After the shock: what gold did after its biggest days
Every COMEX gold day up or down 2% or more since 2000, and what came next: the median and the usual range of the next 1, 5 and 20 days, and how often the next day kept going. Real daily closes, sample size in every cell.
What came after big up and down days
| Day type | Days | Next 1 day | Next 5 days | Next 20 days | Next day same direction |
|---|---|---|---|---|---|
| Up 2% or more | 227 | −0.07% −0.70% to +0.67% · up 48% · n 227 | +0.23% −2.03% to +2.23% · up 52% · n 227 | −0.02% −3.09% to +4.09% · up 50% · n 226 | 49% (n 226) |
| Down 2% or more | 250 | +0.11% −0.48% to +0.94% · up 54% · n 249 | +0.21% −1.76% to +2.35% · up 52% · n 248 | +0.65% −2.54% to +4.06% · up 57% · n 245 | 46% (n 248) |
| Up 3% or more | 62 | −0.06% −0.62% to +0.77% · up 48% · n 62 | −0.27% −2.11% to +2.68% · up 50% · n 62 | −0.20% −2.13% to +3.77% · up 48% · n 61 | 48% (n 62) |
| Down 3% or more | 83 | +0.09% −0.91% to +0.92% · up 52% · n 82 | +0.40% −1.48% to +2.97% · up 54% · n 81 | +0.52% −3.19% to +4.22% · up 54% · n 80 | 48% (n 82) |
| Every trading day (baseline) | 6,582 | +0.05% −0.49% to +0.63% · up 52% · n 6,581 | +0.29% −1.18% to +1.70% · up 55% · n 6,577 | +0.74% −2.03% to +3.90% · up 56% · n 6,562 | — |
Median change from the big day’s close, with the 25th to 75th percentile range, the share of cases that were higher, and n (the number of days with a known result; the newest days do not have a 20-day result yet). Cells with fewer than 10 days are hidden. 6,582 trading days from 31 Aug 2000 to 20 Nov 2026. Gold: COMEX gold futures (GC=F) daily closes via Yahoo Finance, close only.
By decade (2% days)
| Decade | Up 2%+ days | Next 5 days after an up day | Next day still up | Down 2%+ days | Next 5 days after a down day | Next day still down |
|---|---|---|---|---|---|---|
| 2000–2009 | 94 | +0.39% −1.91% to +2.42% · up 53% · n 94 | 51% (n 94) | 104 | −0.09% −2.25% to +2.33% · up 48% · n 104 | 50% (n 103) |
| 2010–2019 | 59 | −0.08% −1.44% to +1.08% · up 49% · n 59 | 43% (n 58) | 68 | +0.35% −1.18% to +2.11% · up 57% · n 68 | 41% (n 68) |
| 2020 to now | 74 | +0.29% −2.95% to +3.33% · up 54% · n 74 | 50% (n 74) | 78 | +0.21% −1.76% to +2.68% · up 53% · n 76 | 45% (n 77) |
Same method, split by the year of the big day. A decade with few big days has hidden cells.
The last five 2% days
| Date | Move that day | Next day | Next 5 days | News |
|---|---|---|---|---|
| 20 Nov 2026 | −4.06% | not yet | not yet | news that day |
| 18 Nov 2026 | −3.16% | −1.51% | not yet | news that day |
| 12 Nov 2026 | −2.88% | +0.03% | −5.14% | news that day |
| 10 Nov 2026 | −2.05% | −0.74% | −4.12% | news that day |
| 4 Nov 2026 | +3.37% | +0.99% | −1.78% | news that day |
From the same data file; “not yet” means the later closes are not in the data yet. News links go to our daily gold news page for that date where one exists.
The short answer
Since 31 Aug 2000 gold had 227 days up 2% or more and 250 days down 2% or more out of 6,582 trading days. The next day went the same way 49% of the time after a big up day (n = 226) and 46% after a big down day (n = 248), close to a coin flip either way. Twenty trading days later the median change was −0.02% after big up days and +0.65% after big down days, against +0.74% for every day in the sample. The table above gives the full picture: the middle result, the usual spread of results and the sample size for every cell.
How to read the tables
- Big day: a COMEX gold close at least 2% (or 3%) above or below the previous trading day’s close.
- Next 1, 5 and 20 days: the change from the big day’s close to the close 1, 5 and 20 trading days later. The bold figure is the median (the middle case); the small range below it is the 25th to 75th percentile, so half of all cases fell inside it.
- Next day same direction: after an up day, the share of next days that closed higher; after a down day, the share that closed lower. Unchanged next days are left out.
- Baseline: the same numbers for every trading day, so you can see whether big days were followed by anything unusual at all.
Why the days are not independent
Big days cluster: 2008 (61), 2026 (50), 2006 (43) alone hold 32% of the 2% days. Big moves arrive in bursts during crises and sharp trend changes, often several in the same few weeks. That means many of the 5-day and 20-day windows overlap and share the same later days, so the real amount of independent evidence is much smaller than the count suggests. Treat each figure as a description of a handful of market episodes, not of hundreds of separate experiments.
Method and limitations
- Data: COMEX gold futures (GC=F) daily closes via Yahoo Finance, 31 Aug 2000 to 20 Nov 2026 (file refreshed 20 Nov 2026). Only the close is used, because open, high and low are missing on many early rows. Futures trade a few dollars from spot XAUUSD; daily percentage changes are almost the same.
- Trading days: rows of the data file. A change across a gap of more than 7 calendar days is not counted.
- Thresholds: 2% and 3% are fixed round numbers chosen before looking at results, not tuned. Gold’s volatility has changed a lot over 26 years, so 2% was rarer in calm years than in 2026.
- Small cells: any cell with fewer than 10 days is hidden.
Frequently asked questions
After a big gold rally, does gold keep going the next day?
In COMEX closes since 2000 the next day went the same way only about as often as not; the exact share and its sample size are in the table on this page. A big day says little about the next day’s direction.
Does gold bounce after a big drop?
Sometimes, but not reliably. The table shows the median and the usual range of the next 1, 5 and 20 days after every 2% and 3% down day, next to the same figures for all days.
Why do the numbers say n?
n is the number of big days behind each figure. Cells with fewer than 10 days are hidden, and because big days cluster in crises, even larger counts hold less independent evidence than they appear to.
Is this a trading signal?
No. It is a description of past daily closes. It gives no entry, stop or target and does not predict the next move.
Educational study only. Not financial advice and not a signal. Past price behaviour does not predict future prices. See our risk disclaimer.