Reading the Ice
How we model individual player shot volume to find mispriced shots-on-goal lines before the puck drops.
Think of a weather forecast. Meteorologists do not just look at today's temperature and guess tomorrow's. They feed atmospheric pressure, wind patterns, humidity, and historical norms into physics-based models to produce a probability distribution: “70% chance of rain, 30% chance of clear skies.” Our SOG props model works the same way, except the atmosphere is a hockey game and the question is whether a player's shot count lands over or under the book's number.
This page explains how the model thinks, from raw player data to the picks that surface in your terminal. We will keep it honest: you will understand what powers the system without us handing out the recipe.
The Challenge
Why SOG Props Are a Different Beast
Imagine predicting how many packages a delivery driver will drop off tomorrow. You know their daily average, but that average masks wild day-to-day swings: route changes, traffic, weather, new stops. Now add the twist that whether they hit 3 deliveries or 2 is the entire bet. That one-shot margin is the SOG props market.
Discrete Outcomes
Shots on goal are whole numbers: 0, 1, 2, 3. A player averaging 2.4 SOG per game can easily produce 1 or 4 on any given night. The integers make the distribution lumpy, not smooth.
Opponent-Shaped
A forward facing an aggressive shot-suppressing defense is a different proposition than the same player against a porous team that gives up 35 shots per game.
Ice Time Uncertainty
A player can only shoot when they are on the ice. Fluctuations in time on ice, power play usage, and coaching decisions create a hidden variable the book cannot fully price.
The opportunity: Most sportsbook SOG lines are derived from simple averages and general market flow. Our proprietary model goes deeper — analyzing thousands of data points per player per game to uncover lines where the book has misjudged the true probability.
The Projection Engine
Counting Events, Not Guessing Averages
Imagine a factory that produces lightbulbs, and you want to predict how many defective bulbs come off the line in a shift. You would not just average past shifts and call it a day. You would model the production rate per hour, factor in the quality of raw materials today, and use a statistical distribution designed for counting events to compute the probability of 0, 1, 2, or 3 defects. Our SOG model follows the same logic: it estimates a player's expected shot rate for tonight's specific conditions, then uses a counting distribution to compute the exact probability of landing over or under the book's line.
The model builds a player-specific projection from multiple input signals, each capturing a distinct factor that shapes tonight's shot volume. These signals combine into a single expected shot count, which feeds the probability engine.
| Signal | What It Captures |
|---|---|
| Player Profile | A composite view of the player's tendencies blending season-long data with recent form to capture both consistency and momentum. |
| Usage & Role | How the player is being deployed tonight — factoring in role context and situational advantages that affect output. |
| Matchup Context | Every opponent presents a different environment. The model adjusts for the specific defensive and tactical tendencies of tonight's opposition. |
| Game Environment | Broader game-level factors like expected tempo and flow that influence total event volume for all players on the ice. |
| Situational Factors | Additional context layers including special teams, lineup changes, and other variables that shift a player's expected output. |
Key insight: The book sets a line based on general averages. Our model asks: given this player's profile, against this opponent, in tonight's specific conditions, what is the actual probability of going under? That specificity is where edge lives.
Player Profile
The model builds a composite view of each player's tendencies, blending full-season data with recent form. This captures both true talent level and any short-term shifts in output.
A player whose recent output has diverged from their season average is a fundamentally different bet than one performing at baseline, even if the line stays the same.
Matchup Context
Not all opponents are created equal. The model evaluates how tonight's specific opposition affects the player's expected output. Defensive environments vary dramatically across the league, and this context is critical to accurate projections.
This is one of the model's most impactful signals. The same player can have materially different probabilities depending on who they are facing.
Usage & Deployment
Output requires opportunity. The model evaluates recent deployment patterns and role context to assess whether a player is getting more or less opportunity than the book's line assumes. Shifts in usage that the market hasn't priced create edge.
Game Environment
Broader game-level conditions influence total event volume. The expected tempo and flow of tonight's specific matchup affect every player on the ice. The model incorporates these dynamics into each individual projection.
The UNDER Edge
Why This Model Focuses on Unders
Picture a casino roulette wheel. The house does not need every spin to go their way. They need the math to be slightly in their favor over thousands of spins. SOG unders work on the same principle: the nature of low-count discrete events creates a structural tilt.
When a book posts a SOG line of 2.5 for a player averaging 2.4 shots per game, the UNDER is not just “slightly more likely.” The discrete counting distribution produces a probability that is often meaningfully higher than the market implies, because most shot outcomes cluster at low integers: 1, 2, and 3 dominate the distribution. The OVER requires hitting 3+, which demands both opportunity (ice time) and execution (getting shots through). The UNDER just needs the natural variance to do its thing.
This is not a blanket “always bet unders” strategy. The model evaluates each player-line-opponent combination individually and only surfaces picks where the statistical probability meaningfully exceeds the market's implied probability.
Statistical Foundations, Not Guesswork
The model uses a well-established statistical distribution specifically designed for counting discrete events. This is the same mathematical framework used in manufacturing quality control, insurance actuarial science, and epidemiology. Rather than assuming a bell curve (which does not apply to whole-number counts), the model computes the exact probability of each possible outcome: 0 shots, 1 shot, 2 shots, and so on. The sum of probabilities at or below the line gives the true P(UNDER).
Finding Value
Where Edges Come From
Sportsbooks set SOG lines from a blend of player averages and market action. They are good, but they cannot individualize every factor for every player every night. We look for the specific gaps:
- Opponent mismatch: The book line reflects the player's general average, but tonight's opponent suppresses shots at a rate well above or below league average. Our opponent suppression factor catches what flat averages miss.
- Ice time shifts: A player's deployment has changed in recent games, but the book line still reflects older ice time patterns. Our projected TOI captures the shift before the market adjusts.
- Pace environment: Tonight's matchup pace differs significantly from the player's season average game pace. The model adjusts shot expectations for the specific game flow.
- Distribution edge: The counting distribution reveals that the true P(UNDER) diverges from the market's implied probability. The gap between our model's probability and the book's fair probability is the edge.
Model vs. Market
The model outputs P(UNDER) for each player-line combination. We strip the bookmaker's margin from both sides of the line to get a fair implied probability. The gap between the model's probability and the fair probability is the edge. If a book prices UNDER 2.5 SOG at -130 (fair implied ~56.5%), and our model says the true P(UNDER) is 63.2%, the edge is +6.7%.
Important: Not every edge is worth taking. The model enforces minimum edge thresholds and line filters. Only picks where the model sees a meaningful probability advantage are surfaced to subscribers.
The Confidence System
Multi-Gate Quality Control
Like playoff roster cuts, every potential pick must survive multiple rounds of evaluation. If it fails any single gate, it is cut. One failed gate and the pick is gone.
Gate 1: Minimum Edge
The model's P(UNDER) must exceed the market's fair probability by a configurable threshold. Marginal edges get rejected outright.
Gate 2: Line Filter
Only specific SOG lines are considered. The model targets lines where the counting distribution produces the most reliable probability estimates.
Gate 3: Daily Cap
A maximum number of picks per day enforces quality over quantity. On busy NHL nights with 15 games, only the highest-conviction picks survive.
Gate 4: Tier Assignment
Surviving picks are ranked and assigned confidence tiers that determine unit sizing. Higher-edge picks receive larger allocations.
Edge-Ranked, Tier-Sized
Picks that survive all gates are ranked by edge strength. The top picks are assigned to confidence tiers that determine unit sizing: higher-edge picks are given more weight, while picks near the threshold receive smaller allocations.
| Metric | How It Works |
|---|---|
| Edge | Model P(UNDER) minus fair (de-vigged) probability. Higher edge = stronger signal. |
| EV | Expected value calculated from the model's probability and the actual UNDER odds offered by the book. |
| Tier | Confidence tier based on edge strength. Higher tiers receive larger unit allocations. |
Important: Individual picks still lose. SOG props are inherently variable because of the discrete, low-count nature of the outcome. The model is designed for expected value over many bets, not guarantees on individual picks.
How We Measure Success
Anyone can cherry-pick a hot week. We believe in full transparency across large sample sizes. Here is what we track and why it matters:
Win Rate by Tier
Higher confidence picks should win more often. We track each tier separately to validate our confidence calibration.
Return on Investment
Win rate alone can mislead. ROI accounts for the odds on each bet, showing actual profit relative to amount wagered.
Closing Line Value
Did we get better odds than the final line? Consistent CLV proves we are identifying real edges, not getting lucky.
Sample Size
Results over 20 picks mean little. Results over hundreds of picks reveal true model performance through the variance.
Closing Line Value (CLV)
CLV measures whether the line moved in your direction after you locked in your pick. If you bet UNDER 2.5 SOG at -130 and the line closes at -150, the market confirmed your read. You got a better price than the final consensus.
Sportsbooks use CLV to identify their sharpest customers. Consistent positive CLV is the single strongest predictor of long-term profitability, because it means you are consistently ahead of the market.
How our picks are timed:
Picks lock 10:00 AM ET daily, hours before most NHL games. This gives you time to place bets before lines move, maximizing CLV capture.
The Bottom Line
SOG props reward those who understand that hockey shots are not random, they are the product of ice time, matchup, pace, and player tendencies. Our model does the heavy lifting of analyzing each factor for every player on every slate, but the philosophy is simple:
- Model the count, not the average: a purpose-built counting distribution that computes exact probabilities for discrete shot outcomes
- Context is everything: opponent suppression, projected ice time, and game pace transform a generic projection into a specific one
- Filter ruthlessly: through multiple independent gates. Only genuine edges with meaningful probability advantages survive
- Size by conviction: tier-based unit sizing allocates more to higher-edge picks and less to marginal ones
- Track everything: so you can evaluate us with full transparency across hundreds of picks
The scoreboard shows goals, but the edge lives in the shots. Our job is to count them before the game starts.
Technical Breakdown
For the quantitative readers: a deeper look at the architecture and algorithms powering our NHL SOG prop predictions.
Counting Distribution Model
The core probability engine uses a statistical framework designed for modeling discrete event counts. Given the player's projected output for tonight's conditions, the model computes the probability distribution across all possible outcomes and derives P(UNDER) from the cumulative result.
This approach correctly handles the discrete, non-negative nature of shot counts, producing more accurate probabilities than continuous estimation methods.
Data Pipeline
Player shot profiles are rebuilt daily from current-season game logs. Team defensive metrics are computed from historical shot-against data. Both datasets are stored in our database and refreshed before each projection run.
Real-time odds are aggregated from licensed sportsbooks via our odds feed. Both sides (over/under) are paired for each line to enable proper de-vigging and fair probability calculation.
Edge Calculation
Edge is the difference between the model's P(UNDER) and the market's fair (de-vigged) probability. Fair probability is computed by removing the bookmaker's margin from both sides of the line. Only picks where the model sees a probability advantage above the configured threshold are surfaced.
Expected Value (EV) is calculated from the model probability and the actual American odds offered, accounting for the real payout structure.
Model Workflow
5:00 AM ET: Settlement: previous day's picks graded against actual box score results
9:00 AM ET: Projection: full pipeline runs (player profiles, defense metrics, odds fetch, projection, probability engine, filtering)
10:00 AM ET: Lock: top picks ranked and surfaced to subscribers
Post-game: CLV tracking and performance metrics updated
Monthly Access
- Predictions only go live when the model finds true edge
- Closing line value tracked on every prediction so you can verify it yourself
- Covers every market we model and we're always adding more
- Cheaper than your average unit size
Annual Access
- Get 4 months free on us when you go annual
- Every new model we ship is included automatically
- Full platform access for less than most services charge monthly
- Models run 365 days, your subscription should too