Study · re-run every night · updated
Do CME gaps always fill? Bitcoin's weekend gaps since 2017, tested
460 CME weekends of Bitcoin since 2017: CME's Bitcoin futures stop from Friday's close to Sunday's open while Bitcoin trades on, and the weekend's move leaves a gap on the futures chart. Within 30 days 85% of the gaps of 0.5% or more were filled, and 87% of the levels just as far the other way were reached too. Which comes first? The gap's level, 42% of 319 times (a coin flip would be 50%): less often than a coin flip, so once CME opens the weekend's move tends to go on, the other way from the saying. Past patterns, not financial advice.
Filled in 30 days
85%
Bitcoin's gaps of 0.5%+
The other way
87%
a level as far the other way
Gap's level first
42%
a coin flip is 50%
Weekend of Oct 11
+1.2%
filled 16 hours after the open
§ 01 · This weekend
This weekend's gap
as of · Binance's hourly Bitcoin price to Oct 15, 2026
CME Group's Bitcoin futures stop trading from Friday 4 pm to Sunday 5 pm Chicago time, while Bitcoin itself trades on. The weekend's move leaves a gap on the futures chart, between Friday's close and Sunday's open, and the saying goes that the price always comes back to fill it. Here every CME weekend since 2017 is measured on Binance's hourly Bitcoin price: the gap's level is Friday's CME close, and beside it sits its mirror, a level just as far from Sunday's open the other way.
The newest weekend, Oct 11, 2026: Bitcoin opened at $83,529, 1.2% above Friday's CME close of $82,514. The gap's level was reached first, 16 hours after the open; its mirror, $84,557, not yet.
| Weekend | Gap | First in 7 days | Hours to fill |
|---|---|---|---|
| +1.2% | the week not over | 16 | |
| +1.7% | the gap's level | 52 | |
| +1.0% | the gap's level | 4 | |
| −0.2% | the gap's level | 1 | |
| −0.0% | the mirror | 5 | |
| +0.3% | the gap's level | 1 | |
| +1.6% | the gap's level | 2 | |
| +0.4% | the gap's level | 2 |
Gap: Sunday's open against Friday's CME closeHours to fill: from Sunday's open to the gap's level, within 7 days
Every weekend since 2017, and the newest one's levels, free as JSON for your own charts and bots: /api/v1/cmegap.json (the free API).
§ 02 · Do they fill?
Do CME gaps always fill?
as of · the gaps of 0.5% or more
the gap's level reachedits mirror reached, as far the other way
Within 30 days, 85% of Bitcoin's gaps of 0.5% or more were filled: the number the saying quotes, and it is true. But 87% of their mirrors, the levels just as far from Sunday's open the other way, were reached in the same 30 days. Bitcoin moves several percent in a month, so almost any level that close gets reached either way: a gap filling is no news by itself. When the gap's level was reached within the week, it took 18 hours in the middle case.
| Reached within | 1 day | 7 days | 30 days |
|---|---|---|---|
| Bitcoin, the gap | 45% | 74% | 85% |
| Bitcoin, its mirror | 56% 328 weekends | 77% 327 weekends | 87% 325 weekends |
| Ether, the gap | 53% | 78% | 87% |
| Ether, its mirror | 56% 235 weekends | 76% 234 weekends | 88% 233 weekends |
§ 03 · Which first
Which level comes first
as of · from Sunday's open, within 7 days
What would make a gap worth trading is its level coming first, before the mirror, more often than a coin flip. On Bitcoin's 319 gaps of 0.5% or more, the gap's level came first 42% of the time: less often than a coin flip, so once CME opens the weekend's move tends to go on, the other way from the saying. The placebo, the same measure on ordinary weekdays with the price 49 hours before as the level, came first 50% of the time.
the gap's level first (faint: z under 2)the placebounder each bar: its weekends
| Gap size | Weekends | Gap's level first | Placebo | Trade toward the gap |
|---|---|---|---|---|
| Bitcoin · 0.25-0.5% | 63 | 45% z −0.8 |
45% | −0.05% won 43% |
| Bitcoin · 0.5-1% | 76 | 39% z −2.0 |
50% | −0.17% won 39% |
| Bitcoin · 1-2% | 102 | 37% z −2.7 |
51% | −0.35% won 36% |
| Bitcoin · 2%+ | 141 | 47% z −0.8 |
49% | −0.38% won 47% |
| Bitcoin · all 0.5%+ | 319 | 42% z −3.0 |
50% | −0.32% won 42% |
| Ether · 0.25-0.5% | 39 | 40% z −1.3 |
55% | −0.07% won 44% |
| Ether · 0.5-1% | 56 | 46% z −0.5 |
45% | −0.06% won 46% |
| Ether · 1-2% | 64 | 42% z −1.2 |
50% | −0.20% won 42% |
| Ether · 2%+ | 110 | 46% z −0.8 |
52% | −0.58% won 47% |
| Ether · all 0.5%+ | 230 | 45% z −1.5 |
51% | −0.35% won 46% |
The trade toward the gap: at Sunday's open, a buy when the weekend fell and a sell when it rose, take-profit at the gap's level, stop at its mirror, else out after 7 days; its average result a trade, before trading costs and funding.
| Bitcoin, year | Weekends | Gap's level first | Trade toward the gap |
|---|---|---|---|
| 2018 | 40 | 52% | +0.28% |
| 2019 | 38 | 47% | −1.10% |
| 2020 | 40 | 35% | −0.33% |
| 2021 | 41 | 54% | +0.27% |
| 2022 | 34 | 40% | −0.45% |
| 2023 | 31 | 26% | −0.78% |
| 2024 | 33 | 36% | −0.98% |
| 2025 | 37 | 38% | −0.12% |
| 2026 | 23 | 41% | −0.15% |
Read the other way round, after the fact: the trade away from the gap (take-profit at the mirror, stop at the gap's level) made +0.32% a trade on Bitcoin before Oct 12, 2026, won 58% of 327 (a luck test of 1.7). It was found by looking at these same weekends, so it is a lead, not a finding: only its record from Oct 12, 2026 on counts, no weekend yet.
§ 04 · How it is tested
How it is tested
as of · methodology
Each weekend: the price at Friday's CME close (the close of the hour that ends at 4 pm Chicago time) and at Sunday's open (the open of the hour that starts at 5 pm), Chicago's clock by the US daylight-saving rule. The gap is the open against Friday's close; its level is Friday's close; its mirror sits as far from the open the other way. From the open on, hour by hour for 7 days, which level the price reaches first, by each hour's high and low: an hour that reaches both counts half to each (which came first inside it is unknown), and neither in 7 days counts as open. A driftless market reaches either first half the time.
Fixed before any number was read: the gaps of 0.5% or more are the test (under 0.25% an hour's range swallows both levels); the gap's level must come first more often than its mirror at a one-sided z of 1.65 or more and above half in both halves of the weekends, more often than the placebo, and the trade toward the gap must make money at a luck test of the same size in both halves. Spot prices, not the futures' own (their premium moves little over a weekend); CME's holiday closes are not modelled. Every number is re-read each night from the hours on file.
§ 05 · Revisions1 dated
Revisions
as of · every change to its headline numbers, dated
This study is re-run every night on the whole history. It was first written on Oct 15, 2026; its numbers last moved on Oct 15, 2026. Each night its headline numbers moved is listed here, newest first.
- Bitcoin's gap level first 42% · within 30 days the gap 85%, the level as far the other way 87%
§ 06 · More
More research
as of