Accuracy
Audited. Honestly.
Worst-case residuals against JPL Horizons across nine reference dates spanning 1899–2101. Sub-arcsec on every body except the geometric Moon — that residual is the published MPP02-vs-DE441 ceiling, and we say so out loud below.
Worst-case residuals
Body by body.
Vs. JPL Horizons.
| Body | Worst |dλ| | Method |
|---|---|---|
| Sun | <0.1″ | VSOP87D Earth heliocentric + FK5 + IAU 1980 nutation + aberration |
| Moon (apparent) | 0.87″ | Home-grown ELP/MPP02 (Chapront & Francou 2003) full 35,901-term series |
| Moon (geometric) | 21″ | Same — DE405/406-fit ceiling vs DE441 (see ceiling doc) |
| Mercury | 0.10″ | VSOP87D |
| Venus | 0.11″ | VSOP87D |
| Mars | 0.10″ | VSOP87D |
| Jupiter | 0.22″ | VSOP87D |
| Saturn | 0.46″ | VSOP87D |
| Uranus | 1.1″ | VSOP87D |
| Neptune | 2.4″ | VSOP87D |
| Pluto | 0.04″ | DOPRI8 8(7) n-body, 9-perturber, JPL Horizons J2000 seed |
| Ceres | 1.5″ | Same integrator |
| Pallas | 2.84″ | Same integrator |
| Juno | 0.46″ | Same integrator |
| Vesta | 2.8″ | Same integrator |
| Chiron | 0.06″ | Same integrator |
| Pholus | 0.05″ | Same integrator |
| Nessus | 0.08″ | Same integrator |
| Eris | 0.09″ | Same integrator |
| Haumea | 0.05″ | Same integrator |
| Makemake | 0.05″ | Same integrator |
| Sedna | 0.10″ | Same integrator |
| Quaoar | 0.04″ | Same integrator |
Live data: /api/v1/accuracy serves the same residuals as JSON for any tooling that wants to verify on every release. Engine version: 9.13.0.
Honest disclosure
The Moon ceiling.
The 21″ geometric-Moon astrometric residual is the DE405/406-fit ceiling shared by every analytical lunar theory in the world that traces back to MPP02 or LEA-406. DE441 (the modern reference) shifted the lunar trajectory by ~10–21″ depending on epoch — no analytical theory fitted to DE-405 can close that gap. We tested the two paths that close it (parameter refit, alternative theory) and shipped the negative result instead of glossing over it.
Production callers hit moonApparentEcliptic (apparent mode), where the Moon is sub-arcsec because the engine's nutation is added correctly. The 21″ only surfaces if you specifically ask for the geometric mean position. The full write-up is at docs/MOON_CEILING.md — including the two paths that close the gap (DE441 binary refit; numerical Moon integration) for any customer who needs sub-arcsec geometric Moon enough to pay for the multi-week engineering it takes.
Source of truth
No black box.
Every coefficient is cited.
IMCCE — Bureau des longitudes (Paris)
Chapront & Francou 2003, A&A 404, 735 — DE405/406 fit
Chapront-Touzé & Chapront 1988 — alternative Moon evaluator
NASA/JPL — J2000 state vectors for every integrated body
Wahr 1981, Seidelmann 1982 — top 15 + extension terms
Willmann-Bell 1998 — every classical chapter
JPL SSD — outer-span fallback on integrated bodies
Doctrine attributions live in SOURCES.md — pre-1929 public-domain only, every traditional rule traced to its primary source (Ptolemy, Vettius Valens, Brihat Parashara Hora Shastra, etc.). No copyrighted modern interpretations are baked in.