---
title: Absorption in Order Flow
slug: absorption-in-order-flow
cluster: learn
description: What absorption means in order flow, how it looks different in the order book and on the tape, the two ways it is actually detected, and how to tell a level that is holding from one that is about to break.
tldr: Absorption is aggressive volume arriving at a price and failing to move it, because resting limit orders on the other side are consuming the flow as fast as it comes. On the tape it shows as a large signed delta with almost no net displacement; in the book it shows as size that keeps refilling at the same price while it is being executed.
published: 2026-07-31
updated: 2026-07-31
author: The Confluence Show Research
schema: Article
keywords: [absorption order flow, absorption trading, iceberg orders, trapped passive orders, absorption exhaustion]
prompts: [What is absorption in order flow?, How do you spot absorption on a chart?, What is the difference between absorption and exhaustion?]
entities: [absorption, order flow, iceberg order, order book, delta, liquidity]
related: [order-flow-trading-explained, footprint-charts-explained, cvd-explained, liquidity-sweeps-explained]
faq:
  - q: What is the difference between absorption and exhaustion?
    a: Absorption describes the passive side winning — aggression arrives and is consumed without moving price. Exhaustion describes the aggressive side running out — the pushing stops because there is no one left to push. They can look similar on a delta chart and are distinguished by what happens on the next test of the level.
  - q: Can you see absorption without order book data?
    a: Partially. The tape alone shows a large signed delta with little net displacement which is the effort-without-result signature. What the tape cannot show is replenishment — whether the resting size is being refilled or simply eaten — and that requires book snapshots.
  - q: Is absorption a reversal signal?
    a: No. It is an observation that one side is defending a price. A defended level can hold and turn the market or be worn down and broken; the observation itself does not distinguish the two which is why the follow-up tests matter more than the original event.
  - q: What is a trapped passive order?
    a: A resting participant who was defending a level that then broke against them. Their position is now underwater at a known price and their eventual exit pushes in the direction of the break which is why the level often behaves differently on the retest.
  - q: Why does absorption sometimes appear on almost every bar?
    a: Because the detection threshold is too loose. A sensitive rule that only asks for concentrated delta and a rejection wick fires constantly including on bars with below-median volume. Requiring genuinely extreme volume relative to a strictly prior window cuts the count down to something meaningful.
sources:
  - label: Binance spot WebSocket market streams (trade and depth feeds)
    url: https://developers.binance.com/docs/binance-spot-api-docs/web-socket-streams
draft: false
---

## What is absorption in order flow?

Absorption is aggressive volume arriving at a price and failing to move it, because resting limit orders on the opposite side are consuming the flow as fast as it comes. It is the clearest case of the effort-versus-result idea: a lot of effort was spent and no result was produced, which means someone on the passive side was willing and able to take the other end of all of it.

The word describes a *relationship*, not a pattern. Every trade has an aggressor and a passive counterparty; absorption is simply the case where the passive counterparty is large enough, and persistent enough, that the aggressor's size disappears into it. In a normal push, aggression consumes the resting size at a price, that price empties, and the market moves to the next one. In absorption, the size at the price does not empty.

This matters because it inverts the naive reading of volume. A surge of buying at a high looks bullish if you only see unsigned volume. If price does not advance while that buying is happening, the correct reading is that a seller is being served at a good price and the buyers are the ones being accommodated. [Order flow](/learn/order-flow-trading-explained) exists largely to make this inversion visible.

Absorption is also the mechanism behind most [CVD](/learn/cvd-explained) divergences. When cumulative delta grinds one way and price refuses to follow, the arithmetic is telling you that something is absorbing. The divergence is the symptom; absorption is the thing.

## Absorption in the book versus absorption on the tape

The two data feeds show absorption differently, and neither one is complete on its own. The tape shows the aggression that arrived and what it achieved; the book shows the liquidity that met it and whether it was replaced. Read together, they answer a question neither can answer alone: was that size being eaten, or refilled?

| | On the tape (trades) | In the book (depth) |
| --- | --- | --- |
| What you observe | Large signed delta with little net displacement | Resting size at a price that does not deplete |
| Direct measurement | Aggressive volume by side and price | Displayed size and how it changes between snapshots |
| Key derived quantity | Delta versus bar range | Replenishment: size refilled versus size executed |
| Blind spot | Cannot tell refilled from never-consumed | Hidden and iceberg size never displays at all |
| Can it be spoofed | No — the trades happened | Yes — displayed size can be posted and pulled |

The practical consequence is a hierarchy of evidence. Executed volume is the hard fact; displayed liquidity is a claim that may or may not be honoured. A wall in the book that vanishes before price reaches it told you about someone's intentions, not about a defence. A wall that is repeatedly executed against and keeps reappearing is a defence, and the arithmetic of that — how much executed versus the largest size ever shown — is exactly how a hidden order gets inferred.

## How is absorption actually detected?

Two ways, and a serious implementation uses the better one whenever the data allows. The stronger method uses order flow directly: it asks whether an extreme signed delta failed to produce the price outcome it should have. The fallback method uses volume and displacement only, for when per-bar flow is unavailable.

**Method A — order flow.** Delta for the bar is compared against a threshold derived from the instrument's own recent history: a rolling percentile of absolute delta over the *strictly prior* window, never including the bar being judged. A ceiling reads as absorbed buying when delta is strongly positive and the bar still closes at or below its open. A floor reads as absorbed selling when delta is strongly negative and the bar closes higher or leaves a lower wick covering a large part of its range. This method is preferred because the aggressor side is measured from the maker flag rather than estimated.

**Method B — volume and displacement.** Where per-bar flow is missing, absorption reduces to the classic reading: very high volume, very little net movement. "Very high" needs two gates, not one — volume above a multiple of the prior window's *median*, which is robust to outliers, and above a z-score threshold against the prior window's dispersion, which is sensitive to how varied that window actually was. "Very little movement" means the close-to-close change is small relative to the prior window's median bar range.

Both gates share one non-negotiable property: the comparison window ends before the bar under evaluation. A threshold computed over a window that includes the bar it is judging will flag that bar for being unusual in a sample it defined. That is not detection, it is circular arithmetic, and it is the single most common flaw in absorption tooling.

## Icebergs, replenishment and trapped passive orders

Absorption raises an obvious question: where is all that resting size coming from? Often from an iceberg — a large order that displays only a small slice at a time and replaces the slice each time it is filled. Icebergs cannot be observed directly; they are inferred from the mismatch between what displayed and what executed.

The inference is arithmetic. Over a moving window of book snapshots at one price, you track cumulative executed volume, cumulative replenished size, and the largest size ever *shown*. When executed volume exceeds the largest displayed size by a large multiple, and the refill rate is high enough that the level was genuinely being restocked rather than eaten once, the level is flagged as a possible iceberg. The Confluence Engine gates that multiple on a percentile of the instrument's own recent history with an absolute floor underneath, so a thin market cannot manufacture icebergs out of small numbers.

The mirror case is a **trapped passive order**. A participant defends a level, the level breaks against them, and their resting position is now underwater at a price everyone can see. That price does not become irrelevant — it becomes the price they need. A trapped buyer sits above the market and their exit pushes down; a trapped seller sits below and their exit pushes up. The engine draws these as markers that only become confirmed once price retests the lost level without reclaiming it, because an unconfirmed trap is a hypothesis and a retested one is an observation.

Both readings depend on keeping book snapshots alongside the trade stream. Neither can be reconstructed from candles, and neither can be reconstructed from trades alone.

## When does absorption fail?

When the aggressive side brings more size than the passive side is willing to keep supplying. Absorption is a defence, and defences are finite. The useful question is therefore never "is there absorption here" but "is this defence holding, weakening, or done" — and that is answered by the *sequence of tests*, not by the original bar.

The Confluence Engine treats an absorption level as an object with a life, not an event. Each subsequent test of the level is sealed as its own observation carrying: which side is attacking, how deep the penetration went in ticks, how much aggressive delta the attacker spent on this attempt, and how much the resting side replenished during it. From those, each test resolves to one of three states:

| State | What the test showed | Reading |
| --- | --- | --- |
| Holding | Shallow penetration, replenishment keeping pace | The defence is still funded |
| Weakening | Deeper penetration or less replenishment than the prior test | Same defender, less capacity |
| Exhausted | Penetration through with no meaningful replenishment | The defence is gone |

Read in sequence, this is far more informative than any single flag. A level tested three times with progressively deeper penetration and progressively weaker refill is telling a coherent story about a defender running down their book. A level tested three times with identical shallow penetration is telling the opposite one.

## Absorption, exhaustion and liquidity sweeps

Absorption, exhaustion and liquidity sweeps get conflated constantly and they are mechanically distinct. Absorption is the *passive* side winning: aggression arrives and is consumed. Exhaustion is the *aggressive* side finishing: the pushing stops because there is nobody left to push, not because someone stopped it. A sweep is aggression that deliberately runs a level to trigger the orders resting beyond it.

| | Who is doing the work | Volume signature | What follows is diagnostic |
| --- | --- | --- | --- |
| Absorption | Passive side | High aggressive volume, little displacement | Whether the level holds on retest |
| Exhaustion | Nobody — flow dries up | Falling volume and falling delta | Whether the range simply stalls |
| Sweep | Aggressive side | Fast displacement through a level then reversion | Whether price accepts beyond the level |

The overlap is real: a sweep frequently ends *in* absorption, when the aggression that ran the level is met by size on the other side of it. That combination — a level run, liquidity taken, and the resulting flow absorbed — is one of the more legible sequences in order flow, and it is covered in [liquidity sweeps](/learn/liquidity-sweeps-explained). What distinguishes it from a plain break is entirely in the aftermath, which is why none of these readings can be evaluated on the bar that produces them.

## How absorption is tracked on The Confluence Show

The Confluence Show is a live, AI-run market analysis broadcast. Its analyst, NAIRO, reads raw trades, order books and positioning 24 hours a day through the Confluence Engine — more than forty analytical layers computed in-house — and absorption is one of the layers that most depends on the engine keeping raw data rather than summaries.

In practice that means three things. Absorption is gated on a single shared criterion, so the level family drawn on the chart and the one the analyst reasons over can never diverge and double-count. Every absorption level is followed test by test, so what gets said on air is "this defence is weakening on its third test", not "there was absorption here". And every reading carries the moment it became knowable, so a level is never drawn as though it had been visible before the data existed.

You can toggle the absorption and [footprint](/learn/footprint-charts-explained) layers yourself in the live room and inspect the bar under discussion while it is being discussed. [How it works](/how-it-works) covers the engine, [NAIRO](/nairo) covers the analyst, and the show streams free on a delay at [watch](/watch). It is educational market analysis, not financial advice, and it issues no signals.