Library/Market Making for Prediction Markets: A Probability-Space Approach
MicrostructureAnalysis

Market Making for Prediction Markets: A Probability-Space Approach

XO Labs·April 17, 2026·Twitter
a logit transformation lost $1,114, but a probability-space engine turned profitable at $453

Why It's Worth Reading

Technical research on adapting the Avellaneda-Stoikov market-making framework for prediction markets. Classical models fail because prediction market prices are bounded probabilities (0 to 1) rather than unbounded asset prices, creating non-constant volatility and guaranteed terminal convergence. After a logit-space transformation lost $1,114 in backtest, XO Labs iterated to a probability-space engine with inventory-skewed spreads, volatility regime detection, and multi-outcome coordination that turned profitable at $453. Includes the full mathematical framework and open problems.

Extensive technical background assumed

Concepts

Platforms mentioned: Polymarket, NEW: XO Market

Related Reading