Research database
The research layer organizes the underlying information used to evaluate each slate: game context, starting pitchers, hitters, recent form, matchup strength, first-inning trends and market-specific model outputs.
Current MLB research
- Game Models and First 5 context
- Hit Rate Matrix
- H+R+RBI Opportunity Engine
- NRFI/YRFI matchup research
- Strikeout Matchups
Betting: prediction first, price second
The betting layer ranks qualified signals using model probability, Decision Score, reliability, lineup context and portfolio controls. A wager still has to pass a separate price check because a likely outcome can be overpriced.
Current betting workflow
- Freeze the final pregame Top 20 snapshot
- Compare model probability with the available price
- Keep market performance and realized betting P/L separate
- Grade the exact listed market and line after results
Freshness and lineup confirmation
Morning runs establish an early baseline. Later runs become more actionable as lineups confirm and the model receives more complete slate information. Update time and lineup confirmation stay visible because they are part of the research process.
Historical integrity
Final pregame rankings are preserved for grading. Results are evaluated without postgame re-ranking so model performance can be judged across meaningful samples.
What stays private
The public site receives ranked outputs, summary metrics and selected explanations. Proprietary weighting logic, source code, raw feeds and internal diagnostics remain outside the public website.