Validation scope
The production calculator uses Astronomy Engine 2.1.19. Expected planetary and Ascendant values were produced separately with Swiss Ephemeris 2.10.03 under the documented Western tropical conventions. Swiss Ephemeris code and ephemeris files are not shipped in the website.
The retained golden set covers different continents, hemispheres, dates, longitudes, daylight-saving states and zodiac boundaries. It also tests exact, approximate and unknown-time behavior rather than checking only a few ordinary midday examples.
| Calculation group | Comparisons | Acceptance rule |
|---|---|---|
| Moon | 26 | Longitude difference no more than 0.05° |
| Rising | 26 | Longitude difference no more than 0.10° plus exact orientation match |
| Sun, Venus and Mars | 78 | Longitude difference no more than 0.05° |
| Mercury through Pluto | 156 | Longitude difference no more than 0.05° |
| Total | 286 | All retained comparisons must pass |
Civil-time and timezone cases
Astronomical accuracy depends on converting a local civil label to the correct instant. The test evidence includes ordinary offsets, half-hour and quarter-hour zones, daylight-saving gaps, repeated clock labels, date-line behavior and historically skipped civil dates.
A repeated local time is not silently resolved to the earlier or later occurrence. Both real offsets are presented for confirmation. A nonexistent local time is rejected, and an entirely skipped date returns no result instead of borrowing an instant from a neighboring date.
- Timezone identifiers use the IANA database represented by the pinned runtime.
- Place results expose country, region and coordinates to avoid same-name guessing.
- Unknown-time calculations enumerate the valid instants belonging to the selected local date.
Boundary and uncertainty tests
Sign classification uses the unrounded longitude before degrees and minutes are formatted. Tests cover values immediately below, at and above every 30-degree sign boundary, including normalization around 0 and 360 degrees.
Approximate-time tests evaluate the complete declared plus-or-minus window. Unknown-time tests cover the complete valid local civil date. When sampled positions cross a sign boundary, the transition is refined to one second and every possible sign is reported.
Supported range and known limits
The verified date range begins on 1 January 1900 and ends at the bounded current date. Rising is supported from 66 degrees south through 66 degrees north inclusive. Outside the verified range, the relevant calculation stops with an explicit message rather than producing an untested result.
The project calculates Western tropical sign placements and one Ascendant convention. It does not validate houses, aspects, nodes, chart wheels, rectification, compatibility or predictions because those features are not part of the product.
Version and regression control
Astronomy, timezone-boundary, Temporal and place-data versions are pinned. A dependency, convention or data update is treated as a calculation change even when the visible layout is unchanged. Affected golden fixtures and boundary tests must pass before publication.
The calculator result audit exposes the engine version, zodiac system, time-confidence path, timezone decisions, sampling cadence and unrounded exact longitude when one instant was selected. This helps reproduce a disagreement instead of asking users to accept a hidden result.
Download the machine-readable summary
The CSV below contains the public comparison totals, tolerance, production engine, independent reference and scope notes used in this report. It contains no personal birth data or private test records.
- Download validation summary CSV
Public machine-readable summary for release v2.3.
What this evidence does not prove
The report shows that the software follows its stated calculation convention within the declared test limits. It cannot certify the historical accuracy of a user's birth record, guarantee that every external calculator uses the same convention or establish astrology as a scientific method.
Astrology is not scientifically established. Interpretations remain a separate, clearly labelled editorial layer for entertainment and self-reflection.