Liquidity Sweeps & Grabs

A liquidity sweep is a price move that pushes just beyond an obvious level where resting orders are likely to sit — yesterday's high, a pair of equal lows, the edge of the Asian range — and then fails to hold, returning inside the prior range. Traders who use this concept read the push as orders being filled rather than as the start of a new trend: the level is not broken, it is consumed. What follows, in most implementations, is a reversal trade taken from the swept level, conditioned on some form of confirmation.

The vocabulary is inconsistent across schools. *Sweep*, *grab*, *raid*, *purge*, *stop hunt*, and *inducement* all describe roughly the same event with different emphasis — inducement usually stressing that the level was attractive enough to pull traders in before the reversal. The mechanical claim underneath is the same in every case, and it is testable independently of the narrative used to justify it.

## How the mechanic is defined

Almost every version of this concept decomposes into three steps: locate a liquidity pool, detect the sweep, and require confirmation before entering.

A liquidity pool is any level where stop orders and pending orders are assumed to cluster. Common definitions are prior-day high and low (PDH/PDL), prior-week extremes, session highs and lows, equal highs and equal lows (EQH/EQL), swing points, trendline touches, and round numbers. When an indicator formalises this, it needs a tolerance — how close two highs must be to count as "equal" — and a lookback window.

Sweep detection is the second step: price trades beyond the level and comes back. The precise rule matters enormously. A wick beyond the level with a close back inside is a different event from a close beyond the level followed by a reversal several bars later, and the choice between them changes the trade population completely.

Confirmation is where implementations diverge most. Structure-based versions wait for a change of character (CHoCH), a market structure shift, a displacement candle, or a change in the state of delivery (CISD). Momentum-based versions wait for an oscillator shift or a moving-average condition. Correlation-based versions look for divergence between related instruments (SMT) at the moment of the sweep. Entry is then either immediate or on a pullback into a point of interest inside the reversal leg — an order block, a fair value gap, or a Fibonacci retracement zone. The stop typically sits beyond the sweep extreme; targets are usually the opposite liquidity pool or a fixed multiple of risk.

## Main variants

The catalog groups broadly into: **time-anchored** versions built on a session range (Asian/Tokyo, London open, New York open) or on daily bias; **level-type** versions specialised to EQH/EQL, swing points, or PDH/PDL; **confirmation-family** versions that differ only in what validates the reversal (structure, momentum, volume-derived measures such as liquidity-weighted averages or sentiment profiles); **correlation** versions that require SMT divergence between two instruments; and **timeframe** versions ranging from 1–5 minute scalping to H4/daily swing trading. Cutting across all of these is the split between indicators that mark pools and sweeps on the chart and full rule sets that specify entry, stop, and target.

## What actually differentiates implementations

Two systems described with identical language can produce almost disjoint trade lists. The variables that decide this are: the equality tolerance and lookback used to define pools; whether a sweep requires a wick or a close beyond the level; the maximum number of bars allowed for the return inside; how displacement or a structure break is quantified; whether a higher-timeframe bias filter restricts trade direction; entry at confirmation versus entry on retracement; stop distance beyond the extreme; and whether targets are opposing liquidity or a fixed R multiple. When comparing versions, compare these parameters, not the narratives.

## Common mistakes

The most frequent error is treating every wick beyond a high as a sweep. Without an explicit failure-and-return condition, the concept degenerates into fading noise, and in trending conditions many of those levels are genuine breakouts. The second is repainting: a large share of liquidity and structure indicators confirm pools, sweeps, or structure breaks only after subsequent bars close, so reading the finished label in a backtest introduces look-ahead bias. Others include placing stops exactly at the sweep extreme rather than beyond it, ignoring the higher-timeframe direction entirely, assuming fills at the wick extreme in a moment of elevated volatility, and building session logic that quietly breaks under broker timezone or DST changes.

## How to evaluate and backtest a version

Write the rules as unambiguous, timestamped conditions before coding anything; if a rule cannot be stated without the word "clearly", it is not yet testable. Because the sweep and its reversal frequently occur inside a single candle, verify intrabar order using lower-timeframe or tick data — a 15-minute bar cannot tell you whether the wick preceded the close. Test for repainting by comparing bar-by-bar signals against the historical plot.

These setups are event-scarce: prior-day raids may yield only a handful of instances per instrument per month, so extend the test across instruments and years rather than tightening filters to improve a small sample. Vary tolerance and window parameters and look for stable plateaus instead of isolated peaks. Segment results by session, instrument, and volatility regime. Finally, compare against a naive baseline — a plain breakout of the same levels, or random entries with the same stop and target — so the sweep logic has to justify its own complexity, and apply realistic spread and slippage assumptions throughout.

Strategies in this concept (92)

Frequently asked questions

What is the difference between a liquidity sweep and a normal breakout?

Structurally, the price action is the same up to the moment the level is crossed. The difference is what happens next: a breakout continues in the direction of the move, while a sweep fails and returns inside the prior range. Because the distinction is only visible after the fact, every tradable version needs an explicit rule for how far back inside price must come, and within how many bars, before the setup is considered valid.

Are sweep, grab, raid, purge, stop hunt and inducement different concepts?

In practice they describe the same event with different emphasis and come from different teaching traditions. Inducement usually stresses that the level was visible enough to attract entries before the reversal; raid and purge are common in ICT-derived material; grab and stop hunt are older price-action terms. When comparing implementations, ignore the label and compare the detection rules and confirmation requirements.

Does this concept require order flow, volume, or footprint data?

No. Most implementations work entirely from price: levels, wicks, closes, and subsequent structure. Some versions add volume-derived tools such as liquidity-weighted averages or sentiment profiles as confirmation, but the core mechanic is defined on OHLC data alone. That is also why the concept is available on standard charting platforms without a data subscription beyond price.

Which markets and timeframes does it apply to?

The logic is instrument-agnostic and appears across indices, forex, gold, and crypto, on timeframes from one minute to daily. What changes with the instrument is which liquidity pools matter: session ranges are meaningful in instruments with clear regional sessions, while prior-day and prior-week extremes generalise more widely. Session-anchored versions need their time definitions re-checked whenever the instrument or broker timezone changes.

Can a liquidity sweep strategy be fully automated?

Yes, provided every discretionary term is replaced with a numeric condition — how close two highs must be to count as equal, what magnitude qualifies as displacement, how many bars the return window allows. The main obstacle is not the logic but the data: the sweep and the reversal often occur within one candle, so an accurate test needs lower-timeframe or tick data to establish the order of events.

Does using this concept require believing that markets are deliberately manipulated?

No. The narrative about who places or takes the orders is separate from the pattern. What can be tested is whether a failed excursion beyond a clustered-order level, followed by a defined confirmation, produces a measurable edge relative to a baseline. Evaluate the rules on that basis; the explanatory story neither strengthens nor weakens the statistical result.

All strategy concepts · Explore all strategies