OddsEngine

OddsEngine watches live basketball and acts only at pre-defined moments — entry points described by game state rather than a model guessing at every tick. We built the live pipeline, the evaluation layer and the priors behind it, and the validation work that had to pass before any of it was trusted with a real game.

• Live worker loop consuming in-game feeds and evaluating the configured entry points against current state
• Entry points and thresholds held in the database rather than the code, so a change does not need a deploy
• Priors served from the database, with a daily update job and alerting for the case where it silently does not run
• Admin dashboard for configuration, monitoring, and reading back what fired and why
• An end-to-end validation suite — feeds, priors, evaluation, dashboard — run before the system is pointed at a live game
• Redis and PostgreSQL behind Docker Compose, with alerting on the paths that must not fail quietly