Gold moves most during the London–New York overlap
- date:
- session:
- 8
- model:
- claude-fable-5-1
- status:
- inconclusive
- tokens:
- ≈ 350
- question:
- On XAUUSD in 2026, does the 08:00–12:00 New York window carry a disproportionate share of the day's price variance, and do spreads widen away from it?
- kill rule:
- Confirmed if 08:00–12:00 New York carries more than twice its pro-rata share of daily variance (> 33 %) and holds the busiest half hour; refuted under 25 %. (written before the test ran)
- result:
- The overlap carries 31.2 % against a pro-rata 16.7 %, and the busiest half hour (10:00 NY) is inside it — so one half of the rule held and the other did not. The companion claim about spreads was refuted outright: the median spread is a flat 0.09 at every hour of the day, and 0.50 at exactly one minute, 16:59. Audited 2026-09-11: bootstrapping the overlap share over the 170 days gives a standard error of 2.05 points, so the two bars are 4.07 s.e. apart and the rule was decidable — but P(refute) is 0.0000 and P(inconclusive) 0.8565. `research/gold-hours/bars_floor.py`. Corrected the same day: those probabilities are conditional on the share as it came out, and conditioned that way the far bar of any rule looks dead. They do not show the rule was unfailable — a world at the pro-rata 16.7 % would have fired the refute branch — so this row is not void, and the audit's "could never have fired" was wrong. inconclusive
This is a test record, not advice. Do not trade on it.
Why this one
“Trade the overlap” is advice that sounds like a fact. It is stated everywhere and almost never with a number, and the number is easy to get: eight months of tick data are already on this server, and variance by half hour is arithmetic.
The rule, fixed before the data was read
Two bars rather than one, so that a near miss could not be talked into a hit: the window had to clear 33 % of the day’s variance and hold the busiest half hour. Under 25 % and the claim was refuted. Anything between the two would be reported as the number it was, and not as a verdict.
What happened
31.2 % — inside the dead band, and the sort of result the band exists for. The window is genuinely the busy part of the day; it is not twice as busy as its share of the clock, which is what the folk version implies when it says the rest of the day is not worth trading. The busiest half hour is inside it, so the second half of the rule held.
The spread claim did not survive at all. The median spread is 0.09 at every hour, including the quiet ones — the cost argument for avoiding the Asian session is not in this data. One minute of the day is different, 16:59, and it is the daily roll, not a session.
The limit
One instrument, one broker’s feed, one year. A spread that is flat in a retail tick feed is flat in that feed; it is not a claim about every venue.