MEASURED · 28 AUGUST 2026
Gold makes its daily low in the Asian session 48% of the time
Every session guide says the same thing: skip Asia, wait for London. So we measured where XAUUSD actually sets its daily high and low across 258 sessions. The session everyone skips prints the day's low almost half the time. The session everyone waits for sets either extreme about one day in ten.
What we measured
One year of XAUUSD five-minute bars from a live MetaTrader 5 feed — 70,952 bars covering 31 July 2025 to 4 August 2026. For each complete session we found the bar holding the day's highest high and the bar holding its lowest low, then recorded which block of the trading day each fell in.
Blocks are in broker server time, using the same boundaries a session filter uses, so the numbers are directly comparable to a setting you would actually configure. Server time is usually GMT+2 or GMT+3 and is almost never your local time — if that distinction is new, see server time vs local time.
Sessions with fewer than 200 bars were dropped, which removes holidays and half-days rather than letting a three-hour Christmas Eve session claim its extremes were “made in London”. That left 258 full sessions, with a median daily range of $82.87.
| Block | Server hours | Sets the HIGH | Sets the LOW |
|---|---|---|---|
| Asian | 00–07 | 35.7% | 48.1% |
| London | 08–12 | 9.7% | 11.2% |
| New York | 13–20 | 40.3% | 34.1% |
| Late | 21–23 | 14.3% | 6.6% |
London is the least likely, not the most
London looks weak in that table partly because it is a shorter block — five hours against eight. So the fair comparison is per hour, and it does not rescue London. It makes the result sharper.
| Block | HIGH per hour | LOW per hour |
|---|---|---|
| New York | 5.0% | 4.3% |
| Late | 4.8% | 2.2% |
| Asian | 4.5% | 6.0% |
| London | 1.9% | 2.2% |
Hour for hour, London is the least likely block of the day to print either extreme. An Asian hour is over three times more likely to set the day's low than a London hour is.
That is not an argument that London does nothing. London is where structure typically forms and where the move between the extremes gets travelled. It is an argument that “wait for London because that is where the action is” is describing a feeling rather than a measurement.
The caveat we have to state ourselves
A large share of extremes land in hour 00 — 14.3% of highs and 20.2% of lows. That is a boundary effect and it would be dishonest to present it as Asian volatility. Hour 00 is the session's opening hour, so an extreme sitting there usually means the day trended away from its opening level and never came back, not that the hour itself was busy.
So strip it out. Excluding the opening hour entirely, the Asian block still sets the daily low in 27.9% of sessions across its remaining seven hours — about 4.0% per hour, still comfortably ahead of London's 2.2%. The finding survives its own worst objection, which is the only reason it is worth publishing.
The day's range is almost never made in one session
In only 14.3% of sessions did the high and the low form inside the same block. In the other 85.7%, the day's range was built across at least two sessions.
That is worth sitting with, because a great deal of session advice implicitly assumes the opposite — that you can pick the right session and be present for the whole move. The measurement says the move is usually handed between sessions, and that a strategy confined to one block is, most days, trading a fragment of the range rather than a smaller version of it.
The median block covers a real share of the day even so: New York 65.5% of the full range, Asia 59.9%, London 39.1%. Those overlap rather than sum — each is that block's own span measured against the whole day's.
Why this does not mean “trade the Asian session”
It would be easy to read this as a case for trading the hours everyone else avoids. It is not, for two reasons.
The first is that an extreme is only identifiable afterwards. Knowing the day's low often forms before 08:00 tells you nothing at 03:00 about whether the level in front of you is it. There is no trade in this finding, and anyone selling you one on the back of a statistic like this is selling hindsight.
The second is cost. Our spread study measured the same year and found spread essentially flat across all 24 hours — median $0.16, and above $1.00 in 3 bars out of 70,546. But as a share of the movement available it is 5.4% at 04:00 against 2.3% at 14:00. The quiet hours are not expensive because the spread widens. They are expensive because there is less movement to pay for it with.
The two findings together
Gold's turning points form disproportionately in the hours where each dollar of movement costs the most to capture. That is the reconciliation, and it is why a session filter should select for cost-efficient movement rather than for where the extremes happen to land. “Trade where the highs and lows are” sounds like an edge and is close to the opposite of one.
It also explains a failure mode worth naming. A strategy that hunts reversals in thin hours will find plenty of genuine turning points and still lose, because the cost of participating exceeds what the turn pays. The signal was real. The arithmetic around it was not.
Method, and what would break this
One symbol, one broker's feed, one year, on server time. Daylight saving moves the boundaries against London and New York twice a year and we have not adjusted for it, so treat the block edges as approximate rather than exact. Extremes are taken from the bar containing them, which places an extreme to a five-minute resolution and no finer.
258 sessions is a real sample but not a large one. A block splitting 35% against 40% is well inside what a year of data can produce by chance, so we are not claiming New York beats Asia for highs. The claim is the wide gap — London against everything else, and the Asian low share — which is far too large to be noise at this sample size.
The script is research-extreme-timing.py and it runs on the same bar archive as the session breakout study and the spread study, so all three are directly comparable. If you want the schedule rather than the statistics, that is gold trading hours.