A BaZi chart is more than eight characters—they're birth time in another notation
Eight characters are only the four pillars’ stems and branches—not a full BaZi chart. This article maps three layers by field: casting conventions (zi-hour day change, true solar time, luck direction/start), natal structure (hidden stems, ten gods, element counts, interactions), and timing (luck-pillar table, annual cycles). It compares birthday-only vs eight-characters-only vs full packet inputs, and why ChatGPT invents ten gods and luck pillars when fields are missing.
Four stem-branch lines are often mistaken for the whole chart. Below: three layers by field—casting conventions, natal structure (hidden stems, ten gods, elements, interactions), luck pillars and annual cycles—and what birthday-only, eight-characters-only, and full-packet inputs each leave out.
Where the eight characters sit
BaZi is four pillars—year, month, day, hour—each with one stem and one branch. Charting runs calendar rules on birth time to get those four pairs, then exports ten gods, luck pillars, and the rest. Eight characters are step one, not the deliverable.
The eight characters do not include (must be computed separately):
- Ten-god tables, element counts, stem-branch interactions
- Luck-pillar sequence and annual cycles
- Stated conventions: day-change rule, true solar time, luck start
A full BaZi chart is usually three layers
A. Casting settings (conventions) B. Natal: pillars and derived structure C. Timing: luck pillars, annual cycles (optional monthly layers)
A. Casting settings
| Setting | What it affects |
|---|---|
| Day-change rule | Especially zi hour—when the day pillar turns |
| Month pillar | Often solar terms, not lunar calendar day 1 |
| True solar / longitude | Clock time vs local solar time—hour pillar |
| Luck direction & start | Gender + year stem polarity; start age from solar terms |
| Unknown birth hour | Approximate flag, default hour handling |
Without stated conventions you cannot tell calculation error from legitimate convention difference.
B. Natal structure
| Block | Contents | Role |
|---|---|---|
| Pillar detail | Stem and branch per pillar | Skeleton |
| Hidden stems | Stems inside branches | Base for ten gods and counts |
| Day Master | Day stem | Reference point for “self” |
| Ten gods | Labels vs Day Master | Structural relations—not preset judgments |
| Element counts | Metal/Wood/Water/Fire/Earth tallies | Counting result; state the method |
| Interactions | Combine, clash, punish, harm, break | Relations between pillars |
| Shen sha (if used) | Rule-matched tags | Optional; schools differ |
Ten-god counts may count stems only or weight hidden stems—the packet should say which.
C. Timing layers
| Block | Contents |
|---|---|
| Luck pillars | ~10-year steps; start age/date; year ranges; ten god per step |
| Annual cycles | Stem-branch of a calendar year on natal + current luck step |
| Monthly (optional) | Month pillars by solar terms—for deeper timing questions |
Natal pillars are stable for a fixed birth time and convention. Annual cycles change every year—they are not embedded in eight characters. “How is this year?” without annual and luck fields lacks data for period-level discussion.
Three kinds of input
| Input | What you actually give | Usually missing | How far it goes |
|---|---|---|---|
| A birthday sentence | Natural-language time | Stems/branches, conventions, structure, timing | Vague talk or guessed pillars |
| Eight characters | Four-pillar encoding | Hidden stems, ten gods, interactions, luck table, annual cycles, conventions | Pillar check at best—not luck/year questions |
| Full chart text | Conventions + natal + timing | — | Natal themes and phase-level questions |
Eight characters save one guess versus a birthday—but ten gods, luck pillars, and annual cycles were not included. Something still has to compute them—or invent them.
Why chat models fill in the chart
1. A language model is not a calendar engine
Charting needs fixed tables and formulas. Models excel at fluent text, not identical pillar output every time. Same birthday, different replies—see our accuracy comparison.
2. Eight characters are too thin
“Wealth vs seal this year” needs Day Master, ten gods, luck sequence, and the target annual cycle. Eight glyphs cover pillars only; empty fields get filled from training patterns.
3. Unknown conventions → hidden defaults
Unstated zi-hour or true-solar rules mean the model picks defaults. Mismatch with your app may be convention drift or error—full text with stated rules tells them apart. See true solar time.
4. Long chats lack one fact source
A luck table invented in turn one may shift in turn three without a single fixed packet to cite.
What's missing is a checkable fact table—not a longer prompt.
Minimal contrasts (fictional)
Eight characters only
Geng-Wu Ji-Mao Jia-Zi Ding-Mao
Readable: four pillars. Not readable without further work: ten gods, hidden stems, luck start, this year's annual cycle, true solar flag. Ask “career this year” and the model must assume luck and annual layers.
Eight characters + two convention lines (still incomplete)
Geng-Wu Ji-Mao Jia-Zi Ding-Mao Day change: zi start · Luck: forward · Longitude correction: yes
Conventions added—still no luck table, still no annual cycle for “this year.”
Typical full-packet sections (outline only)
[Rules & conventions] [Birth & cast] [Four pillars] [Day Master] [Element counts] [Ten-god counts] [Stem-branch interactions] [Luck pillars] (Annual cycle appendix when asking about a given year)
A full cast is multi-section text, not four pillars. Shape of fielded text: AI-readable chart packet.
Common mistakes
| Mistake | Correction |
|---|---|
| BaZi charting = eight characters | Eight characters are step one; structure and timing tables follow |
| Eight characters to AI = chart cast | You supplied pillar codes; ten gods and luck may still be generated |
| “This year” needs no annual cycle | Annual cycles are timing data, not natal eight characters |
| One hour off doesn’t matter | Hour pillar can flip; zi-hour day change too |
| Different sites = someone is wrong | Compare conventions first, then calculation |
Before handing BaZi to AI
Minimum for natal themes:
- Four pillars
- Hidden stems and ten gods (or equivalent relation table)
- Stated casting conventions
Add for phase or year questions:
- Luck-pillar table (including current step)
- Target year annual cycle
One conversation, one packet; ask the model to reason only from listed fields and not re-cast pillars. Steps and prompt samples: BaZi charting tutorial for AI.
Scope
BaZi packets are structural data—stems, relations, timing sequences—not preset lucky/unlucky readings. Useful god, pattern schools, and event timing belong to interpretation layers.
This article is four-pillar charting (birth-time input) only—not divination systems keyed to a cast moment.
Is BaZi charting just eight characters?
If I paste eight characters into ChatGPT, is the chart already cast?
What does a BaZi chart usually contain?
What are ten gods, and how do they relate to the eight characters?
What are luck pillars and annual cycles?
Why do BaZi charts differ across websites?
Checkable full text requires a charting engine first—then paste into AI.