The paper introduces ABIDES (Agent-Based Interactive Discrete Event Simulation), a novel simulation environment designed to support high-fidelity market simulations for AI research. Unlike proprietary tools used internally by trading firms, ABIDES is broadly available and aims to facilitate AI agent development in financial markets. It simulates tens of thousands of trading agents interacting with an exchange agent, incorporating configurable network latencies and modeling NASDAQ's equity trading protocols (ITCH and OUCH) to ensure realism. The simulator's message-based design and validation through example scenarios demonstrate its capability to replicate complex market dynamics.

ABIDES is applied to develop a market impact model, showcasing its practical utility in quantitative finance research. The paper highlights its potential for exploring future experimental problems, such as the development of machine learning-based trading algorithms. By providing an open, high-fidelity platform, ABIDES addresses a gap in available tools and aims to accelerate AI research in market applications, offering a foundation for testing and refining trading strategies in simulated environments that closely mimic real-world conditions.