Durham University, Durham, UK
Detection of Algorithmic Trading
There are ongoing concerns on the growth of algorithmic traders and their effects on the ability of financial markets to perform efficiently their functions. In the wake of increasing frequency of market crashes in major markets, market regulators explore methods to monitor the activity of these fast traders, and their effects on financial markets. In this paper, we propose a method to identify quote volatility episodes consistent with specific types of strategies employed by algorithmic traders, such as undercutting behaviour and quote stuffing. The two examples of behaviour will likely have the effect of increasing quote volatility and execution costs. We propose a ratio to quantify these patterns in the millisecond environment across several asset classes. Specifically, the ratio captures the rapid changes in price quotes and expressed as the rate of oscillation of the best ask and the best bid over extremely short period. We further carry a two-stage Artificial Neural Network experiment on this ratio in order to detect potential commonalities, which may signal that an episode of algorithmic activity is imminent. There are reasons to believe that at least part of algorithmic activity is predictable to some extent. Fundamentally, algorithms are triggered by changes in market conditions. If these conditions are known, it would be possible then to forecast when algorithmic activity is imminent. However, this information is almost certain to be protected as intellectual property because algorithmic traders see the implementation of their strategy as the source of their competitive advantage and naturally hide their algorithms. We further demonstrate that Artificial Neural Network is a useful technique to detect commonalities in market conditions immediately preceding an episode of algorithmic activity. ANN results suggest that the quote volatility ratio appears to be a good filter for signals, and an increase of the ratio threshold seems to improve the detection in ANN but only for some levels.
I am an assistant professor of finance at Durham University Business School. My research interests are in market microstructure and span a wide range of topics including market transparency, high-frequency trading, and systemic risks. I also conduct research in Asset Pricing and Behavioural Economics. My current research examines the contribution of high-frequency trading on systemic risk, by focusing on the underlying causes of flash crashes and the regulatory consequences of these events. More broadly, I look on how financial economics interact with physics and technologies.