Cross-Platform Prediction Market Arbitrage (Polymarket vs Kalshi)

How It Works The same real-world event (e.g., an election outcome or economic indicator) is often listed on multiple prediction market platforms simultaneously. Because user bases and liquidity profiles differ, transient price divergences appear. The system monitors both platforms in real time, identifies matching market pairs, and when the spread exceeds combined fees and slippage, it buys the cheaper side and sells the more expensive side.

Entry / Exit Conditions Entry fires when the absolute price difference between the two platforms for an equivalent outcome exceeds a configurable minimum edge threshold that accounts for USDC gas/settlement costs, platform fees, and expected slippage. Kelly sizing determines position magnitude based on estimated edge and variance. Exit occurs at resolution or when the spread compresses to near zero.

Verified Results No explicit live trading results are disclosed. The presence of an embedded analytics dashboard implies ongoing monitoring and iterative refinement.

Implementation Notes Built in async Python for low-latency simultaneous polling of both APIs. Risk management is integrated at the execution layer, not bolted on afterward. The unified PMXT API (pmxt-dev/pmxt) could serve as an abstraction layer for this type of cross-platform system.

Risks The primary risk is settlement asymmetry: the two platforms may resolve identical-sounding markets differently due to differing rule language, creating a position that is long on one platform and unhedged on the other. Latency differences between platforms can also cause one leg to fill at a stale price. Regulatory changes affecting either platform create tail risk.