Market efficiency in Polymarket's 5-minute BTC markets
2026A market-microstructure study of Polymarket's most competitive market. I replicated the incumbent maker's quote schedule to a median within-slot R² of 0.92, derived a fair value and an inventory-skew rule from first principles, then rejected every candidate edge under placebo, chronological out-of-sample and cost-inclusive tests. The market is efficient at the retail-accessible level and the binding constraint is queue position rather than model quality. The most useful output was a per-slot measurement of how far the replay diverges from live.