MEASURED · 17 AUGUST 2026
Gold gaps over the weekend. We measured 52 of them.
Median gap $7.60. Nearly half exceed $10. The largest in a year was $97.98. None of that would matter much, except for one thing: a stop loss does not work across a gap.
The measurement
Spot gold closes Friday evening and reopens Sunday evening. It does not stop reacting to news in between — it simply cannot reprice until the market returns, and then it opens wherever it opens.
Across one year of XAUUSD from a live MetaTrader 5 feed — 52 weekend breaks:
| Median gap | $7.60 |
|---|---|
| Share exceeding $10 | 48% |
| Largest gap | $97.98 |
| Weekend breaks measured | 52 |
Why that number is worse than it looks
A stop loss is an instruction to close at market once a level is touched. It is not a guarantee of price. Inside a normal session that distinction is small — there is liquidity on the other side and you fill near your level.
Across a gap there is no other side. Price simply is not printed between Friday's close and Sunday's open, so a stop sitting in that range is not touched — it is jumped. You fill at the reopening price, however far away that is.
The part nobody puts together
We separately measured how often a stop is hit by ordinary intraday movement, by sampling an entry every 35 minutes across 70,546 five-minute bars. A $10 stop was hit within an hour 34.2% of the time. That makes $10 a reasonable intraday distance on gold — wide enough to sit outside most noise.
And 48% of weekend gaps exceed $10.
So a stop sized sensibly for intraday conditions is jumped by the weekend roughly half the time. Not hit — jumped. The two figures were measured separately and neither is surprising alone; together they say something specific about holding gold from Friday to Sunday.
What follows, and what does not
It does not mean never hold over a weekend. Gaps go both ways, and a gap in your favour is the same phenomenon. What it means is that weekend exposure is a genuinely different risk from weekday exposure, and should be sized as one rather than assumed to be covered by the stop.
For an automated system the consequence is concrete: a bot that closes before the weekend has made a decision, and a bot that holds through it has also made one — usually without anyone having chosen. Ours enforces a session window rather than running around the clock, which is a smaller claim than it sounds and is mostly about not being in the market for the two windows where execution is worst: the daily rollover and the weekend.
And if you use a prop firm account, this interacts with the drawdown rules directly. A gap that opens through your stop is realised loss you did not choose the size of, on a Sunday, against a limit that does not care why. See prop firm rules.
Where the numbers come from, and what they cannot tell you
One year of XAUUSD five-minute bars — 70,546 of them across 312 sessions, from a single raw-spread MetaTrader 5 account, August 2025 to July 2026. The hourly aggregation and the script that produces it are published under CC BY 4.0.
The honest limits. The per-gap series is not in the published CSV — the summary figures above come from the same archive but the individual 52 gaps are not something you can currently recompute from what we have released. Fifty-two observations is also one year, which characterises a period rather than an instrument, and gold ran from roughly $3,288 to $4,044 across it. Gaps are in dollars, so they are not directly comparable to a period at a very different price level.
One broker, one account type. If you have the same measurement from a different feed, the repository takes issues and pull requests, and a correction gets the same prominence as everything else.