← Models / SOG Model

How the Shots on Goal Model Works

What it does

For every player Hard Rock Bet posts a shots-on-goal line for, the model simulates the game 10,000 times and counts how often the player goes over. That gives a probability for every shot count (0, 1, 2 … 10+) — the bars under each pick — and the probability of going over Hard Rock's line.

Model vs Sim: the probability the picks use (Model) blends the simulation with Hard Rock's own de-vigged price, weighted 35% simulation / 65% market. In a two-season backtest the simulation alone was about as accurate as the closing line, but the blend was measurably more accurate than either — each knows things the other doesn't. The weight is re-fitted every week on graded results. Sim (in each row's detail) is the raw simulation.

How a simulation works

  • Game script — each simulated game draws how many power plays each team gets, how the score unfolds (from Hard Rock's moneyline and total), whether a trailing team pulls its goalie late, and whether it goes to 3-on-3 overtime. Every player on a team shares the same script, so a power-play-heavy night lifts the whole PP unit together.
  • Ice time — the player's share of his team's even-strength, power-play, penalty-kill, 6-on-5 and overtime minutes, from his recent deployment, plus a realistic random swing drawn from real games (including nights cut short by injury).
  • Shot rate — shots per 60 minutes in each situation, estimated from his history and pulled toward forward/defenseman norms when the sample is small. Adjusted for the opponent's shot suppression, the arena's scorekeeper tendencies, home ice, back-to-backs, and score state (trailing teams shoot more).

Reading the board

  • Book — Hard Rock's probability for the over with the vig removed.
  • Model — the model's probability for the over (simulation blended with the market, see above).
  • Proj — average simulated shots; TOI — expected minutes (power-play minutes in brackets).
  • EV — expected profit per 1 unit bet at Hard Rock's price on the model's side.
  • Fair — the price at which the bet would break even by the model.

Best bets & leans

Best bet: EV of at least 5% and the model at least 4 percentage points away from the de-vigged line. Lean: EV of at least 2%. Only lines with both an over and an under price can be picks. Stake is 0.25× Kelly as a share of bankroll — a sizing guide, not a promise.

Settlement & timing

Graded like the sportsbooks grade it: overtime counts, shootouts don't, and a player who doesn't take a shift is a void. Predictions refresh every 30 minutes on game days as lines move and lock at puck drop; the locked prediction is what gets graded. Every pick is tracked on Pick history and Performance, including how the line moved after the pick (CLV).

Honest expectations

Shots on goal are noisy — a 3-shot-per-game player still goes under 2.5 often. The model aims to be well calibrated, not to call every game. Results before live tracking come from a walk-forward backtest priced at Hard Rock's closing lines, using only information available before each game.

Informational only. Not betting advice.

Games
1
Lines priced
11
Best bets
3
Leans
2
Updated
Oct 19, 6:03 AM CT
backtest · priced at close

Best Bets

Stefan Noesen
NJD @ LAK · 5:00 PM CT · F
UNDER 1.5
Price+150
Model46.1%
Book37.5%
EV+15.3%
Fair+117
Stake2.5%
0
1
2
3
4
5
6
7
8
9
10+
Proj 1.63 SOG · 15.4 min (2.5 PP) W 0 SOG · +1.50u
Timo Meier
NJD @ LAK · 5:00 PM CT · F
UNDER 2.5
Price+100
Model54.6%
Book46.9%
EV+9.1%
Fair-120
Stake2.3%
0
1
2
3
4
5
6
7
8
9
10+
Proj 2.28 SOG · 16.7 min (1.6 PP) L 3 SOG · -1.00u
Jack Hughes
NJD @ LAK · 5:00 PM CT · F
UNDER 3.5
Price-110
Model55.7%
Book49.0%
EV+6.4%
Fair-126
Stake1.8%
0
1
2
3
4
5
6
7
8
9
10+
Proj 3.18 SOG · 21.0 min (3.1 PP) L 4 SOG · -1.00u

Full Board

PlayerGameLineOverUnder BookModelProjTOI (PP) PickEVResult
NJD @ LAK 5:00 PM CT 1.5 -200 +150 62.5% 53.9% 1.63 15.4 (2.5) UNDER best +15.3% W 0 SOG · +1.50u
LAK vs NJD 5:00 PM CT 2.5 -145 +110 55.4% 52.3% 2.71 20.2 (3.2) — -11.6% 2 SOG
LAK vs NJD 5:00 PM CT 1.5 -120 -110 51.0% 51.4% 1.76 18.2 (1.5) — -5.8% 2 SOG
NJD @ LAK 5:00 PM CT 2.5 -130 +100 53.1% 45.4% 2.28 16.7 (1.6) UNDER best +9.1% L 3 SOG · -1.00u
NJD @ LAK 5:00 PM CT 3.5 -120 -110 51.0% 44.3% 3.18 21.0 (3.1) UNDER best +6.4% L 4 SOG · -1.00u
NJD @ LAK 5:00 PM CT 2.5 -105 -130 47.5% 41.9% 2.18 19.9 (2.7) UNDER lean +2.7% L 4 SOG · -1.00u
NJD @ LAK 5:00 PM CT 2.5 +100 -135 46.5% 41.8% 2.19 19.3 (2.8) — -16.5% 3 SOG
LAK vs NJD 5:00 PM CT 2.5 +105 -140 45.5% 40.1% 2.15 17.7 (1.9) UNDER lean +2.7% L 3 SOG · -1.00u
NJD @ LAK 5:00 PM CT 2.5 +125 -170 41.4% 38.5% 2.15 19.2 (2.8) — -13.4% 5 SOG
LAK vs NJD 5:00 PM CT 2.5 +130 -175 40.6% 36.7% 2.05 16.1 (0.8) — -15.7% 1 SOG
LAK vs NJD 5:00 PM CT 3.5 +135 -180 39.8% 34.9% 2.78 20.6 (2.7) — -18.1% 2 SOG

Click a row for the shot distribution and model inputs. Hard Rock lines only. Model fitted through 2026-10-18.