NHL Desk Strategy Spec

Version: 5.1 | Date: 2026-04-03 | Status: DEPLOYED

Core Strategy

Trade mispriced Kalshi NHL contracts using Pinnacle as the sharp anchor. De-vig Pinnacle odds to fair probability, compare against Kalshi price, execute when net edge exceeds 4 cents after 7% Kalshi fee.

De-Vig Methods

2-Way Markets (Spreads, Totals) — Shin Method:

fair_prob = (sqrt(z² + 4(1-z)p²) - z) / (2(1-z))

Solved numerically via brentq. When Power and Shin diverge > 2 cents, average them. Otherwise Shin only.

Moneylines — Shin Method (primary): Same formula. Handles lopsided NHL lines better than multiplicative de-vig.

Kalshi Fee & Edge Calculation

win_amount = 100 - kalshi_price
fee_on_win = 0.07 * win_amount
net_edge = (model_prob * (win_amount - fee_on_win)) - ((1 - model_prob) * kalshi_price)

Minimum threshold: 4 cents net edge post-fees.

NHL Distribution Models

Poisson (Goal Totals):

• Lambda calibrated from Pinnacle O/U line
• Search range: [3.0, 9.0]
• OT adjustment: +0.15 to lambda
• Max goals cap: 15

Normal (Spreads):

• Mean calibrated from Pinnacle puckline
• Std dev: 2.0 (fixed)
• Alt lines: -0.5, -1.5, -2.5, -3.5

Market Priority

Edge Scanner Schedule

Data Collection Schedule

Kalshi Series Tracked (15)

NHL-Specific Edge Factors

Goalie Factor (Primary Edge Driver)

• Starter vs backup shifts outcome by 15-20 cents on Kalshi
• Source: DailyFaceoff (4x/day scrape) + NHL API
• AHL callup = +4-5 point O/U swing
• Confirmation tiers: Confirmed (full), Expected (50%), Unconfirmed (no bet)

Back-to-Back Games

• ~3-5 cents underperformance
• If BOTH teams on B2B, effects cancel — skip

Referee Assignments

• Source: ScoutingTheRefs (game day scrape)
• Penalty-heavy refs affect totals and pace

Custom Metrics (10)

1. Goalie-Adjusted xGA — Team_xGA × (Starter_sv% / League_avg_sv%)
2. Special Teams Impact Score — (PP% - Lg_PP%) × PP_opps/game + (PK% - Lg_PK%) × PK_times/game
3. Net Goalie Matchup Delta — Starter_A_GSAx/60 - Starter_B_GSAx/60
4. Fatigue xG Penalty — B2B = -0.25 xG; 3-in-4 = -0.35 xG; with travel modifier
5. PDO Regression Flag — 5v5 PDO outside 0.985-1.015 flagged
6. EWMA Form — 10-game window, alphas 0.10/0.12/0.15 tracked simultaneously
7. Comeback Probability Index — 3rd period scoring + trailing performance + EN pull tendencies
8. Score-State Adjusted Metrics — Separate xGF/xGA when leading, tied, trailing
9. Starter-Backup Delta — Save% gap between starter and backup
10. Playoff Context Flags — Clinched/eliminated/magic number/games remaining

Position Sizing (Quarter Kelly)

full_kelly = (model_prob * net_odds - (1 - model_prob)) / net_odds
bet_size = min(bankroll * full_kelly * 0.25, bankroll * 0.05)

Steam Detection

Rung Distance (Tail Confidence)

Hard No-Bet Rules

• Unconfirmed goalie
• 5+ rung distance
• Kalshi volume < $500
• Net edge < 4 cents
• Implied prob < 15% or > 85%
• Both teams on B2B
• Injury news breaking mid-scan

Key Constants

Settlement & Learning

• Runs at 4AM/10AM/4PM ET
• Tracks per edge: game_id, market_type, kalshi_price, model_prob, raw_edge, net_edge, rung_distance, goalie_status, steam_flag
• Post-settlement: Brier score, CLV tracking vs Pinnacle closing
• Monthly review: P&L per market type, edge accuracy, rung distance performance

Deferred to Later

Recently Completed (Apr 3, 2026)