Results: Compatibility and Relationship Survival
1. Hypothesis
Does an astrological "Compatibility Score" predict marriage longevity? We tested multiple variations:
- V1/V2: Aggregate Cosine Score (Basic aspects).
- V3: Survival Analysis separating Death from Divorce.
- V4: Adding Zodiac Signs (Demographics).
- V5: Removing Generational Planets (Uranus, Neptune, Pluto).
2. Methodology
- Data: Wikidata (N = 2,722 couples).
- Metric: Weighted ElasticNet Regression & Kaplan-Meier Survival Estimators.
- Definition of Success: Relationship duration (Years) and "Survival" (Avoiding Divorce).
3. Key Findings
Experiment V4: The "Pluto Artifact"
When we included Zodiac Signs and Outer Planets, our model achieved an $R^2$ of 0.08 (8% predictive power).
- However, the top predictors were Pluto in Libra (Negative) vs Pluto in Aries (Positive).
- This was a demographic artifact: Pluto marks the generation you were born in. The model simply learned that people born in the 19th century (Pluto Aries) had longer marriages than modern couples (Pluto Libra).
Experiment V5: Traditional Planets Only (The 'Clean' Test)
To eliminate this demographic bias, we removed Uranus, Neptune, and Pluto, leaving only the visible planets (Sun through Saturn).
Results:
- Predictive Power Collapsed: The $R^2$ dropped to 0.01 (1%). The "signal" vanished once we removed the birth-year proxies.
- Top Predictors (Weak):
-
Saturn-Saturn(+0.50): Couples with Saturn in the same place (Same age or 29 years apart). -
Jupiter-Jupiter(+0.47): Couples born in the same year. - Conclusion: Even with only traditional planets, the model only finds innocent proxies for "Age Similarity."
-
- Survival Analysis:
- High Compatibility (Top 20% scores) vs Low Compatibility (Bottom 20%).
- Difference: The divorce rates are identical.
- 5-Year Survival: High Comp (88.1%) vs Low Comp (87.9%).
4. Visualizations
A. Survival Curve
There is no separation between the Blue line (High Comp) and Orange line (Low Comp). Astrological compatibility provides no survival advantage.
B. Regression Coefficients (Traditional Only)
The coefficients are small and dominated by Saturn/Jupiter (Age). No specific aspect (e.g., "Venus-Mars") stands out as a driver of longevity.
C. Distributions (Statistical Detail)
While the KDE plots look very similar, a Kolmogorov-Smirnov (KS) test reveals slight, albeit weak, differences on the edges for specific aspects.
Long Term (>35y) vs Short Term (<7y) Marriages:
| Aspect | Mean Cosine (Short) | Mean Cosine (Long) | Significance (p-value) | Interpretation |
|---|---|---|---|---|
| Sun-Moon | -0.06 (Slight Opp) | +0.02 (Neutral/Conj) | 0.027 (Sig) | Long-term couples have slightly more Conjunctive Sun-Moon aspects. |
| Mars-Mars | -0.05 (Slight Opp) | +0.05 (Slight Conj) | 0.027 (Sig) | Long-term couples have slightly more compatible Mars placements. |
| Venus-Mars | -0.05 | -0.00 | 0.33 (Not Sig) | No difference in "Romantic Chemistry". |
| Sun-Sun | -0.02 | +0.02 | 0.13 (Not Sig) | No difference in Ego compatibility. |
Nuance: Although the "Aggregate Score" failed, the Sun-Moon and Mars-Mars aspects show a statistically significant (p < 0.05) tendency: Long-lasting couples are slightly more "Conjunctive" (aligned) than short-term couples. The effect size is small (Diff ~0.10 Cosine), but it is not zero.
Aspect Distributions
5. Final Conclusion
After rigorous testing—controlling for censoring (death), removing demographic artifacts, and testing non-linear machine learning models—we find no evidence that traditional synastry (aggregate geometric aspects between charts) predicts relationship survival in the Wikidata dataset. However, there is some significance in Sun-Moon synastry and Mars-Mars synastry.