Quick answer: No, not in any way we could measure. We lined up Bitcoin's daily closes from August 2017 to October 2026 against 226 exactly computed full and new moons. In the 7 days after a full moon, BTC averaged +0.2%. After a new moon it averaged +1.4%, and after any day +1.1%. None of the gaps, at 3, 7 or 14 days, came close to statistical significance (lowest p = 0.22).
Ask around crypto long enough and someone will share the secret: buy the new moon, sell the full moon. The theory comes from old trading folklore. It says markets turn euphoric and top out around the full moon, then bottom near the new moon.
It sounds poetic. But does Bitcoin actually obey the Moon? We ran the numbers in June. Then we re-ran them with exact astronomical moon times, three more months of data and proper significance tests. Here's everything, including where our first version was wrong.
How we tested
Everything below can be reproduced from public data.
- Prices: Binance BTC/USDT daily candles from August 17, 2017 (the first day in Binance's BTC/USDT daily history) to October 8, 2026. That's 3,340 consecutive days with no gaps. Each daily close is the last price at 23:59:59 UTC.
- Moon phases: the exact time of every new and full moon, computed with the method in Jean Meeus' Astronomical Algorithms (chapter 49) and converted to UTC. We cross-checked every date against the PyEphem astronomy library, and all of them agreed to within one minute. That gives 113 full moons and 113 new moons, from the new moon of August 21, 2017 to the full moon of September 26, 2026.
- Entry: the Bitcoin close on the UTC calendar day of the full or new moon.
- Horizons: returns 3, 7 and 14 calendar days later, close to close, with no fees.
- Baseline: the same 3, 7 and 14-day returns starting from every day in the dataset.
- Significance: we drew 20,000 sets of 113 random days and checked how often random days landed at least as far from typical as the moon days did. That share is the p-value. To compare full moons with new moons directly, we shuffled the "full" and "new" labels 20,000 times. We also ran a Welch t-test and bootstrapped a 95% range for the gap.
- The original test, extended: we also re-ran our June "half-cycle" test, which measures Bitcoin's move from each new moon to the next full moon and from each full moon to the next new moon.
What Bitcoin did after full and new moons
| Bought at the close on⦠| Held for | Average | Median | Share up | Different from random days? |
|---|---|---|---|---|---|
| Full moon | 3 days | β0.23% | +0.13% | 54.0% | No (p = 0.22) |
| New moon | 3 days | +0.33% | +0.02% | 50.4% | No (p = 0.82) |
| Any day | 3 days | +0.45% | +0.29% | 53.2% | baseline |
| Full moon | 7 days | +0.21% | +0.12% | 50.4% | No (p = 0.32) |
| New moon | 7 days | +1.35% | +0.95% | 56.6% | No (p = 0.75) |
| Any day | 7 days | +1.08% | +0.53% | 53.2% | baseline |
| Full moon | 14 days | +1.87% | +0.13% | 51.8% | No (p = 0.80) |
| New moon | 14 days | +2.12% | +1.24% | 58.4% | No (p = 0.94) |
| Any day | 14 days | +2.20% | +0.89% | 54.0% | baseline |
The p-values in the table are for the average. The same test on the medians and the share of up moves gave p-values of 0.39 or higher in every row.
Comparing the two phases head to head:
| Held for | Full moon minus new moon (average) | 95% range | p (label shuffle) |
|---|---|---|---|
| 3 days | β0.56 points | β2.21 to +1.02 | 0.50 |
| 7 days | β1.15 points | β3.50 to +1.22 | 0.34 |
| 14 days | β0.25 points | β3.94 to +3.46 | 0.89 |
Every 95% range crosses zero. On this data, "after a full moon" and "after a new moon" are statistically the same thing.
The half-moon runs, re-tested
Our original post measured each half of the lunar cycle. Here's that test again with exact moon dates, through September 2026:
| Half of the lunar cycle | Runs | Average | Median | BTC rose | Compounded, holding only this half |
|---|---|---|---|---|---|
| πβπ Waxing (new moon to full moon) | 113 | +2.80% | +1.34% | 57.5% | +696% |
| πβπ Waning (full moon to new moon) | 112 | +1.88% | +0.05% | 50.9% | +164% |
| Any 15-day window (baseline) | 3,325 | +2.37% | +1.01% | 54.0% | n/a |
| Buy & hold, Aug 21, 2017 to Sep 26, 2026 | n/a | n/a | n/a | n/a | +2,002% |
The waxing half did a bit better, which is the direction the folklore predicts. But the gap of 0.92 points per half-cycle is well inside the noise (p = 0.64). The difference in how often each half rose (65 of 113 vs 57 of 112) is noise too (p = 0.35).
Could you actually trade it?
On paper, "buy the new moon, sell the full moon" turned $1 into about $7.96 (+696%). Holding only the opposite half made +164%. Two things undercut that headline:
- Buy & hold crushed both. Over the same span, simply holding returned +2,002%. Each moon strategy sits in cash for half of Bitcoin's climb. That was true in June and it's still true.
- The gap is fragile. Move every moon date one day earlier and the two halves come out at +356% and +351%, almost identical. Move them one day later and you get +508% vs +244%. A real effect shouldn't vanish because the clock moved by a day. Compounding 113 noisy returns magnifies small random differences, and a few huge runs carry much of the result. The single best waxing half was +49.8% (December 14β30, 2020), and the best waning half was +68.9% (December 3β18, 2017).
Is any day of the lunar cycle special?
We tried buying on each of the 30 days of the lunar cycle in turn, counting from the new moon, and holding for a week. The 7-day averages ranged from +0.27% (11 days after the new moon) to +1.60% (22 days after). The new-moon day ranked 10th of 30. The full moon fell 14 or 15 days after the new moon in 95 of 113 cycles, and those two days ranked 12th and 28th.
Every one of those averages sits inside the range that 113 random days produce on their own (β0.62% to +2.79% for a 7-day hold). Nothing about the moon days stands out.
Two more cuts, both showing the same thing:
- By year. On the 7-day measure, full moons "beat" new moons in 6 of 10 calendar years (2017 and 2026 are partial). New moons won 2018, 2019, 2020 and 2022. The "winner" flips around, which is what noise looks like.
- By market regime. When BTC was above its 200-day average, the week after a full moon averaged +1.72% and the week after a new moon +1.86%, almost the same. Below the 200-day average, full moons averaged β1.06% and new moons +0.89% (52 of each). That looks like a pattern, but it's one subgroup out of many we could have sliced, and it isn't significant (p = 0.22).
Why people believe it, and what the data says
The Moon is always close by. A full or new moon arrives about every 14.8 days, so every market top and bottom is within about a week of one. Look back at any big turn and you'll find a moon nearby.
Vivid examples stick. The worst week after any full moon in our data started on March 9, 2020: Bitcoin fell 36.6% in the next seven days during the COVID crash. Moments like that get screenshotted and remembered. The 50.4% of full-moon weeks that went up don't.
Bitcoin's drift hides everything. BTC rose about 1,800% over this period, so almost any window you pick averages a gain. Both halves of the lunar cycle ride that same tide. Our first write-up showed this too.
What this test can and can't say. With Bitcoin's volatility, 113 events can't rule out a small lunar effect. A real edge smaller than about 1.7 points per week would hide inside the random range. What the data rules out is a large, reliable one, the kind that would make "sell the full moon" a usable rule.
Three caveats to keep the numbers honest:
- Multiple comparisons. The forward-return tests alone add up to 21 comparisons (six moon rows times three statistics, plus three head-to-head checks), before the half-cycle, year and regime slices. Run that many and roughly one in twenty should look "significant" by luck alone. None of ours came under 0.2.
- Overlapping windows. The any-day baseline uses overlapping windows, which is why we compared against random sets of days instead of treating every day as independent.
- Regimes dominate. The 2017 and 2020β21 bull runs and the 2018 and 2022 bear markets drive most of the big moves. They swamp anything the calendar might add.
What changed since our first test
Our June version used an average 29.53-day lunar cycle. That method reproduces exactly: 109 full moons, 110 new moons, a +2.5% average waxing run, a +2.1% waning run, and +413% vs +220% for the two half-cycle strategies. The trouble is that real moon phases don't keep to the average: those dates sat up to 18 hours from the true moments, and 60 of the 219 landed on the wrong calendar day. This version uses exact times and runs through October 2026.
We also got one thing backwards in the first write-up. We said the folklore was "backwards" because Bitcoin tended to rise into the full moon. But rising into the full moon and topping there is exactly what the folklore predicts. The accurate summary is that the data leans very slightly the folklore's way, by an amount indistinguishable from chance, and the lean disappears when you move the dates by a day.
Where to go from here
So, is the Moon a secret Bitcoin indicator? No. That's exactly how we treat the cosmic stuff here: a fun lens, not an edge. You can watch today's moon phase next to the Bitcoin price. If you enjoy the esoteric angle, our Astrology lab maps eclipses, zodiac seasons and lunar phases against real price history.
If you want to test a lunar idea (or any idea) without risking money, the $10,000 paper-trading demo is free with no signup, on real Binance market data. If you'd rather build an edge that doesn't depend on the sky, start with the free TA course.
Got a trade the stars supposedly blessed, or cursed? Share your story with the community on the blog, or read more from the WenMoonLambo blog.
Frequently asked questions
Does the full moon affect Bitcoin price? Not in a way we could measure. Across 113 full moons from August 2017 to September 2026, Bitcoin averaged +0.21% over the next 7 days, against +1.08% for any day, and 50.4% of those weeks were up. Random sets of days differ by that much about a third of the time (p = 0.32).
Is there a lunar cycle in crypto? We found none in Bitcoin. Buying on each of the 30 days of the lunar cycle and holding a week gave averages between +0.27% and +1.60%. That's within what random sets of days produce, and the new-moon and full-moon days ranked mid-pack. We tested BTC only, not altcoins.
Does "buy the new moon, sell the full moon" work? Holding Bitcoin only from each new moon to the next full moon compounded +696% from 2017 to 2026, against +164% for the opposite half. But buy & hold made +2,002%, the per-cycle gap wasn't significant (p = 0.64), and shifting the dates one day earlier shrank it to +356% vs +351%.
Is it better to buy Bitcoin on a new moon? Over 113 new moons, buying at the close and holding 7 days returned +1.35% on average, against +1.08% for any day. That gap is well within chance (p = 0.75), and buying one day earlier cut it to +1.03%.
When is the next full moon? As of October 9, 2026, the next new moon is October 10, 2026 at 15:50 UTC, and the next full moon is October 26, 2026 at 04:11 UTC. After that come the new moon of November 9 (07:02 UTC) and the full moon of November 24 (14:53 UTC). The Bitcoin moon phase page shows the current phase.
The bottom line
The Moon is a beautiful clock and a useless trading signal. Across 226 full and new moons and three holding periods, Bitcoin after a moon phase behaved like Bitcoin after any other day. The only lunar "edges" we found disappeared when we nudged the dates by a single day. Enjoy the sky; size your trades on something else.
Data: Bitcoin daily closes (BTC/USDT) from Binance, August 17, 2017 β October 8, 2026 (3,340 days). New and full moon times computed with Meeus' Astronomical Algorithms (ch. 49) and cross-checked against PyEphem; 113 full and 113 new moons. Returns are close-to-close and ignore fees. Random-date and label-shuffle tests used 20,000 draws each. Past performance is not predictive. Entertainment, not financial advice. WenMoonLambo is a paper-trading platform: all trading happens with play money on real market data.