Agentic Market Simulator
A large-scale agentic market simulator, where thousands of AI agents trade as a real market would.
Who it's for.
Stress-test
Risk teams at banks
Test how macro events would hit your portfolio and strategy before they happen.
Backtest
Trading desks and hedge funds
Backtest strategies across many simulated historical paths, not just the one history gave us.
Anticipate
Central banks
See how markets might react to policy decisions and shocks before they are made.
Available as a licensed platform, on-premise or hosted.
How we built it.
A six-part series on building the simulator, from the matching engine to policy design.
Episode 1
The Engine and the Environment
A call-auction clearing engine, benchmarked against zero-intelligence traders before any model is involved.
Episode 2
Universe, Information and Leakage
Real S&P 500 data, a test for lookahead bias, and a probe for whether agents can re-identify anonymized companies.
Episode 3
Actions, Personas and Track Record
An attention layer that decides which agents trade and why, grounded in the no-trade theorem.
Episode 4
Beliefs, Personas and Track Record
Turning model probabilities into valuations and conviction, then running a 22-agent market for a simulated year.
Episode 5
Scaling to a Large Number of Agents
2,000 agents trading the 100 largest US companies from 2019 to 2022, checked against stylized facts of equity returns.
Episode 6
Policy Design: A Short-Sale Ban
Using the simulator to study a policy question: what happens to the market when short selling is banned.