In this thesis we propose an agent-based model for a financial market with a
single asset. The agents are motivated to trade via their personal beliefs about the
future direction of the asset price moves. Additionally, the trades are restricted by
the resources available to agents. The constructed model is used to attempt to gain
some insight into the origin of large price moves in the market (“market crashes”).
Monte Carlo simulations are used to study model behaviour under varying initial
conditions. The model is found to be generally capable of reproducing the stylised
facts of real financial markets. The ubiquity of relatively high incidence of large
price moves in the results of model simulation, together with results from similar
models by other authors allow us to conjecture that such moves are inherent in a
market model based on a heterogenous population of intelligent agents. Finally,
several directions for model improvement are identified.