“integer linear programming cuts parlay collateral by 10-80% without sacrificing settlement safety”
Eric Liu (co-founder @totalistrading) explains how prediction market market makers can reduce required collateral for parlay tickets by computing worst-case portfolio exposure with integer linear programming instead of collateralizing each ticket in isolation. He walks through the ILP formulation, shows how it avoids the infeasible 2^n enumeration, and reports 10-80% collateral reduction from backtests against Totali's markets using HiGHS/scipy. The post argues this enables tighter quotes, deeper liquidity, and more expressive combo bets without sacrificing fully collateralized peer-to-peer settlement.
Some technical background helpful
Platforms mentioned: Kalshi