Bazi Compatibility Method: What Is Calculated and Checked
How AstralNex compares two Bazi charts: cross-chart stem and branch relations, Ten God relations, luck-cycle timing, a doctrine-weighted score with its uncertainty, and the checks every paid report must pass.
Overview
Each person’s Four Pillars are calculated first. The compatibility engine then compares the two charts and keeps every fact it used. The written report explains those facts; it does not create new chart facts or a new score.
What is calculated
- Relations between the two charts’ stems and branches: six harmonies, three harmonies and half combinations, clashes, punishments, harms and breaks, and stem combinations and clashes, each with the pillars involved.
- How each Day Master relates to the other as a Ten God, in both directions.
- Spouse palace (day branch) components, the favourable and unfavourable elements of each chart, and which elements one chart supplies to the other.
- Timing windows where both people’s luck cycles or annual pillars trigger the same relations.
- An overall score with support, friction, timing and Day Master bridge components, a confidence value, the reasons for uncertainty, and a score range.
- A supplementary layer (shared symbolic stars, Na Yin, zodiac animals, spouse star movement, luck-cycle alignment) that is shown for context and never enters the score.
Fixed conventions and inputs
- The day boundary (early or late Zi hour) and true solar time are applied per person and change the pillars used in the comparison.
- Scoring follows one fixed rule set (Ziping core, cross-chart version 2); it is not user-selectable.
- When a birth time is uncertain, alternative charts are flagged and the confidence value drops instead of the result being presented as exact.
How the report is written and checked
- The model receives only the calculated facts and writes the explanation; the score shown in the report comes from the engine.
- Before saving, automated checks look for possible luck-cycle attribution mismatches, branch relations outside the calculated evidence, and invented Na Yin. These prose findings are recorded as quality observations, not reasons to reject the report.
- Residual numbers written as scores are replaced with qualitative bands and recorded as quality findings.
- Each year keeps the engine’s own label (supportive, mixed or defensive): a defensive year cannot be headed, written or advised as a good year for major commitments, a personality trait only one of you carries cannot be written as shared, and English relation names must match the relation calculated for that pair.
- Automated checks compare generated wording with the supplied chart facts and record possible mismatches as quality observations, not accuracy certification. Wording, scope and section-coverage findings do not by themselves reject, remove sentences from, or regenerate a readable report. Empty or malformed output, provider-confirmed truncation and private-prompt leakage still prevent delivery. Authorization and credit settlement remain separate protections.
Publisher verification results
A separately written replay recalculates both people's Four Pillars and every cross-chart branch relation (combinations, clashes, harms, punishments, destructions), stem relation and the day-master relation for 240 seeded synthetic pairs, including early/late 子時 and solar-term boundary births.
- Pairs compared
- 240
- Identical under declared rules
- 240
- Disclosed school differences
- 5
Under the method's declared rules, 240 of 240 pairs match on every compared fact. The record also lists 5 places where other schools read the same chart differently: for example, schools that do not recognise 六破 would drop destruction links in 166 pairs. Those variants are disclosed, not counted as errors.
The pillars use the same open-source calendar library as the method (lunar-javascript); the interaction rules are written independently. True solar time and daylight saving are covered by the single-chart Bazi replay. This checks rule application, observed on 2026-09-27, not the compatibility score, useful-element judgments or any interpretation. the method performed this comparison. The downloadable record is a summary; implementation and full execution fixtures remain private. This is not third-party certification or a public replay kit.
Concepts in the AstralNex Wiki
Before you buy
- Both birth records are confirmed on a free preview before anything is charged.
- A complete report uses 65 credits. The credits are reserved when generation starts and returned automatically if the report cannot be completed.
- The saved report keeps the calculated facts it was written from, so later questions in the report chat read the same facts.
Limits
- The score summarizes one doctrine’s weighting of chart relations. It is not a probability that a relationship will succeed.
- The report describes symbolic tendencies and timing; it cannot know either person’s actual behaviour, history or intentions.
- Birth-time errors can change the hour pillar and some relations; the report states when that uncertainty applies.