Tick arrival comes in bursts, not in a smooth stream
Posted: Sat Oct 31, 2026 6:39 pm
Quote updates arrive in clusters. A burst of ten ticks in a second is often followed by a quiet stretch, then another burst. The pattern resembles self-excitation: activity begets activity, as one update triggers reactions from algorithms and other providers.
The practical consequence is that a rolling average of tick rate hides the structure. A strategy tuned on average activity may be overwhelmed during the bursts and idle between them.
Measuring burstiness is straightforward. Compute the coefficient of variation of tick counts per second. A value above one suggests clustering. Compare across hours and pairs.
Bursts tend to coincide with price movement, so they carry information. A strategy that reacts to the start of a burst, rather than waiting for a bar to close, may capture the early part of a move.
The risk is that bursts also coincide with the worst fills, since spreads and latency are stressed.
Testing both the opportunity and the cost, using tick data, separates real edge from the illusion of reacting to everything.
How clustered is the tick flow on the pair being traded?
The practical consequence is that a rolling average of tick rate hides the structure. A strategy tuned on average activity may be overwhelmed during the bursts and idle between them.
Measuring burstiness is straightforward. Compute the coefficient of variation of tick counts per second. A value above one suggests clustering. Compare across hours and pairs.
Bursts tend to coincide with price movement, so they carry information. A strategy that reacts to the start of a burst, rather than waiting for a bar to close, may capture the early part of a move.
The risk is that bursts also coincide with the worst fills, since spreads and latency are stressed.
Testing both the opportunity and the cost, using tick data, separates real edge from the illusion of reacting to everything.
How clustered is the tick flow on the pair being traded?