An agent-based simulation of double-auction markets

An agent-based simulation of double-auction markets Ted Xiao Guo Master of Science Graduate Department of Computer Science University of Toronto 2005 Agent-based simulation has emerged as a promising research alternative to the traditional analytical approach in studying financial markets. The idea behind agent-based simulation is to constructively model a market in a bottom-up fashion using artificial agents. This thesis aims to accomplish four objectives. The first objective is to present a high-quality software platform that is capable of carrying out agent-based financial market simulations. The second objective is to simulate a double-auction market with this software platform, using homogeneous zero-intelligence agents. The simulation is based on the artificial market model proposed by Smith, Farmer, Gillermot and Krishnamurthy (2003). The third objective is to study the microstructure properties of the simulated double-auction market. In particular, we investigate certain statistical properties for the mid-price, bid-ask spread and limit order book. The fourth and the last objective is to explore and evaluate trading strategies associated with the time-constrained asset liquidation problem through computational agents.