Beyond Forecasting: The Belief-to-Trade Layer in Prediction-Market Agents

2026-07-03

AI Summary

This work addresses the persistent gap between well-calibrated probabilistic forecasts and profitable trading strategies in prediction markets. While existing models can generate accurate probability estimates, they often fail to translate these into effective trading decisions. To bridge this divide, the paper introduces Raven-Agent, the first end-to-end autonomous trading agent specifically designed for prediction markets. Raven-Agent incorporates a novel “belief-to-trade” transformation layer that directly maps calibrated probability predictions into order placement decisions, integrating modules for probability calibration, risk adjustment, and order generation within a unified architecture. The agent is trained via reinforcement learning with historical decision replay. In controlled backtesting, Raven-Agent is the only strategy that simultaneously achieves positive raw returns and positive risk-adjusted returns, significantly outperforming all existing benchmarks.

Abstract

Forecasting future events has attracted growing attention as a testbed for general-purpose AI. A natural way to ground this evaluation is let the models trade in the prediction markets. Trading, however, requires more than forecasting. Moreover, recent benchmarks report a substantial gap between calibrated probability scores and the trading results. We propose Raven-Agent, to the best of our knowledge, the first autonomous trading agent for prediction markets. On a controlled replay over an archived decision set, our architecture achieves the only positive return and the only positive risk-adjusted return among all tested policies. We have released our code in https://github.com/Alchemist-X/predict-raven .

Problem

• prediction markets
• forecasting
• trading agents
• probability calibration
• AI evaluation

Innovation

• prediction markets
• autonomous trading agent
• belief-to-trade layer
• risk-adjusted return
• AI forecasting