Stockyard: A discrete event-based stock market exchange simulator

Abstract

We describe an agent-based stock market simulator built using an asynchronous discrete event simulation framework. The simulator is unique in that it's driven by real-world financial algorithms and protocols; and it's open source. It utilizes an order book bid and ask matching model, and real-world exchange protocols. Our simulation is based on multiple agents interacting through an exchange agent. This method is distinct from those that utilize historical pricing data. Order book execution supports a more realistic interaction between agents. Pricing in our model arises from the dynamics of matching orders in the order book. Our simulator enables the study of market dynamics, and trading strategies using real-world exchange protocols. We present our design and implementation of a market simulator and discuss our initial results using the message protocols defined by NASDAQ: OUCH and ITCH. Our initial results demonstrate StockYard's capability and efficiency in simulating markets with realistic trade volumes.