MEASURED · 200 DAILY BREAKS
What happens to gold’s spread at the 5 pm ET daily break?
It steps up after the break rather than blowing out. On a year of one-minute gold (XAUUSD) bars from an MT5 feed, quotes stopped at 4:57 p.m. ET (3:57 p.m. CT) and came back at 6:00 to 6:02 p.m. ET (5:00 to 5:02 p.m. CT). From February 2026, the narrowest spread in the first minute back was a median $0.40, against $0.28 during the day, and $0.50 or more on 39.8% of evenings. By 6:30 p.m. ET it was back to normal.
Two caveats change how to read that. MT5 bars keep each minute’s narrowest spread, so the widest quote at the reopen was at least this wide and may have been wider. And before February 2026 this account’s spread did not move at the break at all.
How it was measured
We took 356,223 one-minute XAUUSDm bars from an Exness MetaTrader 5 feed, August 1, 2025 to August 4, 2026, and lined up every weekday break by the clock: 4:00 to 4:57 p.m. ET before it, 6:00 to 7:59 p.m. ET after it. That gave 200 Monday-to-Thursday breaks and 53 Sunday opens. Holiday early closes are kept out of the averages.
MT5 stores one spread value per bar. MetaQuotes’ own MQL5 book lists it as the “minimum bar spread”. We checked that on this feed: in all 70,985 complete five-minute bars, the spread equals the smallest of its five one-minute spreads. So every figure below is the narrowest quote in its minute. A single wider tick inside the minute does not show.
The account’s spread changed character partway through the year, as in our economic-release studies. Through January 2026, 99.8% of all minutes recorded exactly $0.16. From February 1, 2026, only 3.1% did. The two periods are reported separately.
The spread minute by minute, February to August 2026
| Time, ET (CT) | Evenings | Median | Avg vs day | $0.50+ | Widest |
|---|---|---|---|---|---|
| 4:00–4:29 p.m. (3:00–3:29) | 103 | $0.28 | 1.03x | 0.2% | $0.50 |
| 4:30–4:49 p.m. (3:30–3:49) | 103 | $0.28 | 1.04x | 0.1% | $0.50 |
| 4:50–4:57 p.m. (3:50–3:57) | 103 | $0.28 | 1.05x | 0.0% | $0.40 |
| 4:58–5:59 p.m. (3:58–4:59) | no quotes | ||||
| 6:00–6:01 p.m. (5:00–5:01) | 94 | $0.40 | 1.54x | 32.5% | $0.66 |
| 6:02–6:04 p.m. (5:02–5:04) | 101 | $0.36 | 1.37x | 22.2% | $0.77 |
| 6:05–6:09 p.m. (5:05–5:09) | 102 | $0.36 | 1.30x | 17.5% | $0.65 |
| 6:10–6:29 p.m. (5:10–5:29) | 103 | $0.28 | 1.15x | 6.0% | $0.60 |
| 6:30–6:59 p.m. (5:30–5:59) | 103 | $0.28 | 1.05x | 0.3% | $0.50 |
| 7:00–7:29 p.m. (6:00–6:29) | 103 | $0.28 | 1.03x | 0.0% | $0.36 |
| 7:30–7:59 p.m. (6:30–6:59) | 103 | $0.28 | 1.02x | 0.0% | $0.40 |
103 weekday breaks. Spread is the narrowest quote in each minute. Avg vs day: the window’s average spread divided by the 8 a.m. to 4 p.m. ET average ($0.27). Evenings: breaks with at least one bar in the window. In GMT, 6 p.m. ET is 22:00 in US summer and 23:00 in winter.
The jump is at the reopen, and it is short. In the first two minutes back the spread averaged 1.54x the daytime level, and 32.5% of those minutes were at $0.50 or wider. From 6:10 to 6:29 p.m. it averaged 1.15x, and from 6:30 p.m. ET it stayed within 5% of the daytime average. The single widest minute in any reopen hour was $0.77.
Small as it is, it is concentrated. The weekday 6 p.m. hour is 3.4% of the period’s minutes but holds 60.6% of its minutes at $0.50 or wider: one minute in 18 in that hour, against about one in 854 between 8 a.m. and 4 p.m. ET. Before the break nothing much happens: the 4:50 to 4:57 p.m. average was 1.05x the daytime level, and no minute reached $0.50.
The first quote after the break
| Reopen | Breaks | Median | $0.50+ | Widest |
|---|---|---|---|---|
| Weekday reopenFeb–Aug 2026 | 103 | $0.40 | 39.8% | $0.77 |
| Sunday openFeb–Aug 2026 | 27 | $0.40 | 44.4% | $0.60 |
| Weekday reopenAug 2025–Jan 2026 | 97 | $0.16 | 0.0% | $0.24 |
| Sunday openAug 2025–Jan 2026 | 26 | $0.16 | 0.0% | $0.40 |
Daytime median: $0.28 from February 2026, $0.16 before.
From February 2026 the first bar back was wider than the daytime median on 87.4% of weekday evenings. The Sunday open looked much the same, slightly more often at $0.50 or wider, on only 27 opens. Before February the break left no mark: the first bar read $0.16 on 99.0% of 97 evenings, and nothing in any reopen hour went above $0.24. Whether the reopen was calm then or the feed simply recorded a flat spread, bar data cannot say.
When the quotes stop and start
The break on this feed is longer than an hour. Before 250 of 259 evening reopens, the last bar was 4:57 p.m. ET (3:57 p.m. CT; 20:57 GMT in US summer, 21:57 in winter). There was no bar at 4:58 or 4:59. A bot set to close at 4:59 p.m. finds nothing to trade against.
The reopen moved during the year. The first bar came at 6:02 p.m. ET on every weekday evening in August 2025, September 2025 and October 2025, and at 6:00 p.m. on every one from June 2026. Over the whole year it was 6:00, 6:01 or 6:02 p.m. ET on 193 of 200 evenings, and later than 6:02 on the other 7. Sunday opens ran later: 6:01 to 6:09 p.m. ET, with 6:05 on 13 of 53.
US holidays moved the break forward. On Labor Day, Thanksgiving, Martin Luther King Jr. Day, Presidents' Day and Memorial Day, the last bar came between 2:14 p.m. and 2:29 p.m. ET and the market reopened at 6 p.m. as usual, so the pause ran 3.5 to 3.8 hours instead of one.
Price across the break moves more than the spread
For a position held through the break, the spread step is the smaller problem. From the last close at 4:57 p.m. to the first price after 6 p.m. ET, gold jumped a median $2.27 over the 200 weekday breaks. The jump was more than $5 on 24.0% of evenings and more than $10 on 11.0%. The largest, on April 7, 2026, was $34.43.
That is less than gold moves in an open hour. Over the last trading hour on the same days, 3:57 to 4:57 p.m. ET, the median move was $3.50. The difference is that nothing can be done during the break. A stop inside the jump fills at the first price after the reopen, at whatever spread the first minute carries. The first minute back also ranged a median $1.75, against $1.06 for an ordinary minute from 4:00 to 4:56 p.m. ET.
What to do with this if you run a bot
Finish before 4:57 p.m. ET. If the bot must be flat over the break, close by about 4:55 p.m. ET (3:55 p.m. CT). On the US holidays above, close before 2:14 p.m. ET. Exness also takes swap at 5:00 p.m. ET, inside the break, from any position still open on an account without swap-free status, and Exness’s swap-free rules set out who has that status and how it can be lost.
Skip the first half hour. From 6:00 p.m. ET (5:00 p.m. CT) until 6:30 p.m. ET the spread ran above the daytime level, and that is where the wide minutes clustered. Holding off new entries until then costs half an hour a day.
Set the spread filter from your own account. On this account a $1.00 filter would never have fired at the reopen. A $0.50 cap, under twice the daytime median, would have blocked the first minute back on 39.8% of evenings. Whatever EA you run, the Tech Kick Trading Bot included, check what its spread filter compares against and which clock its session times use. This feed’s server clock is UTC, so 6 p.m. ET is 22:00 on the server in US summer and 23:00 in winter.
Size held positions for the jump, not the spread. A short’s stop is triggered by the ask, so the reopen spread brings it closer: from February 2026 the first quote’s median spread sat $0.12 above the daytime median. The price jump across the break was more than $5 on 24.0% of evenings. Our stop study by session shows how often ordinary movement reaches a stop of a given size within the hour, and the weekend version of this risk is in gold’s weekend gap.
Do not trust bar spreads in a backtest at the reopen. They are minimums. For an honest cost at 6 p.m. ET, test on real ticks. 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
The bar clock is UTC: the last bar before the break sat at 20:57 UTC in US summer and 21:57 in winter, 4:57 p.m. ET both times. ET, CT and GMT come from the IANA time zone database. One broker, one account type, one symbol. Peak spreads need tick data, which this archive does not hold, and the bars cannot say why the account’s spread changed in February 2026. Our hourly spread study measured the same feed on five-minute bars, where each spread value is the minimum of five one-minute values. Every figure here comes from one script, research-gold-spread-daily-break.py, run on the same bar archive as our other studies.
Common questions
What time does gold stop trading for the daily break?
On the MT5 feed we measured, the last one-minute bar before 250 of 259 evening reopens was 4:57 p.m. ET (3:57 p.m. CT; 20:57 GMT in US summer, 21:57 in winter), and quotes came back at 6:00 to 6:02 p.m. ET on 193 of 200 weekday evenings. Other brokers may differ by a minute or two, so read the last bar on your own chart.
Does the spread widen before the 5 p.m. close?
Hardly. From February to August 2026 the average spread from 4:50 to 4:57 p.m. ET was 1.05x the daytime level, against 1.03x from 4:00 to 4:29, and no minute in that last stretch reached $0.50. The wide quotes come after the break, not before it.
Is a $1.00 spread filter enough to keep a bot out of the reopen?
Not on this account. The narrowest quote at the reopen never reached $1.00 in a year of bar data (widest $0.77), so a $1.00 cap would never have fired. A cap at $0.50 would have blocked the first minute back on 39.8% of evenings from February 2026. Set the filter from your own account's daytime spread, not from a round number.
Is the Sunday open worse than a weekday reopen?
A little wider and a little later, on a small sample. From February 2026 the first Sunday-evening quote was $0.50 or wider on 44.4% of 27 opens, against 39.8% of weekday reopens. Over the whole year, Sunday quotes started between 6:01 and 6:09 p.m. ET. The bigger Sunday risk is the price gap over the weekend, which we measured separately.