Research
What survived. What didn’t.
Long-form writeups of findings from the Astral Edge research program. Every post cites p-values, effect sizes, and sample sizes so you can verify the numbers against the permutation-study source. Full methodology is in /methodology.
- methodology··6 min
We Ran 10,000 Permutation Tests on Astrology
What survived, what didn't, and why most published astrology findings won't replicate
Read → - finding··5 min
The Zodiacal Releasing Peak Signal Is Contrarian. The Data Said So Twice.
A 2026 permutation study finds Brennan's ZR L1 peak doctrine predicting less luck, not more. Update 2026-04-18: the signal weakened at expanded N.
Read → - finding··4 min
Saturn Return Isn't a Crisis. At Least Not in This Data.
Valens said age 58 was a climacteric. We looked at 534 life events. Peaks win.
Read → - finding··4 min
One Astrology Finding Just Survived the Strictest Multiple-Comparison Correction
Vedic benefic mahadasha clears Benjamini–Yekutieli at q=0.05. Here's what that means.
Read → - methodology··5 min
When a Trained Model Performs Worse Than Theory
Our logreg had ROC AUC=0.505. Replacing it with a hand-weighted theory prior raised lottery top-decile prediction 51%.
Read →
⚠ Most of these posts were written against the Phase-0 pipeline. That pipeline's statistical null and discovery/held-out split were both invalidated in external review. When rebuilt correctly (see /methodology), zero hypotheses replicated. Posts that reference specific "survivors" or "held-out lift" numbers carry retraction banners. The methodology lessons still stand; the specific statistical claims do not.