♓ · ♏ · ♊
✦ Investigation · astrology and statistics

Is there a zodiac sign
of serial killers?

I checked a viral claim across 824 people. The real trap isn't which sign comes out on top — it's what people think a ranking proves.

This investigation tests one specific claim. It doesn't claim to "prove" or "disprove" astrology as a whole.

The internet loves assigning a zodiac sign to serial killers. Depending on the post, it's Scorpio — dark, secretive, intense — or sometimes Gemini, Pisces, or Sagittarius. The lists change, but the conclusion stays the same: certain signs are supposedly "more prone to it," "predisposed," or simply more common among violent criminals.

I wanted to start from a more open question: are any zodiac signs actually overrepresented among serial killers? Not just a sign that lands at the top of a ranking, but a sign that appears more often than you'd reasonably expect.

In a list of twelve signs, there will always be a first and a last. That doesn't mean the first one is statistically abnormal.

The raw ranking already tells a different story

After cleaning the corpus, Pisces comes out on top with 94 people. Aquarius, Capricorn and Sagittarius follow with 74 people each. Scorpio sits at 65 — not first, but seventh. Gemini closes out the ranking with 56 people.

It's tempting to conclude: "Actually, the serial-killer sign is Pisces." That would be making the exact same mistake with a new winner. A high count only becomes overrepresentation once it's been compared against a relevant expected distribution.

Three claims that keep getting mixed up

  1. The descriptive ranking. "In this list, this sign comes first." That's an observation about one particular corpus.
  2. Statistical overrepresentation. "This sign appears more often than you'd expect in a comparable population." This claim requires a reference and a test.
  3. Individual or causal predisposition. "Belonging to this sign would increase the odds of becoming violent." This is a much stronger claim, and it doesn't automatically follow from the first two.

The jump from the first level to the third is methodologically unjustified. Even if a genuine statistical overrepresentation were detected, it alone would not demonstrate that a sign causes, predicts, or predisposes someone to criminal behavior.

Before counting anything: fifteen fictional characters

The starting corpus was built from Wikidata. It counted 839 entries classified as serial killers. Checking it turned up a very concrete problem: fifteen entries corresponded to fictional or non-human characters. They were explicitly excluded, without silently editing the source file.

839initial Wikidata entries
−15fictional characters excluded
824humans in the final corpus
139US cases covered by the historical reference

Birth dates were then verified, and every calendar variable — year, month, day and sign — was recalculated from the retained date. This step doesn't make the corpus perfect: Wikidata is neither an exhaustive criminological registry nor a universal definition of the term "serial killer." But it does keep the result from resting on duplicates, artificial dates, or fictional characters.

The statistics that follow, without the jargon

"Expected count"
The number you'd predict for a sign if no sign stood out at all, given the chosen reference.
The χ² test — pronounced "chi-squared"
It rolls every gap between observed and expected counts into a single score. The bigger that score, the more unusual the whole ranking is. The "(11)" is a mathematical setting tied to the twelve categories — you don't need it to understand the result.
The p-value
It shows how surprising a ranking at least this unbalanced would be if no sign actually stood out. A p-value of 0.131 means a result at least this uneven would show up roughly 13 times out of 100 in that scenario. It is not "13% odds that our conclusion is wrong."
The 5% threshold
By convention, people often start calling a result "statistically significant" once p drops below 0.05. That's a benchmark decided in advance, not a magic line between true and false.
The Pearson residual
It measures how far off a sign is, in a unit comparable to its usual variation. A residual of 0.75 is small; around 2, the gap starts to draw attention — without automatically being proof, especially when twelve signs are being examined.
The question, translated: "Are the height differences between the twelve bars actually extraordinary, or do they still look like what chance produces routinely?"

First test: what if every sign were worth one-twelfth?

The first comparison assumes each of the twelve signs should represent exactly one-twelfth of the corpus. That's the simplest hypothesis — and also the most naive.

Full corpus, uniform expectation

In plain terms: a ranking at least this unbalanced could show up roughly 13 times out of 100 even if no sign actually stood out. That's not rare enough to conclude there's an anomaly.

Technical result: χ²(11) = 16.29 · p = 0.131

Put more simply: if you randomly split 824 people across twelve signs, you'd almost never get twelve perfectly equal groups of 68 or 69 people. Some signs would naturally land above average and others below. The test exists precisely to check whether the gaps are too large to be explained by that ordinary variation. Here, they aren't.

Second test: the signs aren't all the same length

The 1/12 expectation isn't quite exact. Zodiac periods don't all span the same number of days, and leap years add a day to Pisces. A second comparison therefore calculated the expected count based on each birth year and the actual number of days tied to each sign.

Full corpus, calendar days

With this reference, a result at least this unbalanced would show up roughly 5 to 6 times out of 100 if no sign actually stood out. That's close to the conventional 5% benchmark, but it still isn't enough to conclude anything — and this comparison still doesn't account for real birth seasonality.

Technical result: p = 0.055

Among the 733 people born in 1900 or later, the result is very close: p = 0.057. But both of these comparisons still share an important limitation. They assume that, aside from calendar length, births are uniform throughout the year. In reality, they aren't.

The real challenge: birth seasonality

Humans aren't born at the same rate every month. That seasonality varies by place and by era. To seriously test a zodiacal overrepresentation, the ideal would be to compare each person to the birth distribution of their own population, in their own time.

But the corpus is international and spans 1710 to 2002. Building a perfectly adjusted worldwide reference would require comparable historical series for dozens of territories across several centuries. Simply using recent French birth data as a control would have introduced a massive time and geography bias.

So I settled on a narrower but more defensible comparison: people whose birthplace was unambiguously tied to the United States, born between 1950 and 1979. For those years, official National Center for Health Statistics tables provide the number of US births by month.

The best-controlled test

This analysis covers 139 people. For each one, the expected distribution accounts for the actual seasonality of US births during their birth year. At the monthly level, p = 0.438: a gap at least this large would show up roughly 44 times out of 100 if the dates simply followed ordinary seasonality. So there's nothing unusual about the observed monthly distribution.

To get back to signs, monthly births were spread across the days of each month, then aggregated by zodiac boundaries. That conversion assumes births are uniform within a given month — an explicitly documented approximation. The zodiacal test gives p = 0.523: a ranking at least this irregular would show up roughly once every two times with no zodiacal preference at all. That's the opposite of a rare signal.

And Scorpio, specifically?

14
observed
vs.
11.47
expected

The observed/expected ratio is 1.22: we counted 22% more Scorpios than expected in this small sub-sample. But that doesn't mean "22% more risk" for any individual. Measured against the variation you'd normally see from one sample to another, this gap is only worth 0.75 units — a small move, not an anomaly. It doesn't constitute evidence of overrepresentation.

The largest positive gap in this sub-sample actually belongs to Leo, with 18 people observed versus 12.44 expected. But that gap is only 1.58 times the usual variation around the expected count. To start talking about a signal, two things would have been needed: first, that the overall distribution across the twelve signs be itself abnormal — which it isn't, at p = 0.523 — and then that one sign hold onto a strong enough gap after accounting for the fact that we're examining twelve of them. So there's no magic number, like "20 Leos," that alone would settle it: the verdict depends on the whole distribution. Here, neither Leo, nor Scorpio, nor any other sign meets those conditions.

What the analyses show, taken together

No overall test crosses the conventional 5% threshold. This doesn't prove every possible distribution is perfectly identical, nor that a tiny association would be undetectable in some other corpus. It means the data and comparisons carried out here provide no statistical evidence that any particular sign is overrepresented.

Why the myth still feels convincing

A ranking is a story-generating machine. It always produces a winner, a last place, and a few gaps that are easy to comment on. If Scorpio lands high, people point to its intensity. If Gemini lands high, people talk about a split personality. If Pisces comes first, there's always a story to invent after the fact about their emotional depths.

This kind of reasoning starts from the result and builds its explanation backward. It never asks whether the gap was expected, whether the reference population is relevant, or whether the same story would have been told with a different ranking.

"A sign is always going to come first, dear. The real question isn't which one — it's whether it's higher than what chance and real birth patterns already lead you to expect. 🐑"

What this investigation doesn't claim to show

Finally, this study isn't trying to "prove astrology wrong." It tests one precise, falsifiable claim: that certain signs are statistically overrepresented among the serial killers in the studied corpus.

So, is there a serial killer zodiac sign?

Not in the data analyzed. Pisces leads the raw ranking, Gemini comes last, and Scorpio sits seventh. But none of the overall comparisons — from the uniform expectation to controlling for real historical US birth seasonality — reveals a statistically abnormal distribution across the twelve signs.

So the conclusion isn't that some other sign would have dethroned Scorpio. It's that the ranking, on its own, never let anyone name a sign "predisposed" to crime. In this corpus, with the comparisons carried out, we found no evidence that any particular zodiac sign is overrepresented among serial killers.

✦ Explore your birth chart as a tool for reflection →

Frequently asked questions

Which sign comes first in the corpus?

Pisces, with 94 out of 824 people. That's a descriptive ranking: the overall tests don't support concluding that Pisces is statistically overrepresented.

Are Scorpios more common than expected?

In the best-controlled US analysis, 14 Scorpios were observed against 11.47 expected. That gap is small, with a Pearson residual of 0.75, and the overall zodiacal distribution is non-significant (p = 0.523).

Why isn't it enough to compare each sign to 1/12?

Because zodiac periods aren't all exactly the same length, and human births vary by month, year and population. A properly matched demographic reference is more meaningful than a theoretical one-twelfth.

Does this investigation prove astrology is false?

No. It addresses one specific claim: whether zodiacal overrepresentation exists among serial killers. It isn't a test of astrological practice or interpretation as a whole.

Sources and method

Observational, exploratory analysis ✦ The results describe this corpus and are neither an individual criminological profile nor a causal claim about astrology. 🌙