Liquidity-sweep reversal

Friday 16 October 2026, 12:02 ET (23:02 ICT). NQ slides into Thursday’s low, 29,622.50, and trades through it. Twenty-two seconds later it prints 29,609.00, 13.5 points under the level. Fifty-seven seconds after the first trade below, it is back above, and the 5-minute bar closes at 29,644.00. A “liquidity sweep” before a reversal, or the first leg of a breakdown? This lesson writes the setup as a rule you can score. One small piece of it is measured; the reversal itself is not.

Key terms

Two events share the word “sweep”. Only one of them has been measured.

Order-level sweep (sweep cấp lệnh): one aggressive order that trades through three or more prices of the book in a single match event Definition. It lasts milliseconds and shows in the tape or as a bubble (M3.11).

Liquidity sweep (quét thanh khoản), the chart setup: price trades briefly beyond an obvious prior extreme (yesterday’s high or low, the overnight high or low) and comes back inside Definition. It lasts seconds to minutes and is built from many trades, order-level sweeps among them.

Reclaim (lấy lại mức giá): a 5-minute close back inside the level within 15 minutes of the first trade beyond it House rule. The window is a convention that makes the check scorable, not a measured cut-off.

Chart first: a sweep and reclaim on a real day

How it works

The story behind the setup

Myth / hypothesis: “Stops rest under every obvious low. A large buyer pushes price through to trigger them, buys the forced selling, and the breakout sellers who joined at the low are trapped; their exits fuel the reversal.” Lore On NQ nobody can see where stops rest or why a large trader acted, so the story cannot be checked from the data, and it explains every sweep after the fact, including the ones that keep going.

There is real evidence that stop orders cluster, in another market, and it points the other way. In one currency dealer’s order book, stop-loss orders sat just beyond round numbers, and once price crossed them the triggered stops tended to speed the move up rather than reverse it MeasuredOther market (Osler 2003, further reading). If stops do sit under yesterday’s NQ low, hitting them first adds selling. That is why this setup needs a reclaim, and why it can fail.

What is measured: order-level sweeps give back a little

On NQ, order-level sweeps through three or more prices continued over the next minute slightly less often than single-price bursts of the same side, size and time of day PlausibleNQ. The wider the sweep, the bigger the give-back, and after fifteen minutes it is gone. That is the pattern of temporary price impact: a large order pushes through a thin book, overshoots, and a little of the push comes back as the book refills.

A spike through a low is built from such sweeps, so some give-back is its normal aftermath. That describes part of a wick, not a 50-point reversal: the average give-back is about 2 ticks over a minute, about a round trip in costs.

What is not measured: the setup’s reversal

Nobody has tested “sweep and reclaim of a prior high or low, then a reversal to a target” against a fair baseline on NQ: not yet measured. The closest tested version, a failed break of the prior day’s high or low travelling to the other side of the range, showed no edge on NQ in a public practitioner study Lore (the evidence box in M5.01). And our lab found that prior-day highs and lows held after a touch no more often than random prices at the same distance Measured: no edgeNasdaq index. The level does not make the trade.

The honest status: a rule-based trade with a close, objective stop and an unproven edge. Trade it small while you score it; do not call it high odds.

The rule

Write all five parts before the open.

• Context: a prior extreme you mapped before the open: yesterday’s regular-session high or low, or the overnight high or low (M4.07). It tells you where to watch, not where price turns House rule.
• Trigger: a trade beyond the level, then the reclaim check: a 5-minute close back inside within 15 minutes of the first trade beyond House rule. No reclaim, no trade. If price stays beyond the level and passes the acceptance checks of M4.01, the idea to watch is the breakout instead (M5.07).
• Entry: after the reclaim close, with a market or stop order. Not a resting limit inside the spike: a limit that fills during a sweep fills because the sweep went through it, and starts the trade at a loss (markout, M3.11).
• Invalidation: one tick beyond the sweep extreme. Add an acceptance exit: two 5-minute closes back beyond the level and you are flat, even if the stop has not traded House rule. Before the click, check the stop distance: if it is wider than your sizing rule allows, there is no trade.
• Target and time stop: a fixed R multiple or the nearest reference inside (VWAP, the initial-balance midpoint), written before entry; flat after 60 minutes if neither the stop nor the target has traded House rule. Compute the break-even win rate after costs (M6.08); if you cannot argue the rule beats it, the math is unknown, not good.

Why learn it anyway: “swept Thursday’s low by 13.5 points, back above in 57 seconds” is a precise account of the auction, the sweep extreme gives an objective stop, and written checks can be scored in replay, the only route from “not yet measured” to an answer.

What the data says

Worked example: Thursday’s low, 16 October 2026 (long side)

Real NQ session from our lab’s tick data; a reading, not a signal. MNQ: $2 a point; costs about $2.50 per round trip (M0.01).

• Map, written before the open. Thursday’s regular-session low: 29,622.50. Scenario: if a trade below it is reclaimed within 15 minutes, buy the reclaim; if price is accepted below, no long.
• 12:02:32 ET (23:02 ICT), first trade below. 29,622.25. No verdict yet.
• 12:02:54, sweep low 29,609.00. 13.5 points under the level. About 3,600 contracts trade below the level in under a minute.
• 12:05, verdict. The 5-minute bar closes at 29,644.00, back above, about 2.5 minutes after the first trade below. Reclaim check passed.
• Plan. Long 1 MNQ at 29,644.00 (first trade at 12:05:00). Stop 29,608.75, one tick under the sweep low: 35.25 points, $70.50. Target 29,697.00, 53 points, about 1.5R: $106. Time stop 13:05.
• Break-even. After costs, a win nets $103.50 and a loss costs $73.00, so the rule needs to win 73.00 ÷ 176.50 ≈ 41% of the time. How often this rule wins on NQ is not yet measured.
• Outcome. The target fills at 12:15:06: +$103.50 after costs.

Now change one line of the plan. Suppose the target had been “the origin of the drop”, the 11:45 high at 29,753.75. The best price after entry was 29,708.25, so that target never traded. Price drifted back, the 12:45 bar took out the sweep low, and the stop filled at 12:48:50: −$73.00. By 13:10 NQ traded at 29,487.75, 135 points under Thursday’s low. Same entry, same day: the exit rule written in advance decided the result. One day is evidence for neither target.

The mirror and the failure

The short side (mirror). Above a mapped prior high the rule reads upside down: a trade above, a 5-minute close back below within 15 minutes, a short after the close, the stop one tick above the sweep high. Flip the figure to see it. The measured result covers buy and sell sweeps alike.

The failure: Monday 12 October 2026. Friday’s regular-session low was 30,991.00, and NQ opened at 31,045.50.

• 09:31:18 ET (20:31 ICT). First trade below the level. Price reaches 30,936.00 at 09:34:44, 55 points under it.
• 09:35. Back above at 09:35:46; the 5-minute bar closes at 31,030.25. Reclaim check passed.
• Plan as written. Long 1 MNQ at 31,031.50 at 09:40:00, stop 30,935.75: 95.75 points, $191.50. Check the stop distance first: an opening sweep can be deep, and if 96 points is wider than your sizing rule allows, there is no trade.
• What happened. The best price after entry was 31,034.00. The 09:40 bar closed back below the level at 30,975.00, and the 09:45 bar closed below again at 30,969.25. The acceptance exit takes you out near 30,969: about −$127 after costs. Without it, the stop fills at 10:05:19 for −$194.

The reclaim was real evidence at the time and the trade followed the rule. The auction changed its answer; the exits kept the loss defined.

Prop-firm note. In an evaluation, place the stop as a real order with the entry. Evaluation accounts usually fill in simulation, and simulated fills are not live fills; tuning tight brackets or breakeven moves to profit from simulated fills is treated as prohibited conduct by firms such as Topstep (PROP1.04, in the Prop Firm course). Check your firm’s help center before you trade a firm account.

Common mistake: Buying the sweep itself, at the wick, while price is still falling. The sweep alone proves nothing: triggered stops can speed a move up, and on 16 October the low that held at 12:03 broke 45 minutes later. The reclaim is your evidence; predicting it instead of waiting for it is the classic early-entry loss.

Common mistake: Treating a small measured effect as a big edge. “Sweeps reverse” is true only for order-level sweeps, by about 2 ticks over a minute, smaller than a round trip, with the forward test pending. It does not justify a 50-point target or a bigger size on the chart setup, whose reversal has not been shown to have an edge.

Common mistake: Resting a buy limit a few ticks under the level to catch the sweep. If it fills, it fills because sellers traded through it, and the trade starts at a loss. Enter after the reclaim close with a defined stop instead.

Drill: score twenty sweeps of a prior extreme

1. Write your reclaim check, entry, stop (one tick beyond the sweep extreme), acceptance exit, target and 60-minute time stop before you open a chart.
2. In replay, find 20 sessions where NQ trades beyond the prior regular-session high or low after the open. For each, record the depth beyond the level in points, the minutes beyond, and whether the reclaim check passed.
3. For every reclaim, record which came first: target, stop, acceptance exit or time stop, and the result in R after costs.
4. Count how often the reclaim check failed (price stayed beyond), and compare your win rate with the break-even rate your stop and target imply.

What we don’t know yet

• Whether sweep-and-reclaim at prior highs and lows beats a placebo (the same rule at random prices) on NQ: not yet measured.
• Whether the slight order-level give-back survives the pre-registered forward test on new sessions: pending in our lab.
• Whether a fade of order-level sweeps makes money after costs: not yet measured.
• Where stops actually rest on NQ: not observable in public data, so not measured.

Recap

• “Sweep” names two things: one order through several prices, and price poking beyond a prior extreme and coming back. Only the first is measured.
• Order-level sweeps on NQ give back slightly over the next minute (Plausible, forward test pending), by about a round trip in costs.
• The sweep-and-reclaim setup is a rule with a close stop whose edge is unproven; prior-day highs and lows held no better than random prices.
• Trade only after the written reclaim check, invalidate one tick beyond the sweep extreme or on acceptance back beyond the level, and fix the target and time stop before entry.

Further reading

• Osler, C. L. (2003), “Currency orders and exchange rate dynamics: an explanation for the predictive success of technical analysis”, Journal of Finance 58(5), 1791–1819 (RePEc). Stop orders cluster beyond round numbers; crossing them speeds moves up.
• Order Flow, Measured: All claims. A public, pre-registered practitioner study on NQ, 2018–2026. Read the sweep row (OB-03) and the failed-break row (AP-04).
• Our lab: NQ order-level sweeps of three or more prices continued about 1.4 percentage points less often than matched single-price bursts over the next minute, a give-back that grows with sweep width and is gone after 15 minutes; forward test pending.

Next: M5.06 moves to the open: how the first minutes of a session start a reversal, and what the measured open-location base rates add to the opening-type story.

Educational content only. Not financial advice. Futures trading involves substantial risk of loss.