An introduction to high-frequency and low-latency trading

Abstract

High-frequency trading (HFT) is fundamentally a problem of modelling market microstructure at extreme time scales. It sits at the intersection of machine learning, time-series modelling and ultra-low-latency systems engineering.

We frame HFT through its quantitative core: how order book dynamics, queue positions, trade arrival processes, and adverse selection are formalised into tractable models, and how these models drive real-time decisions.
From market data to systems engineering and live trading, we examine each component of an HFT system and the quantitative problems it raises — such as short-horizon price prediction from order book dynamics, or latency and statistical arbitrage.

Using real market data, we illustrate the peculiar properties of HFT data — nanosecond timestamps, order-by-order book updates, and exchange messages — and the challenges they pose for classical estimation and machine learning.

We conclude by identifying open problems for the quantitative finance and machine learning communities.

Bio

David has worked as a quantitative researcher at several hedge funds and market-making firms in both Sydney and London. Most recently he was a quantitative researcher at WorldQuant / Millennium Capital, where he designed and live traded signals for low-latency trading in US equity markets. His research interests centre on the application of probabilistic machine learning to the forecasting of market microstructure and the modelling of order books.

Before that he was Head of Quantitative Research at Life Trading in Sydney, and earlier he spent a number of years in London in senior quantitative research roles, leading teams that designed trading systems, and low latency strategies across equities, fixed income and futures.

He holds a PhD in Machine Learning from INRIA, the French national research institute in computer science, and completed postdoctoral research at the University of California, Berkeley, focusing on probabilistic graphical models and variational inference.