The market has a price.
Oracle has a probability.
Project Oracle is an AI-powered prediction engine that analyzes event derivatives and identifies contracts whose market prices differ from statistically-derived probabilities. The objective is not to predict the future — it is to know when the market is wrong.
"Does our probability differ from the market enough to create positive expected value?"
How Oracle works
Prediction markets are efficient — but not perfectly efficient. Oracle runs a disciplined pipeline that turns raw market data into calibrated probability and expected value.
What an edge looks like
When Oracle's probability diverges from the market's implied probability, that divergence is the product — quantified, explained, and auditable.
Roadmap
Seven phases, from data warehouse to probabilistic intelligence.
Market Scanner
Connect to Kalshi. Pull all active contracts. Store them. Display them. The data warehouse.
Prediction Engine
One category first — ATP tennis. Rankings, Elo, surface, form, head-to-head, fatigue.
AI Research Agent
LLMs read news, injuries, travel, interviews — and emit structured features. Never predictions.
Probability Engine
Independent models — gradient boosting, Bayesian, Monte Carlo — combined by an ensemble.
Trading Intelligence
Not "bet on Djokovic" — the why: expected value, confidence, and enumerated reasons.
Portfolio Manager
Kelly criterion, position sizing, correlation detection, exposure and loss limits.
Autonomous Trading
Optional. Rule-bound. Every decision logged, every prediction replayable.
Coverage
Starting with sports and expanding across every event market Oracle can model.