The chi-square says yes.
The right-sized test says no.

Backlog #72: does your zodiac sign predict anything — specifically, does birth month sort people into careers the way sun-sign astrology claims? Wikidata carries exact birth dates for hundreds of thousands of notable people tagged by occupation, which makes this directly testable at a scale no astrologer's anecdote ever is. This run pulled every human item with occupation actor, musician, or scientist and a day-precision birth date (P569) born 1901–2009, and checked each group's birth-month distribution against the one honest null: not an equal 1-in-12 split (zodiac signs are not equal length), but the exact share of the calendar each sign's date range actually covers — 29 days for Capricorn up to 32 for Cancer. One data-quality step taken before any fit ran: January 1 carried 2.5× any other day's count in a spot-check, a known Wikidata placeholder-date pattern (an unknown exact day sometimes gets stored as day-precision January 1 instead of year-only precision) — those rows are dropped by construction, not discovered as a surprise in the results.

The naive test says every group is "significant," loudly. A chi-square goodness-of-fit against the calendar-weighted null rejects the no-effect hypothesis for all three occupations: actors (213,363 people) χ²(11)= 147.8, p=4.3e-26; musicians (66,926) p=2.2e-29; scientists (21,431) p=5.9e-06. Read naively, that looks like three confirmations of astrology. It isn't: every single sign's deviation from its expected share, in every group, is under 0.8 percentage points — actors' biggest gap is Pisces at +0.48 points, against an expected share of 8.2%. At N in the hundreds of thousands, a chi-square has the power to flag a quarter-point wobble as cosmically significant. That is a property of the sample size, not of the zodiac.

Rebuilt as the regression this desk actually reports, the result reverses. There are only 12 independent sign-categories in this test, not 213,363 independent trials — so the honest fit regresses each sign's observed birth share on its expected birth share across those 12 points, where the no-effect prediction is a slope of exactly 1. None of the three groups can rule that out: actors 1.33, 95% CI [+0.64, +2.03] (p=0.31 vs. slope=1); musicians 1.23, CI [-0.15, +2.62] (p=0.71); scientists 0.50, CI [-0.74, +1.74] (p=0.39). Every one of the three intervals contains 1. Per-sign, after a Bonferroni correction for testing all 12 signs in each group, a handful of individual signs do still clear p<0.05 (3 of 12 for actors, 5 for musicians, 3 for scientists) — the table below marks them — but the whole-distribution test built for the actual number of categories here says the pattern is not distinguishable from the calendar alone.

The one thing all three unrelated occupations agree on is the tell. Pisces posts the single largest excess share in every group tested — actors +0.48 points, musicians +0.70, scientists +0.62. Astrology gives no reason actors, musicians and scientists should share a sign bias — performers and researchers aren't claimed to run the same sign. A shared data artifact does give a reason: actors' and musicians' full 12-sign excess-vs-expected fingerprints correlate with each other, r=0.72, p=0.008 — close enough that whatever is nudging actors toward Pisces is very likely also nudging musicians there, rather than two independent zodiac effects landing on the same sign by chance. The scientist correlations are weaker and don't clear this desk's bar (actor–scientist r=0.30, p=0.34; musician–scientist r=0.41, p=0.19), consistent with a weaker or different version of whatever the shared skew is. The most likely source, not independently tested here: Wikidata's notable-people pool is not a random population sample, and its underlying mix of countries, eras, and record-keeping conventions could easily carry its own small birth-month skew that has nothing to do with any person's star sign.

Read plainly: no group's zodiac sign predicts its career once the test is sized to the actual 12 signs being compared, not the inflated count of people in them. The large-N chi-square's unanimous "rejection" is a lesson in statistical power, not a confirmation of astrology — and the one real cross-occupation pattern in the data points at a quirk of this particular notable-people dataset, not at anyone's birth date.