Using Simulated Limit Order Book Data To Evaluate Trade Execution Strategies

Abstract

The electronic nature of markets today allows traders to infer information from limit order book (LOB) data and execute trades based on this information. Agents such as hedge funds are often interested in solving what is known as the optimal trade execution problem, where they attempt to minimize cost and market impact when buying or selling a large amount of inventory. In this thesis, historical LOB data is used to simulate a market environment in which trade execution strategies are tested. First, high-frequency LOB data from the Coinbase Pro cryptocurrency exchange is used to create a modified version of the queue-reactive model developed by Huang et al. (2013). A market simulator is built based on this model where market dynamics update in real-time in response to the agent's actions. The performances of common trade execution strategies are then evaluated in the simulated market environment. The findings from this thesis are relevant for institutions who are interested in using a simulated market environment to test trading strategies before executing them.