Folk causation

Does Mercury Retrograde Break the Internet? The Results

October 26, 2026

The short answer

No. Mercury retrograde does not predict tech outages.

Outages were slightly more common during retrograde in some cuts of the data and slightly less common in others. None of those differences were bigger than what you get by drawing retrograde dates at random, and knowing the retrograde calendar did nothing to help forecast outages in years the model hadn’t seen.

That is the result I predicted before looking.^[1] The point of writing the prediction down first was so that I couldn’t claim that after the fact.

What I tested

The claim: when Mercury is retrograde, technology breaks.

The test: count every incident that six tech companies posted on their public status pages, and compare outage days inside the retrograde windows with outage days outside them.

• The data: 6,306 incidents from GitHub (since 2010),^[2] DigitalOcean (since 2013),^[3] Reddit,^[4] Discord,^[5] Dropbox^[6] and Google Cloud,^[7] through September 2026. Scheduled maintenance was excluded.
• The calendar: every Mercury retrograde from 2010 to 2026, computed from NASA’s planetary data.^[8] As a check, my 2025 shadow dates match the ones astrologers publish exactly.^[9]
• The fair comparison: status pages log more incidents every year as companies grow. So each retrograde was compared with the rest of the same month, not with some quieter year.
• The chance check: I slid the whole retrograde calendar to random dates 10,000 times and asked how often a fake calendar showed a gap as big as the real one. If fake calendars do it often, the real gap means nothing.

What I found

Read the last column as “how ordinary is this?” A gap that random dates produce one time in three is completely ordinary. To count as unusual, a gap needed to show up in fewer than about 1 in 80 fake calendars, after accounting for running four tests. Not one came close.

The direction also flips. If retrograde broke things, the longer shadow window should look worse as well. It looks slightly better.

Can it forecast? I trained a model on 2010–2020 and asked it to predict daily outages for 2021–2026, with and without the retrograde calendar. The real retrograde calendar did worse than two out of three random calendars, and the shadow calendar did no better. By the rule I set in advance, it needed to beat 95% of them.

The refinements believers add

Astrologers add details to the basic claim, so I tested those too:

• Station days (when Mercury appears to stop and turn, said to be the worst): about 3% more outages. Ordinary.
• Major and critical outages only: no consistent direction. In the longest record, serious outages were actually 24% less common during the shadow, which is the opposite of the claim and still within chance.
• The sign Mercury turns in: retrogrades in water signs showed about 20% more outages, the largest lean anywhere. Random calendars produce a lean that size about one time in six, and I ran 18 of these extra tests. Run that many and one of them will look interesting by luck alone. That’s exactly why the plan was written down first.

Exploration: does the sign matter?

This one wasn’t in the plan, so treat it as exploration, not as a test. People often say it depends on the sign, so I split the retrogrades by the sign Mercury turns in.^[10]

• Too few cases to say much. Sixteen years holds only 2–6 retrogrades per sign, so every number swings widely, from roughly half as many outages to more than double.
• Nothing stood out from chance once you account for checking 12 signs.
• Pisces had the biggest lean, about 50% more outages in both data sets, and it’s where the water-sign lean above comes from. Random calendars produce a lean that size about 1 time in 16, so it’s what you’d expect to see somewhere among 12 signs.
• Gemini and Virgo, the signs Mercury rules and supposedly the hardest hit, didn’t stand out. Gemini was up 41% in one data set and down 3% in the other.

This is a different question from your own sign. A GitHub outage hits everyone at once, whatever their birth chart.

Why the belief survives anyway

None of this should be surprising, and it says nothing bad about the people who believe it.

• Tech breaks constantly. These six companies alone logged an incident on about 2 out of every 3 days. Any window will contain outages.
• The window is huge. Retrograde covers about 19% of the calendar. With the shadow, it’s about half of all days. A rule that covers half of all days will catch half of all failures by default.
• We check the calendar after something breaks. Nobody looks up Mercury on a day when everything works.

What this does and doesn’t show

• It does show: across 16 years and six companies, outages don’t bunch up during Mercury retrograde beyond what chance produces, and the calendar can’t predict them.
• It only covers outages companies report publicly. It says nothing about personal tech trouble: your phone freezing, your laptop dying, an email that never arrived. Those never reach a status page. Testing them would mean tracking individual people’s tech problems day by day, which is a study worth doing and a different one.
• It doesn’t test “it depends on your sign.” That claim is about individual birth charts and would need a different, person-by-person study.
• A gap in the data: Google Cloud’s history had to be rebuilt from archived web pages, and the archive was missing early 2017 and December 2022 to April 2024. Those days were left out for Google Cloud only. Every other provider is complete.
• Where I filled in a detail the plan didn’t specify: the forecasting test needed a baseline guess. I used each company’s average outages over the previous year.

The verdict

Hypothesis: Mercury retrograde is predictive of tech outages. Not supported.

The code, the raw data and every number behind this page are public,^[10] so anyone who thinks I got it wrong can check.

A note on how this was written

Based on the way this is written, you can probably tell it was written heavily by AI. The structure of everything is built by me in conjunction with AI, and nothing is published without my final approval.