Randomising Execution To Avoid Detection

An execution algorithm that always slices a large order into equal pieces sent at exact, evenly spaced intervals is easy for other market participants to detect. Predatory strategies specifically look for this kind of regular footprint — a fixed-size child order arriving every few seconds on the same venue is a strong tell that a much larger parent order is being worked, and once detected, other traders can front-run the remainder, pushing the price against the slicing algorithm before it finishes.

Execution randomisation counters this by varying the pieces the algorithm is not required to keep fixed: child order sizes are jittered around a target rather than kept identical, the time between orders is drawn from a random distribution rather than a fixed clock, and orders are spread unevenly across multiple venues rather than always routed the same way. None of this changes the total amount being bought or sold, or meaningfully changes the strategy's intended participation rate — it only removes the predictable pattern that a detection algorithm would otherwise learn to recognize.

The randomisation has to stay bounded, since scattering orders too widely in size or timing undermines the algorithm's own ability to track its intended schedule, such as staying close to a VWAP benchmark.