How DiamondEdge Works
Transparent methodology — we show our work
The Engine

DiamondEdge uses a Monte Carlo game simulation engine to project every MLB game. Instead of plugging numbers into a formula, we simulate the actual game — plate appearance by plate appearance, inning by inning — thousands of times. Each simulation produces a full box score, and the aggregate of all simulations gives us our projections.

This approach captures context that formula-based models miss: lineup order effects, bullpen sequencing, platoon matchups, park factors adjusted for weather, and the natural variance of baseball.

Simulation Pipeline
01
Player Projections
Ensemble of 6 industry projection systems (Steamer, ZiPS, ATC, etc.) using Marcel-style weighted averages. Multi-year regression with 5-4-3 weighting.
02
Matchup Engine
Odds-ratio method combines batter and pitcher rates with league averages. Platoon splits applied for L/R matchups. Park factors on FanGraphs 100-scale.
03
Weather Integration
Real-time weather data from Open-Meteo for all 30 stadiums. Temperature, wind speed/direction, humidity, and altitude all affect ball flight and run scoring.
04
Game Simulation
2,000+ Monte Carlo simulations per matchup. Each sim plays a full 9-inning game PA by PA with realistic baserunner advancement, bullpen usage, and batting order effects.
05
Props & Edges
Player props emerge naturally from the simulation — not from isolated stat models. HR, K, H, RBI, runs, total bases, and stolen bases all flow from game context. Edges identified by comparing sim probabilities to market-implied odds.
Model Performance
HR Projection MAE
1.45 vs Marcel baseline 4.43
67% better than standard projection methods
Run Scoring Calibration
8.6 runs/game (MLB 2024 avg: 8.58)
Simulation Count
2,000+ games per matchup
Projection Sources
6 industry systems ensembled
Honest Backtesting

We validated our model against 245,958 real game-bets from the 2024 MLB season using actual Retrosheet box scores. Here's what we learned:

Blind Betting Loses
-13.6%
ROI betting all overs regardless of edge
Vig kills you. You need actual edge.
3%+ Edge Works
+4.8%
ROI on bets with 3%+ model edge
8,335 bets. Real, sustainable profit.
🎯
Calibrated
1-3%
Actual outcomes within projected probability
When we say 60%, it happens ~60% of the time.
Prop Performance by Type

We tested every prop type against real 2024 results. Here's what we found:

🔥 Best Edges
Singles +15.5% | Doubles +14.9% | Pitcher HA +14.9%
Softer markets, bigger edges
⚠️ Sharp Markets
Pitcher Ks -1.6% | Batter Ks -1.3%
Tight lines, harder to beat
📊 Home Runs
Most popular prop, moderate edge
We project every HR, show best spots

Our approach: We show you props across all types — HR, K, hits, bases — but only the ones with conviction 60+ (high-quality edges). Some markets are sharper than others, so we're pickier there. You see the best bets, period.

What Makes Us Different
Context-Aware Props

Most prop models project each player in isolation. Our props emerge from full game simulations — so a batter's projected stats account for who's pitching, the park, the weather, where they bat in the order, and who's hitting around them.

Transparent Methodology

We don't hide behind a black box. Every projection shows you the inputs and the math. If our model says a player has a 35% HR probability, you can see why — and decide if you agree.

MLB Obsessed

We don't spread across 6 sports and do them all poorly. We do one sport — baseball — and we go deeper than anyone. Every feature is built specifically for MLB betting.

Weather as an Input

Real-time weather data feeds directly into our simulation engine. Temperature, wind, and humidity affect ball flight, which affects HR rates, total bases, and run scoring — all reflected in our projections.