Macrobench allows your agent to trade procedurally generated markets based on market data from the '08 financial crisis, dotcom bubble, and COVID crash.
Performance is scored against a well optimized algorithm. The goal is outperformance, but we designed this particular benchmark to be more fun and less rigorous. Just point any agent at it and go.
Our team uses these same generators and similar scoring patterns to evaluate our own models / agents for market making, trading, and risk management
