Compatibility and Relationship Survival
1. Study Overview (The "Aggregate Score" Test)
This project tests the validity of the most common concept in pop-astrology: The Compatibility Score. Apps and websites often give users a percentage (e.g., "93% Match") based on summing up all the "good" and "bad" connections between two charts.
The Hypothesis: If astrological compatibility is cumulative, then couples with a higher "Total Harmonic Score" (Sum of planetary interactions) should have relationships that last longer (higher survival rate) than couples with a lower score.
2. Methodology: Large-Scale Survival Analysis
A. The Metric: Total Harmonic Score
We utilize a continuous mathematical model to quantify the total resonance between two birth charts using Cosine Similarity. $$ \text{Interaction} = \cos(\theta_{P1} - \theta_{P2}) $$
- +1.0: Conjunction (Unity)
- -1.0: Opposition (Polarity)
- Total Score: The sum of this interaction for all 100 planetary pairs (Sun-Sun to Node-Node).
B. The Dataset: N = 2,722
Unlike our initial pilot study (N=166), this analysis uses a massive dataset of 2,722 couples scraped from Wikidata, providing high reliability for relationship start/end dates.
C. Survival Analysis
We perform Kaplan-Meier Survival Analysis, treating Divorce/Death as the "Event" and Ongoing relationships as "Censored."
- High Compatibility: Top 50% of Total Scores.
- Low Compatibility: Bottom 50% of Total Scores. We then use Kaplan-Meier Curves to visualize whether the "High Comp" group drops off (divorces) slower than the "Low Comp" group.
3. Results Overview: The Null Hypothesis
The Result is Null. There is zero difference ($r \approx 0.009$) between the survival rates of highly compatible and incompatible couples when using an aggregate score.
- The survival curves overlap perfectly for 50+ years.
- This suggests that "Averaging" a synastry chart is mathematically flawed. (See
RESULTS.mdfor discussion on "The Soup Metaphor").
4. Key Files
-
analysis_survival.py: Performs the core analysis on the Wikidata CSV, calculates interactions, and fits the Kaplan-Meier curves. -
Results.md: Detailed report on the null findings and the philosophical implications for astrological calculation. -
survival_curves_new.png: The visual proof of the null result. -
new_couples_wikidata.csv: The massive dataset used for this V2 analysis.
5. How to Run
# Run the survival analysis script
python analysis_survival.py
7. The Regression Formula
We ran an ElasticNet Regression to find the "Best Fit" equation. The result ($R^2=0.02$) confirmed that Personal Planets (Sun/Moon/Venus/Mars) have weighting coefficients of nearly zero. The only predictive factors were Generational (Age) Markers:
Duration = 18.5 + (0.57 * Pluto-Pluto) ... (Detailed formula in RESULTS.md)
Data Provenance
Relationship Compatibility
- Source: Wikidata (Spouses).
- Link: https://query.wikidata.org