Heatmap intelligence panel: liquidity lens, evidence and alerts

The Intelligence button of the heatmap opens the Liquidity lens, a side panel that records the life of each displayed book level (quantity, peak, executions, net changes) and lists candidate events: walls, refills, absorptions, sweeps and MBO refreshes.

Senzoukria · Documentation · Updated September 2026

Where to find it

Where
Heatmap → ◈ Intelligence button (top toolbar) → panel titled Liquidity lens
Refresh
Snapshot supplied at 4 Hz; the panel never subscribes to the tick stream
Defaults
Record evidence on; Minimum detected volume 0 (automatic); Wall quantity threshold 0 (automatic)
Memory
Up to 256 tracked levels, 128 events, 48 quantity samples per level

What it does

The panel is fed by HeatmapIntelligenceModel, a bounded observer of the depth snapshots and trades the heatmap already receives. For every displayed price level it keeps the first and last time it was seen, the displayed quantity, its peak, the volume traded at that level, and the summed net increases and decreases. MBP depth gives aggregated quantities only, so the model never claims to see individual orders, cancellations or intent; the panel repeats this in a note.

Two tabs: Inspector shows the selected level (side, price, sparkline of the displayed quantity, Displayed, Peak, Observed life, Trades at level, Net increases, Net decreases) and a list of Recorded levels, the 12 active levels with the highest peak. Timeline lists the last 80 recorded events, newest first, with a filter by kind and a Selected level checkbox.

Five event kinds exist. Wall observed and MBO refresh observed are marked Observed event. Refill candidate, Absorption candidate and Sweep candidate are marked MBP inference: an absorption candidate means aggressive volume at least equal to the minimum detected volume hit a level between two complete books while the level stayed present; a refill candidate is the same with a displayed size that did not fall; a sweep candidate is a run of at least 3 consecutive levels on one side within 250 ms. Each kind has a 3-second cooldown per level, and nothing is emitted during the first 2 seconds of observation.

Settings

Detection settings (footer of the panel) and panel controls
SettingDefaultWhat it changes
Record evidenceOnEnables the model. Off pauses recording; the inspector then says recording is paused
Minimum detected volume0 (automatic)Aggressive volume needed at a level to emit an absorption or refill candidate, and minimum size of a sweep. 0 = 2 × the median of the observed book sizes, recalibrated at most once per second
Wall quantity threshold0 (automatic)Displayed size from which a level is a wall. 0 = max(3 × median book size, 90th percentile of the book)
In-panel alertsOffShows the newest event in the footer; new events only, at most one every 15 seconds, while the panel is open
Pin levelNoneUp to 6 pinned levels, kept visible on the heatmap even when the panel is closed
Filter event type / Selected levelAll events / offRestrict the Timeline to one kind or to events touching the selected level

How to use it

  • Click a level on the heatmap, or one of the Recorded levels, to load it in the Inspector. Pin it to compare it with others later.
  • Read the coverage line first: Building context (warm-up), Observed window or Incomplete window. An incomplete window means a gap, a partial book or a snapshot older than 5 seconds; levels that ended during a gap are flagged Partial evidence.
  • Save analysis to journal writes the recorded evidence of the selected level to the daily note (capture date in UTC). Explain this zone sends only that evidence to the AI model chosen in the compact model picker, after your click.
  • Thresholds are typed in the native quantity of the book (contracts, or base units on crypto) and stored per instrument in the browser storage of the app.

Limits and pitfalls

  • Only the retained live window is described. Historical heatmap pixels are not order evidence, and the panel says so.
  • Tracking limit reached appears when more levels exist than the model keeps (256, half per side): the deepest levels are not tracked.
  • Out-of-order events ignored counts snapshots or trades whose timestamps went backwards; they are dropped rather than reinterpreted.
  • MBO refresh observations only exist on feeds that report order-by-order refreshes; they do not confirm a native iceberg.
  • Every candidate is a heuristic on aggregated data; it is not a bar-final absorption and not proof of a passive participant.

This page in other languages

Frequently asked questions

Why is the Timeline empty while the heatmap is moving?
Events need a warmed model (2 seconds of continuous books), a complete two-sided book and volumes above the resolved thresholds. With automatic thresholds a uniform book produces few walls by design. Check the coverage line and the Detection settings.
Are alerts sent outside the app?
No. In-panel alerts only appear in the footer of the panel while it is open, and never for events that existed before you enabled them.
Does the Explain button send my chart to the AI?
It sends the recorded evidence of the selected zone (prices, quantities, timestamps, coverage flags) to the model you selected, and only when you click. Nothing is sent while the panel just sits open.

Keep reading

Strategy researchTest a trading strategy, from observation to challenge rulesDefine an observation, question it with the copilot, write a rule and explore costs, drawdown and fictional challenge limits in an interactive example.Quant · Start hereQuant trading for beginners: from idea to backtestDefine a testable rule, choose suitable data and evaluate later periods after costs.Order flow · HubOrder flow trading: the complete guideThe hub. What order flow is, what it shows — aggression, delta, imbalance, absorption, liquidity — and the tools that read it, each linked to a full guide.