Understand iceberg and absorption candidates
What the measurements show, and what sampled depth cannot establish.
Iceberg candidate calculation
For each completed window, Order Flow groups known-direction executions at the same price. It compares adjacent depth snapshots at that price. A refill sample requires executed quantity between those snapshots and at least 80% of the earlier displayed quantity remaining. The event must pass the refill count, execution volume and executed-to-peak-visible thresholds. The marker appears at the last contributing execution.
Why the label says candidate
Different participants can submit orders at the same price. Sampling can hide intermediate cancellations, fills and replenishments. These observations cannot prove a native or synthetic iceberg, determine its remaining reserve, or identify the owner. Order Flow does not expose a native/synthetic selector because the current adapter does not supply the order identity needed to support that claim.
Absorption is separate
Absorption candidates require enough same-direction aggressive volume with price movement inside the configured tick span. A buy-aggressor cluster is measured against the ask side; a sell-aggressor cluster against the bid side. This is an observed volume-and-price pattern, not proof that price will reverse.
Missing or interrupted data
Unknown aggression is excluded from directional detection. Duplicate execution sequences are counted once. Sequence gaps, out-of-order trade timestamps and depth gaps break the analysis history. The engine waits for a fully observed bucket after the break rather than joining observations across missing data.
Still need a hand?
Get in touch with the EdgeX team.