How we translate advanced sabermetrics, simulation data, and market analysis into actionable betting edges. This is where analytics stop being academic and start being profitable.
Our prediction engine processes dozens of data points per game, from starting pitcher xFIP to bullpen fatigue metrics, team-level wOBA splits, park factors, and weather modeling. The goal is not to predict who wins. The goal is to predict where the market is wrong.
Traditional handicapping relies on gut feel, recent form, and narrative. Our model ignores all of that. It looks at what actually predicts outcomes: pitcher quality (xFIP, SIERA, K-BB%), offense production (wOBA, barrel rate, chase rate), and systemic factors (travel, rest days, altitude).
When the model says a team has a 58% win probability but the market prices them at 52%, that is the edge. That gap between model probability and market probability is where every profitable bet lives.
We evaluate starters using xFIP (removes home run variance), SIERA (accounts for batted ball data), K-BB% (the most stable pitching metric), and CSW% (called strikes plus whiffs). These metrics strip out luck and noise to reveal true pitcher quality. When a starter's ERA is 3.20 but his xFIP is 4.40, we know regression is coming, and the market is usually slow to price it in.
Hitting is assessed through wOBA (weighted on-base average), barrel rate (the best predictor of power output), chase rate (discipline under pressure), and platoon splits. A team that ranks 5th in runs scored but 18th in wOBA against same-side pitching is a team that is about to come back to earth.
Every game is simulated thousands of times using Monte Carlo methods. The simulation accounts for lineup-by-lineup matchup data, bullpen availability chains, and in-game leverage scenarios. This produces a probability distribution, not a single prediction, which is critical for properly sizing bets and identifying value in totals markets.
The model's probability output is compared against closing lines from major sportsbooks. We track CLV (closing line value) as our primary measure of model accuracy, because beating the closing line is the only reliable predictor of long-term profitability in sports betting.
Each component of our model is documented in detail. These guides explain not just what we do, but why each metric matters for bettors specifically.
The full breakdown of how predictions are generated, from data ingestion to final probability output. Links to every sub-component.
EngineStep-by-step walkthrough of the prediction pipeline, from raw data to game-day projections.
ArchitectureThe design philosophy behind our model architecture. Feature selection, training data, and validation methods.
MetricsWhich sabermetrics predict outcomes and which are noise. Data-driven analysis of metric predictive power.
SimulationHow we simulate games thousands of times to generate probability distributions and identify value in totals and props.
ModelsThe ecosystem of models we run and how they interact. Ensemble methods and model weighting.
AccuracyUnderstanding the inherent uncertainty in MLB prediction. Why even good models "lose" 40%+ of games.
VarianceThe relationship between expected value and short-term results. Why variance is not a model failure.
MarketHow our projections compare against market-implied probabilities. Where the market misprices and why.
UncertaintyQuantifying prediction confidence intervals. Why probability ranges matter more than point estimates.
CLVThe gold standard of betting model evaluation. How we measure and track CLV across markets.
FundamentalsCore concepts every bettor needs: bankroll management, Kelly criterion, expected value, and line shopping.
We document every prediction and every result. No disappearing picks, no after-the-fact edits. Our Track Record page shows the complete history, including losses.
Transparency is not optional for a betting model. If you cannot verify performance independently, the model is worthless. We publish timestamped predictions before games and grade them publicly after.
Check our Trends page for historical patterns and our Daily Analysis for current model output with full breakdowns.