What we tested
Retrodiction — the engine analyses a birth chart and predicts life events that have already occurred, then compares its predictions against the documented record. This is the gold standard for astrological validation because the engine has no access to the person's biography during analysis.
Sample
157 verified birth charts with 750+ documented historical events — career changes, marriages, health events, financial shifts, relocations, and bereavements. Charts were selected to span a range of ascendants, planetary configurations, and life trajectories (not just "celebrity" charts).
| Rodden Rating | Count | Description |
|---|---|---|
| AA | 88 | Birth certificate or hospital record |
| A | 44 | Quoted from memory by the person |
| B | 19 | Biography or family report |
| C | 6 | Approximate, rectified where possible |
Scoring
Event detection (93.5%). Did the engine identify the correct type of event in the correct time window? Partial credit for events detected within ±2 years of the documented date; full credit within ±1 year.
Magnitude accuracy (83.5%). Did the engine correctly assess the severity or significance? A 5-tier scale (minor → transformative) is used, with partial credit for off-by-one tier errors.
The engine reliably detects when major life events occur in a chart. Magnitude — how big or small an event will be — is harder and remains our primary area of improvement.
What this doesn't mean
Retrodiction accuracy on known charts is not the same as predictive accuracy on future events. Future prediction is inherently harder because it requires timing precision that retrodiction (with its partial-credit windows) doesn't demand. We do not claim 93.5% accuracy on future predictions, and anyone who does should be treated with skepticism.
What the engine does not predict
By policy and by classical scriptural convention, Oracle Engine does not attempt to predict:
- Exact dates of death — classical texts like Ayurdaya estimate longevity ranges, not exact days
- Lottery numbers or specific financial windfalls — the chart shows themes, not outcomes of discrete random events
- Medical diagnoses — disease indications in the chart are shown as areas of vulnerability, not clinical diagnoses
- Actions of other people — Oracle reads your chart, not theirs
How we're improving
Every week, we run the engine against the validation set and log where it got magnitude wrong, where it missed events entirely, and where it false-positived. The engines that contribute most to error get refactored first — typically via stronger scripture cross-referencing and tighter dasha timing rules.
Want to audit the method?
If you're a professional Jyotish scholar or data-oriented practitioner and want to audit our scoring methodology on specific charts, email hello@oraclevedicastro.com. We'll share the validation protocol and scoring rubric in full. Nothing here is hidden.