Multi-Agent Crypto Arbitrage Predictor
Uses a time-series foundation model and multi-agent framework to predict 5-minute BTC/ETH trends and arbitrage across Polymarket and Kalshi
A production-grade system that combines multiple AI agents within the Hermes Agent framework to execute cross-platform arbitrage. It sources data programmatically via Apify and feeds it into the Kronos Time-Series Foundation Model, which predicts 5-minute close trends for BTC and ETH.
Entry: When the model predicts a significant price movement with high confidence, the system identifies price discrepancies between Polymarket and Kalshi for the same crypto event. It executes paired trades to capture the spread. Sizing uses the fractional Kelly Criterion to optimize position sizes while managing risk.
Exit: Positions are closed when the markets converge or when the prediction window expires. The system is designed for automated, high-frequency trading with strict risk management rules. It is still in development but shows promise for latency-sensitive arbitrage.