Sports Betting Models
11 purpose-built Monte Carlo simulation engines, each calibrated for its league's specific dynamics. 10,000+ iterations per game with Kelly Criterion sizing and Bayesian calibration.
Simulation Engines
Each league uses a dedicated engine tuned to its specific data sources, scoring dynamics, and betting markets.
NBA — Possession MC V5.0.2
Possession-by-possession simulation with Beta shooting distributions, score-state dynamics, and 13-player roster modeling. Each game plays out ~200 possessions per simulation across 10,000 iterations.
College Basketball — Savant Ultra v5.0.1
5-on-5 player-level simulation using EvanMiya BPR ratings, adaptive ATS regime thresholds, and matchup variance modeling across 350+ Division I teams.
NHL — V19.1 Pinnacle
MoneyPuck expected goals (xG) integration with real danger zone data, per-zone goalie modeling, and Bayesian shrinkage for shot quality-weighted simulation.
NFL — Elite V1.1 Pinnacle
EPA-driven drive-level simulation with CDF edge calculation, weekly automated roster updates, and play-by-play modeling calibrated to current season data.
Soccer — V16.3 PBP
FBref xG-calibrated play-by-play simulation with isotonic calibration, formation analysis, 3-way moneyline support, and Both Teams To Score (BTTS) modeling.
MLB — Elite Matchup V4.2
Statcast-powered count-state simulation with pitcher-batter matchup modeling, catcher framing adjustments, bullpen fatigue tracking, and park factor normalization.
League of Legends — Championship v2.1
5-layer Glicko-2 rating system with market blend, staleness regression, and patch-aware meta analysis across LCK, LPL, LEC, LCS, and international events.
Tennis — Elo Monte Carlo v2
Hierarchical point-game-set-match Markov simulation with matchup-adjusted serve modeling, surface-specific Elo, empirical momentum coefficients, and a confidence-band abstain gate.
Golf — Tournament MC V3.3
Full four-round tournament simulation with course-archetype strokes-gained weighting, Data Golf variance modeling, dead-heat cut tracking, and field-shared noise correlation.
Counter-Strike 2 — Glicko-2 Map MC v0.1
Glicko-2 team ratings with per-map Monte Carlo simulation, real pre-match veto capture, and series aggregation for match winner, map handicap, and total-maps markets.
WNBA — Possession MC v0.5
Purpose-built possession simulation on Four Factors with opponent-adjusted SRS ratings and small-sample Bayesian shrinkage. Projections currently free while the tracked record accrues.
College Football — PBP v3
Play-by-play Monte Carlo from state-keyed empirical kernels fit on 862,000+ integrity-screened plays. In shadow validation — no picks published until it demonstrates out-of-sample skill.
Shared Infrastructure
Every engine shares the same calibration, sizing, and quality control systems.
Kelly Criterion Sizing
Optimal bet sizing calculated from the edge magnitude and implied probability. League-specific caps prevent overexposure in volatile markets.
Bayesian Calibration
Raw model probabilities are calibrated against thousands of historical outcomes using isotonic regression and Bayesian shrinkage to correct for overconfidence.
Profitability Zone Analysis
15-dimension sub-niche analysis identifies which specific bet types, leagues, and market conditions are currently producing edge and which have gone cold.
Transparent Tracking
Every pick is recorded at publication time. Results are resolved automatically. Full history available on the track record page.
Free calculators & how to read projections
Turn model probabilities into stakes with the free toolset, then verify process on the public track record.