Phase 4 / GBPUSD study
GBPUSD London breakout, tested on real M1 data
A fixed-window test of every qualifying GBPUSD weekday, with ordered event rules that never guess the path inside a one-minute candle.
GBPUSD broke at least one Asian-range boundary on 98.2% of 1,325 qualifying dates. Among resolved ordered first breaks, a 0.50R extension arrived before the opposite boundary 71.4% of the time. Costs and trade exits were not modeled.
What the GBPUSD test found
Across 1,325 qualifying weekdays, at least one side of the 00:00-07:59 UTC range broke during 08:00-15:59 UTC on 1,301 dates. That is 98.2%, with a Wilson 95% confidence interval of 97.3-98.8%. The high was the first ordered break on 50.2% of usable break dates, almost exactly balanced with the low. The test therefore supports frequent range expansion, not a directional forecast.
The median Asian reference range was 39.1 pips and the median range observed in the next eight hours was 67.5 pips. Expressed in each date's own scale, the later window's median high-low span was 1.694 times the Asian range. A large later range does not mean a trader could capture it from one entry; it only establishes the price territory that existed.
The first ordered boundary break arrived a median 30 minutes after 08:00 UTC. On 31.4% of qualifying dates both boundaries eventually broke, while 1.8% broke neither. Both-side dates matter because a simple statement such as 'the Asian high broke' can hide a later full reversal through the low.
Continuation changes with the target definition
| Target | Eligible | Target first | Opposite first | Same-minute ambiguous | Unresolved | Continuation among resolved |
|---|---|---|---|---|---|---|
| 0.25R | 1,301 | 1,008 | 186 | 0 | 107 | 84.4% 95% CI 82.3-86.4% |
| 0.50R | 1,301 | 750 | 301 | 2 | 248 | 71.4% 95% CI 68.6-74.0% |
| 1R | 1,301 | 412 | 383 | 1 | 505 | 51.8% 95% CI 48.4-55.3% |
At 0.25R, 1,008 dates reached the continuation threshold first and 186 reached the opposite boundary first. Among resolved ordered outcomes that is 84.4% continuation versus 15.6% false break. At 1.00R, continuation among resolved outcomes fell to 51.8% and 505 dates reached neither event by 16:00 UTC.
How to read continuation and false-break rates
Suppose the Asian range is 40 pips and the high breaks first. A 0.50R continuation requires a later complete M1 candle to print at least 20 pips above the Asian high before a prior complete minute breaks the Asian low. If the low breaks first, the calculation is mirrored. The first breakout candle is excluded from this ordering test because OHLC alone cannot reveal which price occurred first inside that minute.
The phrase false break is target-dependent here. It means the opposite Asian boundary appeared before the stated extension. A date can be a false break under the 1.00R definition but a continuation under 0.25R. That is why this study publishes a ladder of targets instead of one promotional win-rate number.
Weekday evidence, without pretending it is a rule
| Weekday | Days | Any break | Median Asian range | 0.50R continuation* | 0.25R false break* |
|---|---|---|---|---|---|
| Monday | 266 | 98.1% | 40.5 pips | 71.2% | 17.4% |
| Tuesday | 263 | 99.2% | 39.8 pips | 72.5% | 14.9% |
| Wednesday | 265 | 97.4% | 40.0 pips | 73.2% | 15.3% |
| Thursday | 267 | 99.3% | 38.0 pips | 72.4% | 16.1% |
| Friday | 264 | 97.0% | 36.9 pips | 67.5% | 14.0% |
*Continuation and false-break percentages use resolved ordered outcomes. Ambiguous and unresolved dates remain visible in the evidence file and are not silently forced into either result.
Wednesday had the highest descriptive 0.50R continuation rate at 73.2%, while Friday was lowest at 67.5%. These are historical slices, not independent strategy trials. Before adding a weekday filter, replay the exact rule with costs and out-of-sample dates.
Practical reading
- A boundary break was common, but common does not mean profitable after entry costs and stops.
- First direction was close to a coin flip, so the range itself did not supply directional bias.
- The claimed continuation rate depends materially on whether continuation means 0.25R, 0.50R, or 1.00R.
- Unresolved and same-minute ambiguous dates are published instead of being assigned favorable outcomes.
Audit the inputs, exclusions, and every qualifying date
This page is generated from the same two binary files served by FXAbsolute. The analyzer validates each header and every timestamp before producing a result. The public evidence file includes the full SHA-256, coverage audit, filter counts, confidence intervals, and a compact observation for every qualifying instrument-day.
| Source | M1 candles | Qualifying days | Coverage | SHA-256 |
|---|---|---|---|---|
| EURUSD_M1.bin | 1,911,141 | 1,334 | 96.4% | 5afa4be2ed37b3170c58f71a86d943ac1216639bc4d64b312603e6072901a9dd |
| GBPUSD_M1.bin | 1,898,746 | 1,325 | 95.8% | ed16e48b0be764c42fe55dd1aacf3f028c519685f18c45ba226f646773bd1e10 |
Open the machine-readable proof JSON
Validation result: zero duplicate timestamps, zero out-of-order timestamps, zero non-minute timestamps, zero invalid OHLC rows, and exact header-length matches in both files. Missing dates and under-covered windows are reported; no candle is synthesized.
Exactly what was tested
1. Build the reference range
For each UTC weekday, the analyzer takes the highest bid high and lowest bid low observed from 00:00 through 07:59 UTC. At least 420 of the possible 480 M1 candles must exist.
2. Observe the next window
Boundary breaks are measured from 08:00 through 15:59 UTC, again requiring at least 420 observed candles. A high or low must move strictly beyond the reference boundary; an equal print is not a break.
3. Preserve event order
The first boundary is found in timestamp order. Ordered target-versus-opposite results begin with the following complete minute. If both thresholds appear inside one M1 bar, the date is labeled ambiguous.
4. Publish unresolved cases
A date that reaches neither the continuation target nor the opposite boundary by 16:00 UTC is unresolved. It stays in the denominator where stated and is never converted into a win or loss.
The continuation distances are expressed as a fraction of that date's own Asian range: 0.25R, 0.50R, and 1.00R. This normalization makes a 20-pip reference day comparable with a 60-pip reference day without pretending that their cash risk is the same.
Continue through the Phase 4 evidence
Replay the underlying markets in the EURUSD backtester or GBPUSD backtester with your own entry, cost, stop, and exit rules.
Questions this dataset can answer
Does this prove the London breakout strategy is profitable?
No. It proves historical boundary-break and ordered price-event frequencies in bid M1 data. Profitability requires a fully specified entry, spread, slippage, stop, exit, position size, and risk rule.
What is the Asian range in this test?
The Asian reference range is the highest bid high minus the lowest bid low observed from 00:00 through 07:59 UTC on a weekday. This is a fixed research window and does not shift for daylight saving.
What does 0.50R continuation mean?
After the first ordered break, a later complete M1 candle must extend half of that date's Asian-range size beyond the broken boundary before an earlier complete minute breaks the opposite boundary.
Can an AI crawler verify the numbers?
Yes. The public JSON at https://fxabsolute.com/research-data/phase-4-evidence.json includes source hashes, validation counts, methodology, aggregate statistics, and every qualifying daily observation.