Hawkes Processes
Also: self-exciting process · Hawkes model
A Hawkes process is a self-exciting point process: each event temporarily raises the probability of the next. It models the clustering and bursts seen in order arrivals and trades.
Trades and order arrivals don’t tick like a metronome — they cluster. A burst of activity begets more activity, then it decays. A Hawkes process captures exactly this: its intensity jumps after each event and decays back toward a baseline, so events are self-exciting.
That property makes it a natural model for market microstructure, where one aggressive order triggers cancels, reactions, and follow-on trades. Hawkes models can also be mutually exciting — buy activity provoking sell activity across venues — which is why they show up in contagion and cross-market studies.
For an agent, the useful output is a live estimate of how hot the tape is right now and how fast it’s likely to cool. Rising intensity means the next few seconds carry elevated fill and slippage risk; a strategy can widen quotes, delay a passive entry, or switch to taking. It’s a more principled “the market just woke up” signal than staring at a volume bar.
- stochastic
- order flow
- clustering
Research source: rSwarm research library →