Methodology & Accuracy

Celestium

Last updated: July 6, 2026

An astrology app is only as honest as its astronomy. This page shows exactly how Celestium computes your chart, which traditions inform the interpretations, and the measured accuracy of every number we show you; verified, not asserted.

The Engine

Celestium computes every position on your device with its own astronomical engine. Nothing is fetched from a horoscope feed and nothing is estimated by an AI. The methods are the published, peer-reviewed standards of positional astronomy:

  • Sun, Mercury through Neptune: VSOP87 planetary theory (Bureau des Longitudes, Paris), with nutation and precession-to-date applied so positions sit in the tropical zodiac of the moment.
  • Moon: Jean Meeus' 60-term lunar series (Astronomical Algorithms, ch. 47).
  • Pluto: Meeus' dedicated periodic series, valid 1885 to 2099.
  • Lunar nodes: the true node (mean node plus periodic corrections), the count most working astrologers use.
  • Chiron: Keplerian orbital elements calibrated against Swiss Ephemeris J2000 values.
  • Houses and angles: sidereal-time-based Ascendant and Midheaven with historical timezone resolution for the birth moment, including DST edge cases.

Verified Accuracy

We diff the engine against Swiss Ephemeris, the reference standard used by professional astrology software, across a 13-date grid spanning 1900 to 2090. The full artifact (every date, every body, every delta) is regenerated with the test suite and ships in the open. Current maximum absolute differences across the entire 190-year grid:

BodyMax differenceIn plain terms
Sun0.0016°about 6 arcseconds
Moon0.017°about 1 arcminute
Mercury–Neptune≤ 0.015°under 1 arcminute
Pluto0.031°under 2 arcminutes
True Node0.097°under 6 arcminutes

For scale: a chart wheel drawn at poster size would place these differences well inside the width of the glyph's stroke. Chiron is verified separately against astro.com reference charts (within 0.5°).

Your Chart, Your Method

Celestium does not lock you into "our" house system or a single tradition. Placidus is the default; the choice stays yours, and your chart re-renders under the method that resonates with your practice. The same principle guides where the app is heading: astrology first and deep, with other calendrical wisdoms joining only when we can do them authentically.

The Interpretive Layer

Interpretations are composed from a curated text corpus grounded in the tradition (including public-domain professional sources), matched to the exact placements, dignities, and aspects your chart actually has. The composition is deterministic; what personalizes it is the astronomy, not a language model guessing. Optional AI synthesis exists as a separate, consent-gated layer that uses your own API key and always interprets engine-verified positions; it never calculates them.

Honest Limits

  • Pluto positions are computed for 1885–2099; outside that range the series degrades and we say so rather than extrapolate silently.
  • Houses require a birth time. If yours is unknown, house-dependent readings should be treated with care; first-class support for time-unknown charts is on our roadmap.
  • Astrology itself is a symbolic tradition, not a predictive science. We hold the astronomy to scientific standards and the interpretation to the standards of the tradition; the difference matters and we keep it visible.

← Back to Home