✦ Chart Literacy

How we test our engine against a century of sky

Before this calculator went live, its output was checked against an independent astronomy library across twenty birth charts spanning more than a hundred years — here's exactly how, and where the small remaining error lives.

The short answer

Every placement this site computes traces back to a vendored, MIT-licensed ephemeris engine derived from JPL planetary data — the same class of source professional astronomy software uses. Before launch, its output was checked against a fully independent astronomy library across twenty birth charts spanning over a hundred years, and the worst disagreement found across the whole set was under a hundredth of a degree.

A birth chart calculator is, underneath the wheel graphics and the sign names, a piece of applied astronomy: it converts a date, time, and place into a set of planetary positions. Getting that conversion right is a solved problem in the sense that professional planetary theory has existed for decades — but it's also a problem where a small, silent bug (a wrong sign convention, a missed time-zone correction, an outdated reference epoch) can quietly shift every result by degrees without anyone noticing until a chart looks wrong to someone who already knows their own placements. This page describes how we checked for exactly that.

The engine underneath

Every placement on this site is computed by a vendored copy of Astronomy Engine — a single-file, MIT-licensed library built on JPL-derived planetary theory, the same category of source that drives observatory-grade software. It's pinned to a specific version and checksummed in the codebase, so the exact math being run today is the exact math that was verified, not something that's silently drifted since.

The golden-matrix check

Before the calculator shipped, its output was run against a completely independent astronomy library — a second, unrelated implementation of the same underlying planetary theory — across twenty real birth charts chosen to stress-test the edge cases that break naive implementations: dates from the early 1920s through the present day, both northern and southern hemispheres, wartime clock-shift rules, half-hour time zones, and leap days. Every sign, every retrograde flag, every rising sign, and every midheaven had to agree between the two independent computations before anything went live.

The worst disagreement found across that entire set was under a hundredth of a degree of ecliptic longitude — small enough that it has never once changed which sign a placement fell in, across any of the twenty test charts.

A technical choice most tools get wrong quietly

One specific decision matters more than people expect: which reference frame the tropical zodiac is measured against. The correct astrological convention measures sign positions against the equinox of the birth date itself — not a single fixed reference point in time. Using a fixed reference point instead (a shortcut some tools take) introduces an error that grows every decade, small enough to miss for a recent birth date and large enough to flip a cusp-date sign for an older one. This engine uses the date-of-birth convention throughout, verified to agree with the reference computation to about a thousandth of a degree.

What runs where

None of this computation happens on a server. The engine runs entirely in your browser — the same code, the same math, whether it's drawing a chart for a date from decades ago or one from this afternoon. Nothing you type is transmitted anywhere in the process, which is a separate claim from the accuracy claims above but one this page can point to just as directly: open your browser's network panel while the calculator runs, and there's nothing to see leave.

The edges, honestly

Twenty test charts is a real check, not an exhaustive one — it can't catch every possible edge case across every possible date and location on Earth, particularly historical ones with unusual local time conventions. What it does establish is that the engine handles the categories of edge case most likely to break a naive implementation, checked against an answer computed a completely independent way. If you ever find a placement here that disagrees with a source you trust, the methodology page has the full detail on how to check it — and we mean that as a real invitation, not a formality.

Check it yourself

Read the full methodology page — every technical claim there names the specific check behind it.

Read the full methodology

Sources & method

  • Astronomy Engine (Don Cross) — MIT license, JPL-derived planetary theory
  • Independent astropy-based golden-matrix verification, 20 charts, 1923–present

Written and kept by Chart & Moon — the house byline; no fictional authors, ever.