How Meridian casts a chart
Chart mathematics is code, never a language model. The same birth data returns the same chart every time, and every convention that could reasonably differ is chosen explicitly and recorded on the chart.
Deterministic by construction
Eight separate engines compute the charts: BaZi, Zi Wei Dou Shu, Vedic, Western, Human Design, numerology, Qi Men Dun Jia and Da Liu Ren. Each is ordinary code. None of them asks a language model anything. Astronomical positions come from a full ephemeris, and Chinese calendar conversions from proper calendrical computation rather than the approximations most tools use.
This matters for a reason that is easy to miss. A language model asked to compute a chart will produce something chart-shaped every time, confidently, and it will differ between runs. It is generating plausible output, not calculating. Deterministic code either returns the right answer or fails loudly.
Tested against published references
Each engine is checked against charts with known correct answers, taken from published sources in its own tradition. The suite runs on every change. A change that would alter an existing chart fails the build rather than silently shifting people's readings.
That last part is the reason to bother. Most chart software has no way to tell the difference between a fix and a regression, so corrections quietly change what earlier users were told.
The conventions we chose
| Question | Meridian's rule |
|---|---|
| Clock time or solar time | True solar time, from birth longitude and the equation of time |
| Day boundary | 23:00, keeping the 子 hour whole |
| Historical daylight saving | Resolved for the birth date and country before conversion |
| Unknown birth time | Three pillars computed, hour left empty, never defaulted |
| Unknown birth place | Chart marked as built on a fallback latitude, and angle-dependent readings dropped |
| A method that cannot be computed | Named and reported, never substituted |
Where it can still be wrong
Worth stating, because a page that lists only strengths is advertising.
- A chart is only as good as the birth data. An hour of error in a remembered time changes the hour pillar, and no amount of rigour downstream repairs that.
- Historical time zone data is very good but not perfect for every country and year.
- Where traditions disagree we picked a convention. Picking one does not make the others wrong.
- Interpretation is bounded by what the sources actually say. Where a tradition is silent, we say so rather than inventing a reading.
Related
Cast a chart and inspect every convention used
Worked example computed by Meridian's chart engines on 1 August 2026. The engines are deterministic, so the same input returns the same chart.
This page explains calculation conventions. It does not interpret a chart and is not advice.