Modelling the trading behaviour in high-frequency markets

We use an agent-based approach to model trading behaviour in high-frequency markets. This study focuses on the Foreign Exchange (FX) market. The initial part of this study is to observe the micro-behaviour of traders to define the stylized facts of their trading activities. This is performed using a high-frequency dataset of anonymised individual traders' historical transactions on an account level, provided by OANDA Ltd. This dataset is considered to be the biggest available high-frequency dataset dealing with the individual FX market traders' trading activities. The second step is to build agent-based models of traders. The traders are modelled to respond to physical time to account for the different market seasonalities. The key elements in modeling the traders are zero-intelligence directional-change events trading strategy, historical prices, actual FX market traders' behaviour, limit orders, FX market trading sessions and market holidays. Using the identified stylized facts of FX market trading activity, we evaluate the trading agents' collective trading behaviour. The results of this comparison indicate that the trading agents' collective trading behaviour resembles, to a certain extent, the collective trading behaviour of FX market traders.