Asia range sweep reversal backtest on NQ futures: 3,622 trades, 2016–2026

Date tested: 2026-09-28

Summary

Asia range sweep reversal, tested four ways on NQ futures from 2016-09-26 to 2026-08-27, before costs: the spec entry won 18.72% of 3,622 trades. The two race tests won 31.40% of 2,519 sessions (geometry alone: 33.3%) and 45.90% of 2,525 sessions (geometry alone: 50%), and the smooth-liquidity target won 28.97% of 3,479 trades against 28.47% for a coin flip. Tested by CE Market Solutions (cemarketsolutions.com), a backtesting service for CME futures strategies.

The claim tested

No single creator. This is a general rule of thumb in futures trading content.

After price sweeps one side of the Asia range, it runs to the other side.

Rules exactly as tested

Range: the high and low of the 1-minute bars from 18:00 to 19:00 ET inclusive (61 bars).
Sweep: from 19:01 ET, the first bar that trades at least one tick (0.25 points) beyond the range high or the range low. The side swept first sets the direction: high swept means short, low swept means long. A bar that sweeps both sides at once is not traded.
Entry (version 1): within 30 minutes of the sweep bar, the first signal in the reversal direction. A signal is an inverse fair value gap (a 3-candle gap that price then closes back through) or a change in state of delivery (CISD, first-open variant).
Stop (version 1): the lowest low (long) or highest high (short) from the sweep bar through the signal bar.
Target (version 1): the side of the Asia range that was not swept.
Exit (version 1): target, stop, or flat at the 16:59 ET session close. No partial exit and no move of the stop to breakeven.
Re-entry (version 1): after a full stop-out, one more signal may be taken, so at most 2 entries per session, inside the original 30-minute window and after the first trade has exited.
Version 2 (race, no entry rules): from the sweep bar to the session close, a win is price touching the opposite end of the Asia range, a loss is price touching a level 0.5 range-widths beyond the end that was swept. Whichever is touched first decides.
Version 3 (race, no entry rules): the same race with the win level moved to the midpoint of the Asia range.
Version 4: the version 1 entry and stop, with the target set to the closest confirmed, unswept liquidity pool (1-minute bars, lookback from the prior regular session to the entry) that is at least 0.5R from the entry. If no pool is at least 0.5R away, no trade is taken.

Assumptions where the rules were ambiguous:

Where a single bar touches both the stop and the target after the fill, the trade is scored as a loss.
The inverse fair value gap uses the test engine’s own gap geometry. The sweep-and-pool search, the origin gate and the gap-in-leg check were removed for this run, and there is no limit on gap age.
In versions 2 and 3, a bar that touches both levels at once is its own bucket and counts as a loss. Sessions where neither level is reached are left out.
Win rate from geometry is the win rate a driftless random walk gives for those distances: 0.5 / 1.5 = 33.3% in version 2, 50% in version 3. Version 4 uses the average of stop / (stop + target) over its trades as the coin-flip rate.

Data

Instrument: NQ (E-mini Nasdaq-100) futures, continuous contract, CME Group data from Databento (GLBX)
Bar size: 1 minute
Date range: 2016-09-26 to 2026-08-27 (the last ~4 weeks of the ordered window were not in the data)
Asia range: 18:00 to 19:00 ET
Sample size: Version 1: 3,622 trades. Version 2: 2,519 sessions. Version 3: 2,525 sessions. Version 4: 3,479 trades.
Costs: Gross. No commission or slippage is subtracted from any figure. The test specified $2.50 per side plus 1 tick of slippage; none of it is applied here.
In-sample or out-of-sample: The whole history was run in one pass and no part was held out, so every figure is in-sample.
Data gaps: 2020-07-01 is excluded because its 18:00–19:00 range hour is missing from the source data. Versions 2 and 3 also leave out 29 sessions with real data gaps. Version 1 keeps them: without those 29 sessions it has 3,578 trades, an 18.78% win rate and an average of +0.0641R.

Results

Asia range sweep reversal, four versions, NQ 1-minute data, 2016-09-26 to 2026-08-27, gross
Variation	Trades or sessions	Win rate	Win rate from geometry or a coin flip	Average R	Total R	Max drawdown
Version 1: spec entry (sweep, inverse fair value gap or CISD, stop at the sweep-to-signal extreme, target the other side)	3,622 trades	18.72%	not reported	+0.0588 (95% interval −0.0275 to +0.1470)	+213.12R	88.34R
Version 2: race to the other side of the range, loss 0.5 range-widths past the swept end	2,519 sessions (791 wins, 1,728 losses)	31.40%	33.3%	not reported	not reported	not reported
Version 3: race to the range midpoint, loss 0.5 range-widths past the swept end	2,525 sessions (1,159 wins, 1,362 losses, 4 same-bar touches counted as losses)	45.90%	50%	not reported	not reported	not reported
Version 4: version 1 entry and stop, target the closest unswept liquidity pool at least 0.5R away	3,479 trades	28.97%	28.47% (coin flip)	+0.0348 (95% interval −0.0341 to +0.1067)	+121.19R	not reported

Version 1 by half: 2016–2020 had 1,565 trades with an average of −0.0281R; 2021–2026 had 2,057 trades with an average of +0.1250R.
Version 1 exits: 2,944 stops, 676 targets, 2 flat at the close. Median R of the 676 targets: 3.940.
Version 4 by half: 2016–2020 had 1,474 trades with a 25.92% win rate against 26.86% for a coin flip; 2021–2026 had 2,005 trades with 31.22% against 29.64%.
Versions 2 and 3 leave out 9 and 3 unresolved sessions and 1 session with no sweep.

Questions and answers

How was the Asia range sweep reversal backtested?

It was run four ways on NQ futures 1-minute data from 2016-09-26 to 2026-08-27. Version 1 trades a sweep of the 18:00–19:00 ET range with an inverse fair value gap or CISD entry, a stop at the sweep-to-signal extreme and the other side of the range as the target. Versions 2 and 3 are plain races with no entry rules, and version 4 changes the target to the closest unswept liquidity pool at least 0.5R away.

What sample size was used?

Version 1 has 3,622 trades, version 2 has 2,519 sessions, version 3 has 2,525 sessions and version 4 has 3,479 trades, all within 2016-09-26 to 2026-08-27.

Did the Asia sweep backtest include trading costs?

No. Every figure is gross. The test specified $2.50 per side plus 1 tick of slippage, but none of it was subtracted.

What win rate did price reaching the other side of the Asia range have?

In the race with the win level at the opposite end of the range, 31.40% of 2,519 sessions were wins, against 33.3% from geometry alone. With the win level at the range midpoint, 45.90% of 2,525 sessions were wins, against 50% from geometry alone.

Want your own strategy tested the same way? Every version is $19 at cemarketsolutions.com.

More tests like this on Instagram and TikTok: @cemarketsolutions.