MEASURED · 250 TRADING DAYS
What share of gold’s daily range happens during New York hours?
Most of it, not all. Over 250 trading days of XAUUSD on a MetaTrader 5 feed, New York hours (8 a.m. to 5 p.m. ET, 7 a.m. to 4 p.m. CT) spanned a median 72.0% of the day’s high-to-low range and carried 43.4% of its movement in 9 of the 23 trading hours. That is more than its share of the clock, and all of the surplus came before noon ET.
The rest of the range is made overnight. On 38 of the 250 days New York never traded outside the range set between 6 p.m. and 8 a.m. ET. On days with a US jobs report, CPI release or Fed decision, New York’s span rose to a median 84.7%.
How the day was cut
The data is 71,402 five-minute XAUUSDm bars from an Exness MetaTrader 5 account, 2025-08-01 to 2026-08-04, the archive behind our spread and volatility by hour study. The feed runs on UTC. Every bar was converted to New York time, so the sessions follow US daylight saving instead of sliding an hour twice a year.
A gold day runs from 6 p.m. ET (5 p.m. CT) the evening before to 5 p.m. ET (4 p.m. CT; 21:00 GMT in US summer, 22:00 in winter), where the feed itself pauses: on ordinary weekdays the 5 p.m. ET hour had no bars, in summer and in winter. Days missing more than an hour of their 276 five-minute bars were dropped: 11 in all, 9 of them US holidays or early closes and the rest part-days at the ends of the archive. That leaves 250 full days.
| Session | New York (ET) | Chicago (CT) | GMT, US summer / winter |
|---|---|---|---|
| Asia | 6 p.m. to 3 a.m. | 5 p.m. to 2 a.m. | 22:00-07:00 / 23:00-08:00 |
| London | 3 a.m. to 8 a.m. | 2 a.m. to 7 a.m. | 07:00-12:00 / 08:00-13:00 |
| New York | 8 a.m. to 5 p.m. | 7 a.m. to 4 p.m. | 12:00-21:00 / 13:00-22:00 |
| New York morning | 8 a.m. to noon | 7 a.m. to 11 a.m. | 12:00-16:00 / 13:00-17:00 |
| New York afternoon | noon to 5 p.m. | 11 a.m. to 4 p.m. | 16:00-21:00 / 17:00-22:00 |
“London” means London’s morning before New York arrives. Its 8 a.m. is 3 a.m. ET, except on the 20 days when the US and UK clocks change on different dates, when it is 4 a.m. ET.
Two ways to measure “share of the range”
Span is the session’s own high minus its low, divided by the whole day’s high minus low. A session that printed both the day’s high and its low scores 100%. Sessions overlap in price, so the three spans add up to more than 100%.
Movement adds up the high-to-low range of every five-minute bar in the session and divides it by the same sum for the whole day. This one does add up to 100%, so it can be set against the clock: New York holds 108 of the day’s 276 bars, 39.1%.
| Session | Median span | Span, share of day (middle half) | Share of movement | Share of the clock |
|---|---|---|---|---|
| Asia | $51.24 | 60.7% (44–76%) | 37.2% | 39.1% |
| London | $30.82 | 37.1% (27–50%) | 19.4% | 21.7% |
| New York | $59.39 | 72.0% (58–86%) | 43.4% | 39.1% |
| New York morning | $49.49 | 58.9% (45–76%) | 24.5% | 17.4% |
| New York afternoon | $30.38 | 35.7% (27–47%) | 18.9% | 21.7% |
Middle half: the 25th to 75th percentile of the daily shares. Share of movement is pooled over all 250 days; the median day’s share is within 1.5 points of it for every session.
Our daily high and low timing study asks a different question, which block prints the high and which the low, on the broker’s clock.
More than nine hours alone would give it
Nine of 23 hours would cover much of the day’s range even if gold moved at an even pace. To test that, we shuffled the order of each day’s bars 200 times and measured New York’s span again. With the timing removed, New York spanned a median 64.1% of the day. The real figure was 72.0%.
On movement, New York carried 43.4% of all five-minute movement on 39.1% of the clock, and the surplus came from the morning. From 8 a.m. to noon ET, four hours, 17.4% of the clock, gold made 24.5% of the day’s movement. The three hours from 8 to 11 a.m. ET alone made 19.2%. For most of the year that window also holds the London afternoon gold auction at 10:00 a.m. ET; our London fix study measures what that minute does on its own. Noon to 5 p.m. ET made 18.9% on 21.7% of the clock, less than an even spread would give it. The quietest hour of the whole day started at 4 p.m. ET, with 2.8% of the day’s movement.
Day by day, New York’s span was wider than Asia’s on 62.0% of days and wider than London’s on 90.0%. It carried the most movement of the three sessions on 72.8% of days, and spanned at least half the day’s range on 84.0%.
What New York adds to the overnight range
A bot that trades New York starts with a range already on the chart. From 6 p.m. to 8 a.m. ET gold ranged a median $60.59; the full day ranged $88.09. New York traded above the overnight high on 53.2% of days and below the overnight low on 39.6%. It broke one side only on 76.8%, both sides on 8.0%, and neither on 15.2%.
The part of the day’s range that only New York reached, beyond the overnight high or low, was a median 26.7% of the day. So by 8 a.m. ET the overnight hours had already drawn a median 73.3% of the day’s range. New York’s own span is about as large, 72.0%, because the two overlap: New York usually trades back through much of the overnight range.
Which way a break goes is a separate question; our session breakout study found it close to a coin flip.
News days, and how steady it is
On the 28 days in the sample with a US jobs report, CPI release or Fed decision (dates from our NFP, CPI and FOMC studies), New York spanned a median 84.7% of the day and carried 48.7% of its movement. On the other 222 days: 70.8% and 42.8%. New York stayed inside the overnight range on 2 of the 28 release days (7.1%), against 16.2% of other days. 28 days is a small sample, and the release days were bigger days overall: a median range of $100.30 against $86.85.
| Period | Days | Median day range | New York span | New York movement | Inside overnight range |
|---|---|---|---|---|---|
| Aug to Sep 30, 2025 | 41 | $42.38 | 75.4% | 43.6% | 22.0% |
| Oct to Dec 31, 2025 | 62 | $77.37 | 76.4% | 44.2% | 16.1% |
| Jan to Mar 31, 2026 | 61 | $132.64 | 75.2% | 43.8% | 11.5% |
| Apr to Jun 30, 2026 | 62 | $100.40 | 64.4% | 41.9% | 17.7% |
| Jul to Aug 4, 2026 | 24 | $88.09 | 66.4% | 43.7% | 4.2% |
The size of the day changed a lot, from a median $42.38 to $132.64 by quarter. New York’s share moved much less: its span between 64.4% and 76.4%, its movement between 41.9% and 44.2%. The first and last rows are part-quarters.
What to do with this if you run a bot
Size targets to the session, not the day. A bot that trades only 8 a.m. to 5 p.m. ET had a median $59.39 of room, not the day’s $88.09. New York printed both the day’s high and its low on only 20 of 250 days. A take-profit or stop set as a multiple of daily ATR is sized for a range a New York-only bot rarely sees in full.
If the strategy needs movement, look at the morning. 8 a.m. to noon ET (7 to 11 a.m. CT) carried 24.5% of the day’s movement in four hours. The afternoon carried less than its share of the clock, and the same spread is paid against a smaller move there (see the hourly spread study). A noon ET cut-off gives up 18.9% of the day’s movement; whether that helps depends on the strategy, and only your own backtest can say. The same movement takes out tight stops: see how often a $3, $5 or $10 stop is hit within an hour, session by session.
Expect days with nothing to break. A bot that trades a break of the overnight range during New York hours had no break at all on about one day in seven (15.2%), and a break on both sides on about one day in twelve (8.0%).
Put session times on the New York clock. This feed’s server clock is UTC, so New York’s 8 a.m. is 12:00 on the server in US summer and 13:00 in winter. A filter typed in server hours drifts an hour against New York twice a year. Check which clock your EA’s session inputs use, the Tech Kick Trading Bot included. Trading from the US? Read whether US traders can use MT5 trading bots first: the instrument measured here is a spot gold CFD.
Method and limits
One broker, one spot CFD, one year: 250 full days from 2025-08-04 to 2026-08-04. Extremes are placed to the five-minute bar, not the tick. The movement measure adds bar ranges, so back-and-forth chop counts: it measures activity, not distance travelled. The control keeps each bar’s shape and shuffles only the order, 50,000 shuffled days with a fixed seed. The release-day split rests on 28 days, and one year cannot tell a seasonal pattern from a one-off. Every figure comes from one script, research-gold-new-york-session-range.py, run on the same archive as our other studies.
Common questions
What are New York trading hours for gold?
In this study, 8 a.m. to 5 p.m. ET, which is 7 a.m. to 4 p.m. CT and 12:00 to 21:00 GMT in US summer (13:00 to 22:00 GMT in winter). It ends at gold's daily pause: on this MT5 feed no bars printed from 5 to 6 p.m. ET on ordinary weekdays, and the next day's Asian session starts at 6 p.m. ET.
Is the London and New York overlap the busiest time for gold?
On this feed, yes. The three busiest hours were 8 to 11 a.m. ET, London's afternoon, and they carried 19.2% of the day's five-minute movement on 13.0% of the clock. The busiest single hour started at 10 a.m. ET, with 6.9%.
Does gold move more in the Asian session or the New York session?
New York, on most days. Both sessions are nine hours long here. New York's span was wider than Asia's on 62.0% of days: a median 72.0% of the day's range against 60.7%. It carried 43.4% of the movement against Asia's 37.2%, so Asia still carried more than a third.