Gold trading hours:what time does gold open? The four sessions and the best hours, measured

People who search for "gold trading hours" want two things: what time gold opens, and what time is best to trade. The most common answer online is "eight in the evening to midnight". That is not wrong, but it merges three different questions into one: which window has the biggest range, which window has the tightest spread, and which window suits 60-second trades versus only 5 minutes or more. In our October 2026 measurements the three answers do not agree — especially the third, which is exactly where many traders get direction right and still lose money.

This article does four things: it sets the four sessions (Sydney, Tokyo, London, New York) out in Beijing time in one table, recalculated on the post-2026-10-25 EU winter-time basis; it measures the realised range of each window from our own quote series over the past three weeks and 15 complete trading days; it measures the spread in each window (the conclusion is slightly counter-intuitive); and it gives the holding period that fits each of the four windows, plus the exact times of the weekend close and the daily maintenance break.

This update (2026-10-25 02:40 Beijing time): the timetable and overlap windows in section 1 have been recalculated for the EU winter-time switch, and the "what time does gold open" / "best time to trade gold" wording has been written into the subheadings and the FAQ; the samples in the range and spread sections have not been recalculated — they remain the 2026-10-23 23:58 sampling.

1. What time does gold open? The four sessions in Beijing time

Gold is quoted continuously around the clock, but "continuous 24 hours" does not mean every hour is worth trading. The market splits geographically into four sessions, each backed by completely different money: Sydney is Australian and New Zealand physical flow, Tokyo is Asian physical flow plus yen carry trades, London is European institutions and the fixings, New York is US institutions, the futures front month and data releases. So the full answer to "what time does gold open" is four times, not one.

Basis and sources (verifiable yourself): Beijing time = UTC+8, and "Beijing" throughout means UTC+8. EU summer time ends at 2026-10-25 01:00 UTC (last Sunday of October), London returns to GMT and 08:00 GMT = 16:00 Beijing; US summer time ends at 2026-11-01 02:00 EDT (first Sunday of November), when New York moves from 20:00–05:00 to 21:00–06:00; Australia has been on AEDT since 2026-10-04 (UTC+11), and 08:00 AEDT = 05:00 Beijing; Japan does not observe summer time, so Tokyo is unchanged all year. The "10-24 and earlier" column is the summer-time basis used in the previous article.

While we are here, count how many hours you can actually place an order: from the Monday 08:00 open to the Saturday 04:00 close the span is 116 hours; deduct the daily 04:00–08:00 maintenance (4 hours × 5 days = 20 hours) and roughly 96 hours remain tradeable (the exact close and maintenance fields are quoted in section 7). So "gold trades 24 hours" describes quoting, not order entry — which is why so many orders left overnight never fill.

The rows that go wrong most often are London and New York. From 09:00 Beijing today (2026-10-25) Europe is on winter time, so London moves from 15:00–24:00 to 16:00–01:00; after 1 November the US follows and New York moves from 20:00–05:00 to 21:00–06:00. In other words the widely copied "London 15:00, New York 20:00" table is only valid during northern-hemisphere summer time. Conversely, in the seven days from 25 October to 1 November, Europe has switched and the US has not: the Europe-US gap is only four hours and the overlap stretches to 20:00–01:00, the widest of the year. In those days the market visibly "keeps moving all evening".

One more trap for overseas Chinese readers: if you watch the market from Singapore or Kuala Lumpur the time zone matches Beijing and the table above applies directly; from Sydney subtract 3 hours (AEDT is 3 hours ahead of Beijing), from London subtract 8 (GMT now, 7 during summer time), from New York subtract 12 (EDT; 13 after the switch to EST on 1 November). If you live across time zones, just use the live session clock, which draws the four sessions in your own local time.

2. The best time to trade gold: the 20:00–01:00 overlap, core 21:00–24:00

Only two overlaps are real. Sydney and Tokyo overlap 08:00–14:00 (Sydney opens 05:00 and closes 14:00; Tokyo opens 08:00 and closes 17:00), and London and New York overlap 20:00–01:00 (2026-10-25 to 11-01 basis). There is in fact no Sydney-London overlap: Sydney closes at 14:00 and London does not open until 16:00, a two-hour gap (one hour even under summer time). The much-quoted "21:00–24:00 Beijing" is the core three hours of the second overlap: under summer time it is the last three hours of the overlap (20:00–24:00), under winter time the first three (21:00–01:00). Whichever way the clocks switch, those three hours always fall inside the overlap, which is why the figure keeps being quoted.

There are only three mechanisms behind the thicker overlap:

• Two sessions are quoting at once. Market makers and liquidity providers are online together, order-book depth stacks up, the same market order eats fewer levels and slippage is smaller.
• Institutional rebalancing windows collide. The London 15:00 afternoon fix (LBMA Gold Price PM), the New York 08:30 data window and the 21:30 US equity open all crowd into these few hours (23:00, 21:30 and 22:30 Beijing under winter time; each an hour earlier under summer time).
• US economic data clusters in 20:30–22:00 Beijing (summer-time basis). Non-farm payrolls, CPI, retail sales and jobless claims almost all land inside that 90 minutes, and when a number prints, quotes gap and volume doubles. After the US switches on 1 November the whole window shifts an hour later to 21:30–23:00 — anyone trading the data needs to reset the alarm that day.

But "thicker" is a statistical statement; it does not pay out every day. The next two sections show the average and today separately.

3. Measurement one: the realised range of each session over the past three weeks

Window note: the 20:00–24:00 "overlap" row is bucketed on the summer-time basis of 2026-10-04 to 10-23 and does not exactly coincide with today's overlap of 20:00–01:00 (from 10-25) — the core three hours, 21:00–24:00, sit inside both and stay comparable across the switch; the extra 00:00–01:00 hour falls outside this sample, so there is no measurement for it and this article puts no estimate on it.

Three things have to be read out of this table:

• "Trade in the evening" has data behind it. The 20:00–24:00 overlap averages 7.82 USD, 2.5× the quiet window. The same order has 2.5× the room in the evening as it does at dawn, while the spread is unchanged (next section) — that is a free improvement in the odds.
• The wildest hour inside the overlap is 20:00–21:00, not after 22:00. The whole window is 7.82 and the core three hours 7.54, and the 0.28 difference sits entirely in 20:00–21:00 — because 20:30 is the US data window (summer-time basis; 21:30 under winter time). So "21:00–24:00 is best" should be rephrased as "be in position by 20:00; sustained moves only start after 21:00".
• The Asian session is not dead. 08:00–15:00 averages 6.37, higher than London's early window at 4.78 and close to 80% of the overlap. Hour by hour, 09:00–10:00 averages 8.51 USD, the third-highest hour of the day (behind 20:00–21:00 at 8.64 and 21:00–22:00 at 8.55), because after the Tokyo 09:00 open (08:00 Beijing) Japanese and Asian physical buying only really connects in the second hour.

"M15 range" measures how far price travels in 15 minutes, not which way. It sets how much room an order has; it does not tell you whether to be long or short.

4. Measurement two: how far each window travelled today (2026-10-23)

Today was a one-way down day: the 00:00 Beijing open was 4,052.74 and it closed at 4,003.29 (23:55), with a high of 4,053.17 and a low of 3,975.74 — a range of 77.43 USD and a decline of 1.22%. Sliced by window:

The 5-minute and 1-minute figures come from /api/candles?tf=300 and tf=60 (last bar 15:55 UTC = 23:55 Beijing); today's 1-minute series only reaches back to 07:19 Beijing, so 00:00–07:19 was rebuilt in 1-minute buckets from the site quotes table — the two methods cross-check to within 0.005 USD over the overlapping period (08:00–24:00).

Two points have to be made or the table will be misread:

• Today's largest window was indeed the overlap (57.67 USD), and it contained the low of the day: 22:00–23:00 Beijing fell from 4,022.90 to 3,975.74, 47.16 USD in a single hour.
• But the second-largest was in Asia (45.43 USD), from a 38.82 USD one-hour slide at 09:00–10:00. So "the overlap is thickest on average" is a probabilistic edge, not a promise that the biggest window is always in the evening. Watch only the evening and you would have missed the whole 09:00 move today.

Look again at 04:00–08:00: the range is only 11.38, the smallest of the day, and there is not a single quote between 05:00 and 06:00 (detailed in section 7). That window is not quiet — it is untradeable.

5. Measurement three: spreads — the answer is not the textbook one

The textbook says the overlap has the tightest spread, because it has the most market makers and the most competition. That holds for ECN quotes, but not for XAUAD's current feed (Swissquote), because that feed quotes a fixed spread.

The implication for "what time to trade" is direct: with a fixed spread you cannot save on the spread by changing hours, you can only earn on the range. The entire point of choosing a window is to have that 0.662 USD sit inside a larger price move rather than a smaller one. If the feed ever switches back to a floating spread, the overlap's spread advantage returns; the test is simple — sample the spread at the open, at midday and before the close, and if all three numbers match it is fixed, if they differ it is floating.

How to read a spread, how to tell fixed from floating and how to measure it yourself with 200 samples is broken down in the next article, gold spread comparison: how far apart are the brokers, and how to measure your own (the cost structure is also worked through in gold short-term tactics: duration, size and cost). For now one conclusion is enough: your cost is a constant, your opportunity is a variable.

6. What duration fits each of the four windows

Only range and spread together give a judgement you can trade on. Everything in the table below is from today's sampling (2026-10-23 23:58), where "share" = 0.662 ÷ the median range of that window:

Four readings:

• A 60-second trade only makes sense 20:00–24:00. The spread eats 26.8%, leaving about seventy per cent of the room; put the same trade at 04:00–08:00 and the spread eats 68.6%, leaving thirty per cent, so the win rate would have to approach seventy per cent to be positive — a bar nobody clears consistently.
• 5-minute trades are spread far more evenly (12.3%–33.7%), so whether you switch windows matters about half as much for a 5-minute trade as for a 60-second one. The longer the duration, the less the window matters.
• 15:00–20:00 is the trap that looks busy and is not worth it. A 21.50 window range is not small, but the move grinds: the 1-minute median is only 1.410, so a 60-second trade gives up 47% to the spread, and being half a beat late on direction or rhythm grinds away the rest.
• Data days need special handling. Do not open a trade in the 3–5 minutes around the 20:30 and 21:30 releases (NFP, CPI, FOMC): spreads, slippage and gaps all stop obeying the usual rules then. The mechanisms are in the Fed, FOMC and gold and CPI and gold.

To add a directional filter on top, stack on the measured correlation of the dollar index with real yields: the window decides whether the odds are good enough, macro decides which side you are on, and neither substitutes for the other.

7. Weekend close and maintenance: the exact times you can verify

This section comes entirely from API fields and quote records, not from "generally speaking". Three facts:

Put together, the practical meaning matters far more than the words "closed at the weekend":

• The platform closes earlier and opens later than the market. Market data stops at 05:00 Saturday Beijing and resumes at 06:10 Monday; the platform stops at 04:00 Saturday and resumes at 08:00 Monday. In those two hours, 06:10–08:00 Monday, the market is already moving and you cannot get an order in — any gap there is yours to absorb, and a stop-loss cannot protect you.
• The daily 04:00–08:00 maintenance covers the 04:59–06:00 quote gap. There are no quotes in that period, so orders do not fill and stops do not trigger (there is no price to trigger them). Treating "set a stop and go to bed" as a seatbelt fails in those hours.
• There is nothing to trade on Sunday evening. The real first window of a new week is the Monday 08:00 open, and by then you have already missed 06:10–08:00. Anyone holding over the weekend hands an unmanageable exposure to those two hours.

8. Four common mistakes

• Treating "London opens at 15:00" as a constant. After 25 October it is 16:00, and after 1 November New York becomes 21:00. You get it wrong once in March and once in October–November every year, and each time you are an hour out of the market.
• Assuming the overlap spread must be tighter. That does not hold on a fixed-spread feed: all 24 hourly medians measured here sit between 0.662 and 0.700, with no systematic difference between the small hours and the overlap. The spread advantage only exists on floating-spread accounts and ECN quotes.
• Treating the Asian session as dead. The 15-day mean is 6.37 USD, 09:00–10:00 alone averages 8.51, and 09:00–10:00 today covered 38.82. It is not short of moves; the moves are brief and need a longer duration to capture.
• Ignoring the 04:59–06:00 quote gap. Orders, stops and "the opening price" are all void in that period, and that is also where Monday's gap happens.

A word on how this article divides the work with the existing how to calculate gold volatility: ATR and session range measured: that piece covers "roughly how many dollars a 60-second move is in each session and what duration to pair with it", an order-of-magnitude read; this one covers "an hour-by-hour timetable, how the daylight-saving switch works, the exact close and maintenance times, and a reproducible measurement series". The two use different bases, so the numbers differ — treat the measurements here as authoritative: first use that article to pick your cycle, then this one to pick the moment you press the order button. How much each of these four windows fell this month and how the range compressed is broken down day by day and hour by hour in gold down 8.2% in a month: a data review from 4,331 to 3,975.

9. FAQ

Q. What time does gold open? How many hours a day does it trade?
A. Spot gold is quoted almost continuously from the Monday open to the Friday close, so "what time does gold open" has four answers. In Beijing time: Sydney 05:00, Tokyo 08:00, London 16:00 (winter-time basis from 2026-10-25; 15:00 under summer time), New York 20:00 (21:00 from 2026-11-01). The high-liquidity window is the London-New York overlap, 20:00–01:00, with a core three hours of 21:00–24:00. Note also that no orders can be placed during the daily 04:00–08:00 maintenance, so there are about 96 tradeable hours a week.

Q. What is the best time to trade gold?
A. The core three hours are 21:00–24:00 Beijing, and the whole overlap is 20:00–01:00 (2026-10-25 to 11-01 basis). Measured on XAUAD's own quote series over 15 complete trading days in the past three weeks, the 15-minute mean range in 20:00–24:00 is 7.82 USD, 2.5× the quiet 04:00–08:00 window (3.13 USD); 21:00–24:00 alone is 7.54 USD. The larger the range, the smaller the share the spread takes.

Q. Can you really not trade the Asian session (Beijing daytime)?
A. You can, but extend the duration. Measured on 2026-10-23, the 1-minute median range in 08:00–15:00 was 1.672 USD, so a 0.662 USD spread takes 39.6% and 60-second trades are expensive; the 5-minute median range in the same window was 3.993 USD, where the spread takes 16.6%, so 5 minutes or more is reasonable. 09:00–10:00 is often the strongest Asian hour (15-day mean 8.51 USD).

Q. How do I change the timetable after the daylight-saving switch?
A. The EU moved to winter time at 2026-10-25 01:00 UTC, so London goes from 15:00–24:00 to 16:00–01:00 Beijing; the US moves at 2026-11-01 02:00 EDT, so New York goes from 20:00–05:00 to 21:00–06:00. After the switch the London-New York overlap is 20:00–01:00 (widest during the seven days from 10-25 to 11-01) and narrows to 21:00–01:00 after 11-01. Tokyo is unchanged all year (Japan has no summer time) and Sydney moves an hour earlier when Australia entered summer time on 4 October. The Beijing-time window for US data shifts an hour later too (20:30 → 21:30).

Q. When does it close for the weekend, and when can I trade on Monday?
A. XAUAD closes per its API configuration: Friday 20:00 UTC to Monday 00:00 UTC, i.e. stop 04:00 Saturday and resume 08:00 Monday Beijing; daily maintenance is 04:00–08:00 Beijing. Our quote records show the feed's last Friday tick at 21:00 UTC (05:00 Saturday Beijing) and its resumption at 22:10 UTC Sunday (06:10 Monday Beijing), so between 06:10 and 08:00 on Monday the market is moving but the platform is not yet open.