← Back to research
research··7 min

Serial-Killer Astrology Claims Under Control-Matched Tests

N=60 natal charts, 59 dated attacks, and 83 expanded transit events: what survived, what failed, and why this cannot be used as behavioral screening.

Safety boundary: this is retrospective research on deceased or public historical criminal figures. It is not a diagnostic tool. It has zero authority to score, screen, assess, or predict behavior for living people.

True-crime astrology is full of confident claims: Scorpio is overrepresented, mutable signs hide double lives, Mars-Pluto makes violence obvious, the full moon triggers attacks, and minor bodies like Nessus or Lilith reveal the shadow material. Those claims are easy to cherry-pick from individual charts. The harder question is whether they survive controls.

We ran four separate retrospective tests:

Test Sample Control strategy Main readout
Natal chart scan Tier 1 AA N=46; combined N=60 Same-day random-time controls for houses; year-matched random-date controls for slower configurations Primary AA cohort had no FDR survivors. Combined cohort had one Mars-Neptune hard-aspect pass.
Event-level transit scan 59 confirmed murder/attack dates across four offenders Shuffled the actual event dates across the four natal charts Neptune hard transit to natal Sun/Moon survived; Mars, Pluto, Saturn, lunar, and mundane Mars-Uranus triggers failed.
Spree-peak day scan 59 attack dates vs quiet days inside active spree windows Compared murder dates against each offender's own non-murder days during the same spree range No FDR survivors. Full/new moon, Mercury retrograde, Mars-Saturn tension, and tight Neptune peaks failed.
Expanded asteroid scan 64 subjects and 83 confirmed events Year-matched natal controls plus shuffled-event transit controls No natal asteroid passes. Two transit asteroid anomalies survived: Lilith overrepresented and BU48 underrepresented.

The short version

The obvious true-crime astrology memes mostly failed.

Scorpio emphasis did not hold up. Mutable dominance did not hold up. Mars-Pluto and Mars-Saturn natal hard aspects did not hold up. Full moons did not hold up. Mercury retrograde did not hold up. Spree-peak day triggers did not hold up. Nessus, Apophis, Lucifer, Chiron, and most other minor-body natal claims did not hold up.

The strongest surviving signals were narrower and less cinematic:

Finding Observed Null mean Lift p-value FDR read
Combined natal Mars-Neptune hard aspect 23.3% 11.7% +11.7pp 0.0070 Passed in combined cohort; primary AA cohort did not pass FDR
Event-level Neptune hard transit to natal Sun/Moon 57.6% 30.3% +27.4pp 0.0010 Passed q <= 0.10
Asteroid transit: Mars/Saturn to natal 1181 Lilith 28.9% 17.8% +11.2pp 0.0050 Passed q <= 0.10
Asteroid transit: Mars/Saturn to natal 1998_BU48 7.2% 21.4% -14.2pp 0.0010 Passed q <= 0.10, but in the underrepresented direction

That is not a license to claim "we can detect criminals in charts." We cannot. The sample is retrospective, selected, public, small, and ethically constrained. The only responsible interpretation is: under matched controls, many popular claims vanish, and a few narrow retrospective anomalies are worth writing down for replication.

Natal charts: most memes failed

The natal scan started from verified serial-killer birth data, then excluded pre-1900 births because the standard DE421 ephemeris used by the research stack covers 1899-07-29 through 2053-10-09. The primary cohort used only Tier 1 AA birth-record subjects (N=46). The combined readout used Tier 1 and Tier 2 subjects (N=60).

The test used two controls:

  • Same-day random-time controls for house and angle claims. These keep the date constant and test whether the birth time creates unusual house placements.
  • Year-matched random-date controls for slower planets, aspects, and signs. These control for generational and seasonal sky patterns.

The headline failures are important:

Claim Combined observed Combined null p-value Result
Sun, Moon, or Ascendant in mutable signs 70.0% 70.5% 1.0000 Failed
Sun, Moon, or Ascendant in Scorpio 20.0% 24.4% 0.4376 Failed
Mars-Pluto hard aspect 5.0% 13.4% 0.0649 Failed
Mars-Saturn hard aspect 11.7% 17.7% 0.2168 Failed
Mars or Saturn in 8th/12th whole-sign houses 35.0% 31.2% 0.5564 Failed

The one combined-cohort survivor was Mars-Neptune hard aspect:

Cohort Observed Null mean Lift Cohen's h p-value Result
Primary AA N=46 23.9% 12.2% +11.7pp +0.309 0.0140 Raw signal; did not clear the primary FDR gate
Combined N=60 23.3% 11.7% +11.7pp +0.311 0.0070 Cleared q <= 0.10 in the combined readout

This is a replication lead, not a product claim. The primary high-confidence cohort is the more conservative readout, and it did not produce an FDR-clean natal finding.

Event transits: Neptune survived, Mars did not

The event-level transit scan used 59 confirmed murder or attack dates across Ted Bundy, Jeffrey Dahmer, Richard Ramirez, and David Berkowitz. The control design shuffled the exact event dates across those four natal charts, preserving the calendar dates and the birth charts while testing whether the personal transit-to-natal alignment was unusual.

That matters because slow planets can look meaningful when a researcher picks one famous chart and one famous date. Shuffling the same dates across the same charts asks a cleaner question: was this person's chart unusually activated on that date compared with the same date assigned to another chart?

Transit claim Observed Null mean Lift p-value Result
Mars hard transit to natal Sun 13.6% 13.2% +0.3pp 1.0000 Failed
Pluto hard transit to natal Sun/Moon 0.0% 5.5% -5.5pp 0.0470 Raw underrepresentation; FDR failed
Saturn hard transit to natal Mars 0.0% 6.8% -6.8pp 0.0370 Raw underrepresentation; FDR failed
Mars hard transit to natal Mars 10.2% 10.4% -0.2pp 1.0000 Failed
Moon hard transit to natal Mars/Saturn 16.9% 24.6% -7.7pp 0.0919 Failed
Mars or Saturn in natal angular houses 50.8% 56.4% -5.5pp 0.3956 Failed
Neptune hard transit to natal Sun/Moon 57.6% 30.3% +27.4pp 0.0010 Passed q <= 0.10

The practical reading is subtle: the common "Mars made them do it" style of transit claim did not survive. Neptune-to-luminary activation did, but it is a wide-window retrospective association, not a day-level trigger or a behavioral forecast.

Spree peaks: no day-level trigger survived

To test exact attack timing, we compared the same 59 attack dates against quiet days inside each offender's own active spree window. This is a stricter question: inside the same life period, same historical era, and same natal chart, do attack days differ from non-attack days?

No pre-registered spree-peak hypothesis survived FDR correction.

Spree-peak claim Observed Null mean Lift p-value Result
Mars-Saturn tension or Mars to natal Mars 16.9% 23.0% -6.0pp 0.2847 Failed
Full or New Moon alignment 16.9% 13.1% +3.8pp 0.4605 Failed
Moon conjunct natal Mars/Saturn 5.1% 5.6% -0.5pp 1.0000 Failed
Mercury retrograde 11.9% 19.6% -7.8pp 0.1269 Failed
Neptune tight hard transit to Sun/Moon 11.9% 19.4% -7.6pp 0.1489 Failed

This is the key guardrail for the Neptune result above. A wider Neptune transit-to-luminary association appeared in the event-level shuffle, but when we asked whether the exact peak days inside a spree were distinguishable from quiet days, even a tighter Neptune version failed.

Asteroids: no natal passes, two transit anomalies

The expanded scan tested 12 minor bodies across natal and transit hypotheses: Ceres, Pallas, Juno, Vesta, Chiron, Pholus, Nessus, Apophis, Lucifer, Lilith, 1996_TL66, and 1998_BU48.

This is exactly where the look-elsewhere problem gets dangerous. If you inspect enough minor bodies, something will always look archetypally perfect in one famous chart. The only way to keep that honest is to predefine the hypotheses and correct for the number of tests.

Natal asteroid claims produced no FDR-clean findings. Transit asteroid claims produced two:

Asteroid transit claim Observed Null mean Lift p-value Result
Mars/Saturn hard transit to natal 1181 Lilith 28.9% 17.8% +11.2pp 0.0050 Passed q <= 0.10
Mars/Saturn hard transit to natal 1998_BU48 7.2% 21.4% -14.2pp 0.0010 Passed q <= 0.10, underrepresented

Notice the second result: it passed because it was lower than expected, not higher. That is one reason these outputs should stay in the research bucket until independent replication exists.

What this changes

This work does not make the product more predictive. It makes the product more careful.

The landing-page-safe claim is:

  • Many easy true-crime astrology claims failed under matched controls.
  • A few narrow retrospective anomalies appeared.
  • None of these findings may be used to assess living people.
  • The results are useful mainly as a methodological warning: controls matter.

That warning is the point. Without matched controls, astrology writing can make almost any famous chart look inevitable. With matched controls, the story gets smaller, stranger, and more honest.

Source files

The generated research artifacts are:

  • services/astro-api/research_output/serial_killer_analysis.json
  • services/astro-api/research_output/serial_killer_analysis.md
  • services/astro-api/research_output/serial_killer_events.json
  • services/astro-api/research_output/serial_killer_events.md
  • services/astro-api/research_output/serial_killer_spree_peaks.json
  • services/astro-api/research_output/serial_killer_spree_peaks.md
  • services/astro-api/research_output/serial_killer_asteroids.json
  • services/astro-api/research_output/serial_killer_asteroids.md

The scripts live under services/astro-api/scripts/serial_killer_*.py.