Using LLM Agents to Study Trading Strategies in Simulated Stock Markets

Summary

This paper describes a simulated stock market in which large language models act as competing agents with different strategies, information, and endowments. The market includes a persistent order book, market and limit orders, partial fills, dividends, and equilibrium clearing. Agents communicate structured trading decisions while also producing natural-language reasoning. In experiments, the models followed assigned roles such as value investor, momentum trader, or market maker. Simulated outcomes included price discovery, bubbles, underreaction, and strategic liquidity provision.

The framework is presented as a way to investigate financial theories and agent behavior when closed-form analysis is unavailable, and to run experiments that would be costly with human participants. It can also examine how responses change across market conditions and how shared prompts may create correlated behavior that affects stability. These are results within a simulation: the excerpt does not show that LLM agents reliably trade profitably in live markets or that the simulated dynamics fully reproduce real markets.

Key ideas

• The simulated exchange models an order book with market and limit orders, partial fills, dividends, and clearing.
• LLM agents can follow assigned value, momentum, or market-making strategies in the described experiments.
• The simulated market displays price discovery, bubbles, underreaction, and liquidity provision.
• The framework supports experiments on agent behavior and financial theories that lack closed-form solutions.
• Correlated behavior induced by shared prompts may affect simulated market stability.
• Simulation findings alone do not demonstrate profitable performance in live trading.

Tags

# Can Large Language Models Trade? Testing Financial Theories with LLM Agents in Market Simulations


# Can Large Language Models Trade? Testing Financial Theories with LLM Agents in Market Simulations









This paper presents a realistic simulated stock market where large language models (LLMs) act as heterogeneous competing trading agents. The open-source framework incorporates a persistent order book with market and limit orders, partial fills, dividends, and equilibrium clearing alongside agents with varied strategies, information sets, and endowments. Agents submit standardized decisions using structured outputs and function calls while expressing their reasoning in natural language. Three findings emerge: First, LLMs demonstrate consistent strategy adherence and can function as value investors, momentum traders, or market makers per their instructions. Second, market dynamics exhibit features of real financial markets, including price discovery, bubbles, underreaction, and strategic liquidity provision. Third, the framework enables analysis of LLMs' responses to varying market conditions, similar to partial dependence plots in machine-learning interpretability. The framework allows simulating financial theories without closed-form solutions, creating experimental designs that would be costly with human participants, and establishing how prompts can generate correlated behaviors affecting market stability.

Shown in full with attribution under the source's licence. Licence: abstract CC0

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.