pm-AMM: A Uniform AMM for Prediction Markets

The article introduces the pm-AMM, an automated market maker designed for prediction markets whose outcome tokens resolve to $1 or $0. It argues that existing AMMs fit these tokens poorly, because outcome token volatility depends on the current probability and the time to expiration, and liquidity providers are essentially guaranteed to lose their value at expiry. The static pm-AMM is derived from a Gaussian score dynamics model and is described as a uniform AMM, meaning its loss-vs-rebalancing is proportional to portfolio value regardless of price. The dynamic version reduces liquidity over time so that expected loss stays constant until expiration.