Market data & features
Build time-aware datasets from prediction-market prices, order books and onchain activity. Study how information moves through markets and which observable patterns may carry a signal.
Polygraph develops prediction-market trading strategies using machine learning to investigate potential information advantages and test their value through paper simulations.
Market participants differ in what they know, how they interpret it and how quickly they act. We study asymmetric information: whether observable trading activity and market responses reveal signals that prices have yet to fully reflect.
Using market and onchain data from prediction markets such as Polymarket, we investigate pricing discrepancies and translate candidate signals into rules for when to enter, which side to take, how much simulated capital to allocate and when to exit.
Our research goal is positive risk-adjusted returns after costs. Strategies must be evaluated on later, unseen data and through paper simulations; profitability remains to be demonstrated.
How could this generate revenue?
If the research produces useful, reliable signals, potential commercial applications could include paid access to market analytics and research dashboards, API subscriptions for model outputs, or licensing analytical tools to other teams.
These are possible applications of the research, rather than announced offerings or commitments to launch. Their viability would depend on demonstrated usefulness, demand and applicable requirements. Current evaluation remains limited to paper simulations.