Does Mercury retrograde move Bitcoin?

The idea is everywhere in crypto circles: Mercury turns retrograde, contracts glitch, traders misread the news and the market slides. “…during bear markets, Bitcoin crashes with each Mercury in Retrograde,” NewsBTC told its readers in 2020, and the theme keeps coming back (NewsBTC, 2022, Bitcoinist, 2024). And Bitcoin's history seems to oblige. When Mt. Gox went dark on 25 Feb 2014, Mercury was retrograde. TerraUSD lost its peg on 9 May 2022, with Mercury in its pre-shadow. And the worst day in the price record, 12 Mar 2020, came two days after Mercury turned direct.

But the same calendar holds the good days. Bitcoin first traded above $100,000 on 5 Dec 2024, with Mercury retrograde, and the fourth halving came with Mercury retrograde. Mercury spends 48% of all days between the start of a pre-shadow and the end of a post-shadow, so any list of famous days should land about half inside, crashes and milestones alike. Of these 20, 13 did: 5 of the 9 crashes and 8 of the 11 milestones.

A list of famous days is a list someone chose. The lore is a claim with dates attached, so it can be checked on all of them. Mercury's retrogrades come from the ephemeris; Bitcoin's price comes from 2,597 daily closes, 8 Sep 2019 to 17 Oct 2026.

Some retrogrades did line up with a fall: −20.1% in Pisces, 17 Feb – 10 Mar 2020. Others caught a rally: +63.1% in Aquarius, 30 Jan – 21 Feb 2021. Picking either one out is how a pattern gets made. Looking at all of them at once is how it gets tested.

Lay the retrogrades over each other, each starting from the close before its first retrograde day, and they fan out both ways. The median one ended +0.4%; the mean, pulled up by the 2021 rally, +4.3%.

Behind them is the range that 8 in 10 stretches of the same length fall in, starting on any day at all. By definition 2 in 10 ordinary stretches end outside it, so about 4.4 of 22 retrogrades would by chance. 5 did: 4 above it and 1 below.

Bitcoin rose over 12 of the 22 retrogrades and fell over 10. For comparison, of every 23-day stretch in the series, starting on any day, 54% ended higher. Retrogrades land where ordinary 23-day stretches land.

Volatility finds nothing either. By the usual measure, the standard deviation of daily returns, Bitcoin moved 2.95% a day during retrogrades against 3.26% on other days: about 10% calmer, a gap that turns up by chance 43% of the time. By the median day's move it was 1.43% against 1.36%, a touch livelier. Neither is the chaos the lore promises.

Counting ups and downs throws information away, so the sharper test uses every day. The average day inside a retrograde was +0.12%; the average day outside, +0.07%. Is that gap something, or noise?

To find out, slide the whole retrograde calendar along the prices by one day, then two, and so on through every one of the 2,596 possible shifts, and measure the gap each time. The real calendar ranks 1,955 of 2,596: a p-value of 0.75. Mercury's real dates do no better than the same windows dropped anywhere else.

How big would an effect have to be to show? Nineteen in twenty shifted calendars put the average retrograde day between −0.30% and +0.32% a day from the average other day. Over a typical 23-day retrograde that comes to −7% or +8%: a retrograde would have had to move Bitcoin about that much, every time, to stand out. A small effect could hide here; the dramatic one the lore describes could not.

One caution runs the other way. Mercury's retrogrades come round every 116 days or so, so a shift near a multiple of that lands close to the real calendar again: 268 of the 2,595 shifts put more than half the retrograde days back on retrograde days. That costs the test some power to find a real effect; it does not make a false alarm more likely.

The same two tests, gap in the average day and gap in volatility, for every planet's retrogrades, plus Mercury's whole shadow window. The slow planets are retrograde for months at a time, so their rows split the series into long blocks; the shifted calendars keep those blocks intact, so the p-values account for it.

Run enough tests and some will look striking by luck. With 9 rows and two tests each, the chance of at least one result under 5% with nothing going on is about 60% if the tests were independent; they overlap, so a little less. The fair bar for any one of the 18 is 0.05 ÷ 18 = 0.0028.

Read down the volatility column and every planet's retrograde days were calmer than the days around them. That is not the planets. A planet looks retrograde when Earth overtakes it, around opposition, when it stands opposite the Sun; for the slow outer planets that comes round at nearly the same date every year. Across the series Saturn was retrograde through most of Jul–Oct, Uranus of Sep–Jan, Neptune of Jul–Nov and Pluto of May–Sep.

Bitcoin has seasons too, at least in these years. Its four calmest months were Jul, Sep, Oct and Dec; its wildest was Mar, at 5.19% a day, the month of the 12 Mar 2020 crash, and none of the four is retrograde in Mar. So their rows measure Bitcoin's quiet season. Mercury's retrograde comes round every 116 days and lands on different dates each year, which is what makes it the fair test.

One row stands out. Across the whole shadow window, pre-shadow to post-shadow, the average day was −0.05%, against +0.20% on the other days: p = 0.02. Held only during the shadows, $100 of Bitcoin would have turned into $52; held only outside them, $1,511; held throughout, $784.

It is still not evidence. It was found by looking through 18 tests, where a p that small turns up by chance about 33% of the time, and it misses the corrected bar of 0.0028. It also weakens: in the first half of the series (2019–23) shadow days averaged −0.11% against +0.24%, p = 0.04; in the second half (2023–26), 0.00% against +0.16%, p = 0.22.

Look inside the window and it turns against the lore. The retrograde itself, the part the stories blame, was the best stretch of it: +0.12% a day, against −0.22% in the pre-shadow and −0.12% in the post-shadow. And the shadow's lead rests on a few violent days. The worst day of the series, 12 Mar 2020 (−40.0%), fell in Mercury's post-shadow, two days after it turned direct; drop that one day and p rises to 0.07, and the post-shadow's average goes from −0.12% to +0.02% a day.

A version of the test that counts only each day's rank, blind to how big it was, gives p = 0.12, and Bitcoin rose on 50% of shadow days against 51% of the others (p = 0.47). The shadow was not worse so much as wilder: it holds 7 of the series' ten worst days and 7 of its ten best. And the price record it was found in is not one market mood but four falls of more than half (2019–20, 2021, 2021–22 and 2025–26) and the rallies between.

A pattern found by looking has to survive data it was not found in. Bitcoin traded long before Binance listed its perpetual, so the same tests were run on Bitstamp's daily closes from 1 Jan 2012 to 7 Sep 2019: 2,807 days and 24 more Mercury retrogrades, none of which the shadow result had seen.

Mercury's retrogrades again did nothing: Bitcoin rose over 12 and fell over 12, p = 0.67. And the shadow result reversed. Shadow days averaged +0.49% against +0.08% on the others (p = 0.07). Sitting out the shadows, the rule that had paid since 2019, would have turned $100 into $302 over those years; holding only in the shadows, $69,413; simply holding, $209,688. That is what a lucky split looks like.

Prices are the daily closes of the BTCUSDT perpetual contract on Binance, in UTC days, from 8 Sep 2019, the first day of Binance's history for it, to 17 Oct 2026. It is priced in Tether (USDT), not dollars, and kept near the spot price by funding payments: on the same days its close sat a median 0.06% from Bitstamp's spot BTC/USD, and never more than 1.6%. The earlier check uses Bitstamp's spot closes from 1 Jan 2012, the first full year of its history. On three days in January 2015 Bitstamp recorded no trades; those days carry the last price. Retrograde and shadow dates are computed by astronomy-engine (VSOP87), the ephemeris behind every date on this site, as described in the methodology.

A retrograde's move runs from the close of the day before its first retrograde day to the close of its last; a day is retrograde when Mercury is retrograde at noon UTC. Only retrogrades wholly inside a price series count: 22 in the main one, 24 in the earlier. Averages are of daily log returns. The famous days are dated from the sources linked in their table, and their phase is taken at noon UTC. Seven years is a short record for a market this young and this volatile: a small effect could hide in it, but not the dramatic one the lore describes.