lookback windows.
Low Confidence
Mixed signals or thin data. May have positive EV but inconsistent hit rates or small sample size.
Ready to find value?
Start exploring today's NHL props and edges.