A liquidity sweep trade fades a move that pokes beyond a high or low and closes back inside. On NQ futures it only showed an edge at the prior day's regular-session high and low, and even there it was modest and depended on the period. Fading sweeps of ordinary intraday swing highs and lows, or of equal-high "liquidity pools", lost money consistently.
What we tested
Three versions of the same idea, all on one-minute NQ with realistic costs, entries on the close that reclaims the level, and a stop beyond the sweep:
- Sweeps of any recent intraday swing high or low (the external range liquidity, ERL), targeting the nearest fair-value gap inside the range (the internal range liquidity, IRL).
- Sweeps of equal highs or lows: two or three swing points at nearly the same price, often called buy-side or sell-side liquidity.
- Sweeps of the prior day's regular-session (09:30–16:00) high and low.
Intraday swings: a clear loss
Fading every sweep of an intraday swing point took 11,294 trades over three years of one-minute data and lost 243R (R = the amount risked per trade). On 15-minute bars it still lost 62.8R. Adding the fair-value-gap target made it rarer, not better: only 20 to 48 trades in three years, with no reliable edge.
Equal highs and lows: no direction
Fading a sweep of a double top or bottom lost 106R. We then tried the exact opposite, buying into the sweep as a breakout, and that lost too. When both a rule and its mirror image lose, the signal carries no direction, only trading costs.
The one positive cell was going with the break of three or more stacked equal highs: +19R over three years, but on only 96 trades. That is an ordinary momentum breakout, the same family that already works better elsewhere.
Prior-day high and low: the only place it worked
Sweep-and-reclaim at the prior day's regular-session high or low was positive in 2023–2026, about +0.07R per trade, positive in each of those years. Two caveats: with realistic fills at the next minute's open the result dropped by roughly 30% (the stops are tight, so a tick matters), and on 2014–2020 data, which the rule had never seen, it lost money.
The likely reason it worked at all: the prior day's extremes are levels the whole market watches and defends. An arbitrary swing high from 40 minutes ago is not.
What to take from it
- Use sweeps of major levels (prior-day high/low) as context, not every swing on the chart.
- Treat "liquidity" labels on intraday swings as a map of where stops may sit, not as trade signals.
- If you trade sweeps, the read that makes them work, how price behaves after the poke, is discretionary. The mechanical version didn't hold up.
Simulated results on historical data, after commissions. They don't guarantee future results; see the risk disclosure and CFTC Rule 4.41.