Prediction Markets
LucidiMarket
LucidiMarket is the Institute's research program on prediction markets: how markets aggregate dispersed beliefs into probabilities, when that process stays informative, and when it breaks. We pair theoretical work with large-scale agent-based simulation, and publish the manuscripts that come out of it.
Research
We study the mechanisms, limits, and failure modes of prediction markets as instruments for collective intelligence: price formation, information aggregation, arbitrage capacity, trader composition, and systemic risk. The aim is a rigorous account of when market prices can be trusted as probabilities.
Simulation
Live markets never reveal interim truth, so we build controlled simulators with a known latent probability path. Our stack spans a microstructure limit order book and a vectorized PyTorch batch-auction engine that scales to a million agents on local hardware, giving us ground truth to measure market behavior against.
Manuscripts
Working papers and methods companions from the program, including studies of composition-driven systemic risk and the reproducible simulation stack behind them.
Read the manuscripts