Key takeaways
- In forex the fade is a coin flip. 9,598 resolved breaks across 13 FX pairs reversed 49.4% of the time. z = −1.25, p = 0.213. The test is designed to resolve at 50%, so that is nothing.
- In crypto the break follows. ETH/USD continued 58.2% (z = −4.23, p < 0.001) and BTC/USD 55.0% (z = −2.67). ETH/USD is the only instrument in the study to clear a Bonferroni correction.
- Indices whipsaw constantly. US30 breaks both sides of the Asian range on 68.1% of days; US100 and US500 on 62.8%. "The Asian range got swept" describes most index days, not special ones.
- The pooled number is misleading. All 22 together give 48.95%, p = 0.005 — but that significance is entirely crypto. Drop two instruments and it vanishes.
- Quiet crosses barely break at all. AUD/JPY whipsaws on 25.3% of days, AUD/USD on 30.5%. Same rule, completely different frequency.
First, in plain English
While London and New York sleep, price usually drifts in a narrow band. Take the high and low of that quiet stretch and you have the Asian range. The popular rule says the first push out of it during the London session is a trap — a sweep of the stops resting outside the box — and that price reverses back through the range afterwards.
To test "reverses" you need to define it against something, or a win rate means nothing. So we set up a symmetric race with no free parameters.
Step 1: draw the box. The high and low of 00:00 to 06:59 UTC. Say EUR/USD ranged 1.1000 to 1.1020 — a 20-pip box.
Step 2: wait for the first break. Between 07:00 and 20:45 UTC, note the first time price trades outside the box. Suppose it goes up through 1.1020.
Step 3: race two outcomes that are exactly the same distance away. From that break level:
- Continuation wins if price reaches 1.1040 — one full box-height above the break.
- Reversal wins if price reaches 1.1000 — the opposite side of the box, which is also exactly one box-height away.
Both targets sit one range height from the entry. That symmetry is the whole point: it means a market with no pattern in it resolves at 50.00%, and any deviation is a real signal rather than an artifact of picking a generous target.
The answer in forex: 49.4% reversal. On 9,598 resolved races across 13 FX pairs, the fade came in six-tenths of a point below a coin flip, which is not statistically distinguishable from one. Whatever is happening at the Asian range in forex, it does not survive being written down as a rule.
How this was measured
- Data. Dukascopy bid-side M15 OHLC for 22 instruments — 13 FX pairs, gold, silver, Brent crude, four equity indices and two crypto pairs — spanning 2021-12-27 to 2026-08-14. Coverage is capped at 100,000 bars per instrument, so FX begins in August 2022 and crypto in October 2023.
- The Asian range. High and low of 00:00–06:59 UTC, fixed year-round. Hours are UTC throughout so the definition never drifts with daylight saving; see our session hours guide for how local clocks map onto it.
- The break. The first M15 bar between 07:00 and 20:45 UTC to trade through either side. Whichever side goes first sets the direction.
- The race. Continuation target one range height beyond the break level; reversal target the opposite side of the range, also one range height away. Symmetric by construction, fair at 50%.
- Minimum data. A day needs at least 20 of the 28 possible Asian-session M15 bars and 40 bars in the break window, so thin holiday sessions do not enter the sample.
- Unresolved days. Days where neither target is reached by the end of the UTC day are reported but excluded from the percentage — hence 17,593 resolved out of 23,520.
- Significance. One-sample z-test against the designed-fair 50%, since the study's symmetry supplies the baseline rather than a separate control arm.
The result
Sorted by reversal rate. The whipsaw column — how often price breaks both sides of the box on the same day — turns out to be the more useful number for most readers.
| Instrument | Days | Resolved | Reversal | Continuation | Breaks both sides | z vs 50% | Significant? |
|---|---|---|---|---|---|---|---|
| US30 (Dow) | 1,128 | 1,076 | 51.6% | 48.4% | 68.1% | +1.04 | ns |
| EUR/CHF | 1,041 | 922 | 51.5% | 48.5% | 55.3% | +0.92 | ns |
| USD/CAD | 1,041 | 902 | 50.9% | 49.1% | 52.3% | +0.53 | ns |
| Silver | 1,091 | 711 | 50.6% | 49.4% | 38.7% | +0.34 | ns |
| NZD/CHF | 1,041 | 717 | 50.5% | 49.5% | 40.0% | +0.26 | ns |
| US500 (S&P) | 1,143 | 1,066 | 50.5% | 49.5% | 62.8% | +0.31 | ns |
| US100 (Nasdaq) | 1,143 | 1,057 | 50.4% | 49.6% | 62.8% | +0.28 | ns |
| AUD/JPY | 1,041 | 477 | 50.1% | 49.9% | 25.3% | +0.05 | ns |
| AUD/USD | 1,040 | 551 | 50.1% | 49.9% | 30.5% | +0.04 | ns |
| CAD/CHF | 1,041 | 945 | 49.9% | 50.1% | 57.8% | -0.03 | ns |
| Brent crude | 1,197 | 1,067 | 49.6% | 50.4% | 55.8% | -0.28 | ns |
| AUD/CAD | 1,041 | 586 | 49.3% | 50.7% | 31.7% | -0.33 | ns |
| EUR/GBP | 1,041 | 926 | 49.2% | 50.8% | 54.0% | -0.46 | ns |
| EUR/USD | 1,040 | 850 | 48.8% | 51.2% | 48.5% | -0.69 | ns |
| USD/JPY | 1,040 | 578 | 48.8% | 51.2% | 29.9% | -0.58 | ns |
| GBP/USD | 1,040 | 861 | 48.3% | 51.7% | 48.3% | -0.99 | ns |
| GBP/JPY | 1,041 | 663 | 47.2% | 52.8% | 34.8% | -1.44 | ns |
| Gold | 1,090 | 719 | 46.3% | 53.7% | 37.5% | -1.98 | * |
| GER40 (DAX) | 1,125 | 931 | 46.3% | 53.7% | 47.3% | -2.26 | * |
| EUR/JPY | 1,041 | 620 | 45.8% | 54.2% | 30.9% | -2.09 | * |
| BTC/USD | 1,037 | 705 | 45.0% | 55.0% | 37.1% | -2.67 | ** |
| ETH/USD | 1,037 | 663 | 41.8% | 58.2% | 32.2% | -4.23 | *** |
Pooled across all 22: 23,520 days, 17,593 resolved races, 48.95% reversal, z = −2.80. Pooled across the 13 FX pairs only: 9,598 resolved, 49.36%, z = −1.25, not significant.
Why the pooled number is the wrong number
Taken together, all 22 instruments give a reversal rate of 48.95% with p = 0.005 — apparently significant, and apparently evidence that the break continues rather than fades. Publishing that as the headline would be a mistake.
Remove BTC/USD and ETH/USD and the pooled figure moves to 49.41% with p = 0.134. Remove Brent as well and it is 49.40%, p = 0.139. Take only the 13 FX pairs and it is 49.36%, p = 0.213. The significance was never a property of the Asian range; it was two crypto instruments with a genuine effect being averaged into twenty that have none.
This is worth dwelling on because it is the most common way an honest backtest produces a dishonest conclusion. Pooling across instruments feels more rigorous than testing one. It is only more rigorous if the instruments actually share the behaviour being measured.
The crypto result, and why we are hedging it
ETH/USD continued through the first Asian-range break 58.2% of the time across 663 resolved races, z = −4.23. BTC/USD continued 55.0%, z = −2.67. Applying the Bonferroni threshold for 22 tests — p < 0.0023, or |z| > 3.05 — ETH/USD is the single result in this entire study that survives. Gold, GER40 and EUR/JPY all show nominal significance and none of them clear the correction.
So the finding is real as far as the test goes, and it points the opposite way to the rule: in crypto the Asian-range break is a continuation signal, not a trap.
Three reasons to hold it loosely anyway. Crypto coverage begins only in October 2023, so this is under three years against roughly four for FX. That window is also an unusual one by any measure, and a momentum-friendly regime would produce exactly this result without the mechanism being durable. And crypto trades through the weekend, so its "Asian session" is a block of hours with no institutional handover in it at all — the label barely applies.
The result is worth trading against only after someone re-runs it on a longer sample. It is published here because it is what the data says, not because we are confident it will hold.
The number most readers should actually use
Set the fade question aside; the whipsaw column is more practically useful than the reversal column.
On the equity indices, price breaks both sides of the Asian range on the same day 68.1% of the time for US30, and 62.8% for both US100 and US500. Two days in three, the phrase "the Asian range was swept" is true in both directions. It cannot be a selective signal if it happens on nearly every day — the same trap our London opening range study found, where the range broke on 99.8% of days.
At the other end, the quiet crosses barely leave the box at all: AUD/JPY whipsaws on 25.3% of days, AUD/USD on 30.5%, USD/JPY on 29.9%. The identical rule, applied to a different instrument, fires less than half as often. Any Asian-range rule that was tuned on an index and then moved to a cross is a different strategy with the same name.
The spread across instruments — 25.3% to 68.1% — is also a reasonable proxy for how much overnight participation each market gets, which is the same thing our hour-by-hour volatility study measures directly.
What this does not say
It does not say the Asian range is meaningless. Ranges are real, the stops above and below them are real, and a 41.5% whipsaw rate in FX means the level is being interacted with constantly.
What it says is that the first break, taken mechanically, in either direction, is not predictive in forex. Practitioners will point out that the real rule includes a higher-timeframe bias, a specific killzone window, or a requirement that the sweep take out a prior day's level too — the last of which we measured separately in the liquidity grab study, where it also came out at random. Those objections are legitimate and this test cannot settle them. What it does supply is the number each of those filters has to beat: 50%, by more than the spread.
Limitations
Stated plainly, because a study without them is marketing.
- The Asian range is fixed at 00:00-06:59 UTC year-round. It does not track the Tokyo cash session, and it does not shift with daylight saving.
- M15 resolution. A break and its opposite target inside the same bar cannot be ordered.
- Coverage is capped at 100,000 bars per instrument - roughly four years for FX, under three for BTC/USD and ETH/USD. The crypto finding rests on the shortest sample in the study.
- Unresolved days are excluded from the percentages rather than scored. On the quieter crosses that removes a large share of the sample - AUD/JPY resolves only 477 of 1,041 days.
- No spread, commission or swap applied. A symmetric race at 50% becomes a losing trade once costs are paid.
- Bid-side data. Ask prices are not replayed, so a break measured on the bid is marginally early for a long and late for a short.
- One range definition and one target multiple. A half-range target or a 2x extension is a different study.
Download this dataset
Every figure on this page as raw CSV and JSON, regenerated whenever the archive updates. Free for any use, including commercial, with attribution to FxBacktest.
- asian-range-first-break-by-instrument.csvAsian range first break by instrument
- asian-range-break.jsonAll tables plus metadata, one file
Cite or republish this data
This research is free to quote, screenshot, or republish — in an article, a video, a newsletter, or a course — as long as you credit FxBacktest with a link back to this page. Copy the attribution below.
Frequently asked
Does the Asian range sweep actually reverse?
In forex, no. Across 9,598 resolved breaks on 13 FX pairs, the first break of the Asian range reversed 49.4% of the time on a test designed to resolve at 50% in an efficient market. z = -1.25, p = 0.213, not significant. Fading the Asian range break is a coin flip before costs and a loss after them.
How often does price break both sides of the Asian range?
44.9% of days across the full sample, but the split by asset class is wide. Equity indices whipsaw most - US30 breaks both sides on 68.1% of days, US100 and US500 on 62.8% - while the quieter FX crosses sit near 25% to 31%. On an index, the phrase 'the Asian range was swept' describes most days rather than a select few.
Does the Asian range work in crypto?
It works in reverse. ETH/USD continued through the first break 58.2% of the time (z = -4.23) and BTC/USD 55.0% (z = -2.67). ETH/USD is the only instrument in this study that survives a Bonferroni correction for 22 tests. The caveat is that crypto coverage only begins in October 2023, so this is under three years of an unusual period.
What time is the Asian range in this study?
00:00 to 06:59 UTC, fixed year-round, with the first break measured between 07:00 and 20:45 UTC. A fixed UTC definition never drifts with daylight saving, which makes the measurement reproducible but means it does not track the Tokyo cash session exactly. See our forex trading sessions guide for how local clocks map onto UTC.
What data was used?
Dukascopy bid-side M15 OHLC for 22 instruments spanning 2021-12-27 to 2026-08-14. Coverage is capped at 100,000 bars per instrument, so FX pairs begin in August 2022 and the two crypto instruments in October 2023. Days needing at least 20 Asian-session bars and 40 bars in the break window were kept.