CUE Research · Study 15

Serial Killers

Birthdays of people recorded as serial killers, read the same way as every other group.

The question: Do any numbers, signs or birthdays appear more often among serial killers than chance would give?

What stands out

No clear differences. Across every group of 50 or more and every measure, the numbers are within what chance produces.

Worth watching, not enough to claim

  • Sun sign

    Everyone: Pisces appears 92 times against about 67.1 expected (1.37×).

  • Sun sign

    Born since 1900: Pisces appears 81 times against about 59.7 expected (1.36×).

  • Birth quarter

    Born since 1900: Jan–Mar appears 210 times against about 179.6 expected (1.17×).

  • Birth quarter

    Everyone: Jan–Mar appears 234 times against about 201.9 expected (1.16×).

  • Day energy

    Born since 1900: 6 day energy appears 92 times against about 71.7 expected (1.28×).

Everyone: nothing clear. Jan–Mar is worth watching (234 against about 201.9 expected) but not strong enough to call.

Everyone Expected, random birthdays 1710–2002 Clear difference Worth watching

Context What explains it

This study exists because the question is asked often, and it deserves a careful answer rather than a sensational one. Records are uneven: some countries and eras are far better documented than others.

A pattern in a group like this would describe the records, not explain anyone’s actions.

Source The data

  • Serial killers. Retrieved 2026-09-28. View source
  • People Wikidata records with the occupation “serial killer”, generally those convicted of or confirmed as responsible for multiple murders. Records are uneven by country and era.
  • Verification: 817 of 1300 included. A person is included when Wikidata records one birth date at day precision.
  • Expected values: every day of the years each group was born in, weighted by how many people were born in each year, run through CUE’s engine. Weighting matters for Chinese signs and Life Paths, which depend on the year.
  • A difference is called clear only in groups of 50 or more and at least three standard errors from expected. Because each study tests about a hundred values, a clear difference needs 3.5 standard errors; 2.5 to 3.5 is only “worth watching”.

Reproduce it: the raw data snapshots and build scripts are in CUE’s code repository (scripts/data and scripts/build-*-study.mts).

Further reading

  • Wikidata, occupation records