Statistical Methods in Finance 2025

Hawkes processes in high-frequency markets

By: Aditya Nittur Anantha
Algoquant Ltd.

Hawkes processes capture how order-flow events trigger further activity across types and venues, which makes them a natural language for microsecond market dynamics. With exponential kernels they admit a finite-dimensional Markov state, so intensities can be updated and simulated recursively with tight latency budgets. This talk outlines design patterns for applying Hawkes models to high-frequency trading and real-time execution for market-making: constructing streaming intensity forecasts, coupling event-history signals with limit-order-book state for execution choices, and using fitted kernels as diagnostics to down-weight transient bursts consistent with manipulation-like activity. I will emphasize parsimonious choices that matter in production: including short rolling windows, treatment of intraday seasonality, walk-forward estimation, and compute constraints. These ideas build on the literature that documents Hawkes excitation in finance and specific examples to applications in high frequency trading.