Project 08: Tropical vs. Sidereal Zodiac #

Source: bigastrologybook.com/2/research/19/project-8 Archive Date: 2026-03-21 Book: The Big Astrology Book of Research by Renay Oshop Dataset: 86 verified celebrity birth charts (from Project 06); 8,500+ trading days for GARCH extension

Research Question #

Which zodiac system is empirically superior: Tropical (seasonal, Western astrology, 0° Aries = Spring Equinox) or Sidereal (constellation-based, Vedic astrology, 0° Aries = first star of Ashwini)?

Due to axial precession, the two systems have drifted apart by approximately 24° — nearly one full sign. In 2025, the Lahiri ayanamsa (the standard Indian offset) is 24.21°, meaning roughly 80.7% of people have a different Sun sign in the Sidereal system than in the Tropical. This is astrology's central unresolved schism.

Hypothesis #

If one zodiac system reflects real physical or biological influences, models using that system's sign placements should predict observable outcomes better than models using the other. A null result in both would indicate that neither 12-fold zodiac categorization carries the relevant information.

Background: The Precession Problem #

The divergence between Tropical and Sidereal zodiacs is caused by Earth's axial precession — the slow wobble that shifts the equinoxes backward through the constellations at ~50.3 arcseconds per year. One complete cycle takes approximately 25,772 years (the Platonic Year). The two systems were roughly aligned around 285 CE; they have been drifting apart ever since.

Different Sidereal schools use slightly different ayanamsa values:

The spread across systems (~1.7°) is enough to reassign sign membership for people born near cusps, adding further ambiguity to any empirical test.

Experiment 1: Sun Sign Alone as Profession Predictor #

Method #

Using 86 verified celebrities from Project 06, a classifier was trained to predict professional category from Sun Sign (0–11) under each of three zodiac systems. Validation: Leave-One-Out Cross Validation.

Baseline: The largest category in the dataset ("Arts") accounts for 24.4% of observations. A classifier that always guesses "Arts" achieves 24.4% — this is the floor below which a model is less useful than no model at all.

Results #

All three systems performed worse than simply guessing the most common category.

The Symmetry Insight #

The exact tie between Tropical and Lahiri Sidereal at 17.1% is not a coincidence — it reflects the structure of error exchange. When the Sidereal shift reassigns a celebrity from Sign A to Sign B, it may correctly reclassify one person while misclassifying another. The errors trade symmetrically between systems, leaving the overall accuracy unchanged. This demonstrates that neither system is extracting information from the data; they are simply shuffling the same errors around.

Interpretation: Sun Sign alone contains zero information about profession in this dataset. Knowing someone's Sun Sign (in any zodiac) makes you less likely to guess their profession correctly than ignoring it entirely. This is a strong, direct null result — not "no significant difference," but actively harmful to prediction.

Experiment 2: Harmonics + Sun Sign (Controlled Test) #

Project 06 established that harmonic aspect strengths (H4 = squares, H5 = quintiles, H7 = septiles) do carry modest predictive information about profession. This provides a stronger test: when the harmonic signal is already in the model, does adding Sun Sign help?

Adding Tropical Sun Sign to the harmonic baseline provided zero lift. Adding Sidereal Sun Sign provided a marginal 2.3% lift — approximately 2 additional correct classifications in 86 — while remaining well below the naive baseline.

Experiment 3: Full Chart Comparison — The "Graha Shootout" #

The Sun Sign may be too coarse. A fuller test uses all nine traditional planets (Sun, Moon, Mercury, Venus, Mars, Jupiter, Saturn, Rahu/North Node, Ketu/South Node) as sign-placement features.

For the first time in this project, a zodiac-based model exceeded random guessing: the Sidereal Full Chart model hit 26.7%, narrowly beating the 24.4% naive baseline. The Tropical Full Chart remained below it.

Feature Importance: Which Planet Drives the Sidereal Advantage? #

Random Forest feature importances, aggregated by planet (Sidereal model):

The Sun Sign — the single most discussed and debated element in popular astrology — ranks last among the traditional planets. Mercury and Mars do the most work.

Statistical Caveats #

• Small N: N=86 produces wide confidence intervals on LOO-CV accuracy (~±10 pp at 95%). The Sidereal advantage of 2.3% over the naive baseline in Experiment 3 is not statistically significant on its own.
• Multiple comparisons: Three zodiac systems × three experiments = nine hypothesis tests. No Bonferroni correction would elevate the marginal Sidereal result to significance.
• Generational confound: As in Project 07, outer planet sign placements are correlated with birth era. The Mercury and Mars results are less contaminated by this confound (Mercury stays in one sign ~20 days; Mars ~45 days), making them the more interesting leads.
• Selection bias: The 86 celebrities are not representative of any population. Fame itself is a filter.

Conclusion #

The tropical vs. sidereal debate — which has divided astrologers for centuries — yields a clear empirical answer for the specific task of profession prediction: neither Sun Sign system contains meaningful information when used as a 12-fold categorical variable. Both systems perform worse than random guessing when classifying profession by Sun Sign alone.

When extended to full-chart analysis using all nine traditional planets, the Sidereal system shows a marginal advantage over Tropical (26.7% vs. 20.9% accuracy). This is small and statistically fragile, but consistent across multiple experiments.

More importantly, the analysis identifies where zodiac sign information might reside: not in the Sun (which ranks last in feature importance), but in Mercury (how someone thinks and communicates) and Mars (how someone acts and fights). If any zodiac-based signal exists in profession data, it appears to be embedded in these faster-moving planets rather than in the annual Sun cycle that dominates popular astrological identity.

The signal in astrology, if it exists, lies in continuous angular relationships (Project 06) and specific planetary placements — not in the 12-fold categorical bins of zodiac signs.

Archived code and raw data outputs preserved in backup/.

Tropical vs Sidereal Zodiac #

Research Question #

"Which Zodiac system is the 'Real' one?"

This is the great schism in astrology:

• Tropical (Western): Based on the Seasons. $0^\circ \text{Aries}$ = Spring Equinox.
• Sidereal (Vedic/Eastern): Based on Constellations. $0^\circ \text{Aries}$ = First star of Aries (Ashwini).
• The Divergence: Due to axial precession, these systems have drifted apart by ~24° (nearly one full sign).

We test which system aligns better with observed reality by asking a Machine Learning model to predict Profession (Scientist vs Artist) using only the Sun Sign calculated in each system.

Methodology #

1. Dataset: The Project 6 Verified Celebrity Data (N=86).
2. Systems Tested:
   • Tropical: Used in US/Europe.
   • Sidereal Lahiri: Official Indian standard.
   • Sidereal Fagan-Bradley: Reconstructed Babylonian/Western Sidereal.
3. Experiment:
   • Calculate Sun Sign (0-11) for every celebrity in all 3 systems.
   • Train a Classifier to predict their profession.
   • Check which system yields the highest accuracy.

Results #

See RESULTS.md.

Execution #

Run the comparison:

python3 analysis_zodiac_test.py

This generates:

• zodiac_comparison_results.csv: Accuracy scores.
• zodiac_comparison.png: Bar chart of performance.

Data Provenance #

Astronomical Data #

• Source: Swiss Ephemeris.
• Link: https://github.com/astrorobotics/pyswisseph
• Reference: Dieter Koch and Alois Treindl (Astrodienst).