concept

Liquidity raid

Price reaches past a visible high or low to trigger resting stop orders, then reverses. The core ICT microstructure primitive. Every sweep-based model on this site is a liquidity-raid trade at some level.

updated 2026-04-23
ictliquiditysweepstops

The idea

Resting stop orders cluster just past obvious reference levels:

When price pushes through one of those levels, those stops trigger. The stops are market orders — sell-stops above a low become market sells when the low breaks; buy-stops above a high become market buys. Both deliver instant fills against whoever is positioned on the other side.

A liquidity raid is the move that goes through the level specifically to trigger that order flow, then reverses. It’s not a failed breakout. It’s a successful inventory collection, after which price rotates back toward the opposite pool.

Why raids happen

To fill a large resting order you need counter-flow. If you want to buy 1,000 contracts and the visible book has 100, you can either:

  1. Walk the price up, buying what’s offered, and move the market against yourself — paying slippage every step.
  2. Wait for a move to a level where sell-stops are clustered, let those get triggered, and buy from the stops at the price that triggered them.

Option 2 is what the raid pattern is modeling. The mechanic explains why raids tend to cluster at the “obvious” highs and lows — that’s exactly where retail stops sit.

The signature

A raid looks different from a breakout:

RaidBreakout
Close relative to levelBack throughPast
Wick vs bodyBig wick past, small bodyBody past
Volume on the breachSpikeElevated but continuous
Next few candlesRotate backContinue
Prior-bar contextTight range into the levelMomentum into the level

On a chart, the classic signature is a long wick that pokes past a high or low and then a strong opposite-direction candle. On a 1-minute chart during a London sweep, it often resolves in 3–5 bars.

Where raids show up on this site

Basically everywhere:

The recurring pattern: different times of day have different raid resolution behavior. Some raids reverse (classic AMD), some continue (PRELON SWEEP). The model pages quantify which is which.

The two critical mistakes

Mistake 1: “Every sweep is a reversal.” Not true. At least on NQ, continuation after a sweep is roughly as common as reversion — the split depends on session and range context. Our AMD Early Pull walk-away covers a case where the reversal thesis was catastrophically wrong (13σ against) and the continuation side was the actual edge.

Mistake 2: “The wick size doesn’t matter.” It does. Depth past the level correlates with conviction — too shallow is noise, too deep can overshoot the reversion. Our CRT model applies an explicit depth adjustment.