Forecasting & AI agents
We study how AI systems — LLMs, ensembles, and multi-agent pipelines — perform as forecasters on real prediction markets, and how to evaluate them rigorously using proper scoring rules.
View work →Research
ForesightFlow organizes its work into five tracks. Each track has its own research questions, methods, and open problems.
Mechanism design
We study how prediction market rules — resolution typology, oracle design, and trading constraints — affect price accuracy, manipulation resistance, and the quality of the information aggregated.
View work →Microstructure
We adapt classical market-microstructure theory — PIN, VPIN, Kyle's λ, order imbalance, and variance-ratio diagnostics — to the discrete-outcome, on-chain CLOB structure of decentralized prediction markets.
View work →On-chain forensics
We develop wallet clustering, funding-flow analysis, novelty scoring, and cross-market behavior methods to attribute trades to economic agents rather than pseudonymous addresses.
View work →Quantitative strategies
We study alpha generation, market making, arbitrage, and execution on the hybrid CLOB structure of decentralized prediction markets — venues with unusual liquidity dynamics and binary payoff structures.
View work →