SECTIONS
Overview
Moneyball uses a Bayesian prediction framework — the same architecture as our NCAA tournament engine — adapted for MLB's unique statistical richness. A quantitative simulation engine produces the prior probability from 25+ sabermetric inputs, and Claude Sonnet applies qualitative adjustments for factors the math can't fully capture.
Baseball is the most quantifiable major sport. Decades of sabermetric research have produced metrics that genuinely predict performance better than traditional stats. Moneyball's edge is combining the best of these metrics into a unified simulation framework, then layering in real-time context: starting pitcher news, bullpen usage, lineup changes, weather, and sharp money movement.
MONEYBALL PREDICTION ARCHITECTURE
┌─────────────────────────────────────────────────────────────────┐
│ 15+ DATA TOOLS │
│ MLB Stats API · ESPN · The Odds API · Baseball Savant │
│ FanGraphs scrapers · weather API · injury reports │
│ Vegas lines · starting pitcher logs · bullpen availability │
└──────────────────────────┬──────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────┐
│ PHASE 1: DATA GATHERING │
│ Parallel fetches: team stats · starter ERA/FIP/xFIP │
│ · bullpen ERA · lineup · park factors · weather │
└──────────────────────────┬──────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────┐
│ PHASE 2: PYTHAGOREAN BASELINE │
│ Run differential → expected win% (fundamental predictor) │
│ RS²/(RS²+RA²) = Pythagorean expectation │
│ Adjusted for strength of schedule, park effects │
└──────────────────────────┬──────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────┐
│ PHASE 3: PITCHING ADJUSTMENT │
│ SP FIP/xFIP vs opposing lineup wOBA/wRC+ │
│ Bullpen ERA + usage + days rest for each arm │
│ Platoon splits, strikeout rates, walk rates │
└──────────────────────────┬──────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────┐
│ PHASE 4: SITUATIONAL FACTORS │
│ Ballpark run factor · wind speed/direction · temperature │
│ Home field advantage · recent form (L10) · day/night split │
│ Head-to-head matchup history · travel/rest advantage │
└──────────────────────────┬──────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────┐
│ PHASE 5: 10,000-GAME MONTE CARLO │
│ Each sim: compute per-inning run distributions │
│ SP × bullpen × lineup × park × weather × variance │
│ Output: win% ± 1% (95% CI) │
└──────────────────────────┬──────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────┐
│ PHASE 6: QUALITATIVE LAYER (Claude Sonnet ±5pp) │
│ Undisclosed injuries · lineup decisions · narrative context │
│ Sharp money signals · manager tendencies · weather details │
└──────────────────────────┬──────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────┐
│ FINAL PREDICTION │
│ Win% · Moneyline value · Run line edge · Total lean │
│ Parlay eligibility · Confidence tier · Key factors │
└─────────────────────────────────────────────────────────────────┘⚾The Sabermetric Advantage
Baseball's granular statistics — pitch-by-pitch data, batted ball profiles, spray charts, platoon splits — give prediction engines far more signal than any other sport. A pitcher's FIP predicts future ERA better than ERA itself. A team's Pythagorean record predicts future wins better than actual record. Moneyball is built on these insights.