BaZi conventions

Put one birth moment into two BaZi calculators and you can get two different charts. Sometimes a different day master, the character the whole reading is built on. That is not a bug in one of them. Four conventions sit underneath every BaZi chart, each with real history behind it, and calculators apply them silently and differently.

Below is our answer to each of the four. They are declared in the disclaimer of every response, which makes changing one a breaking change rather than a refactor.

1. Where the year starts: Li Chun or Chinese New Year

Our answer: Li Chun (立春), the exact instant.

BaZi is a solar system living inside a lunisolar calendar. Its month pillars key to the twenty-four solar terms, which are positions of the Sun rather than phases of the Moon, and the year follows the same logic: it turns at Li Chun, around 4 February. Chinese New Year is a different animal: the first new moon of the lunar year, landing anywhere from late January to late February.

A birthday in the gap belongs to one animal by one rule and another by the other. Neither is a mistake. We take Li Chun, and the instant comes from the ephemeris rather than from the date “4 February”.

2. Where the day starts: 23:00 or midnight

Our answer: 23:00 on the birthplace clock.

The traditional day is twelve double-hours, and the first of them, the zi hour, runs from 23:00 to 01:00. Every other double-hour sits tidily inside one calendar day. The zi hour straddles midnight, and that single structural fact is the whole cause of the dispute: if the zi hour opens the day, a birth at 23:30 already belongs to tomorrow.

This is the axis that moves the day master. Move the day pillar and you have not adjusted the chart, you have replaced it.

3. Whether the clock is corrected for the birth longitude

Our answer: on request, and only for the day and the hour.

For most of this system’s history the question could not arise: local time was solar time. Standard zones are a railway-age invention, and once a country runs on one clock, that clock matches the Sun on a single meridian only. China is the extreme case, one zone across some sixty degrees, so a birth in Kashgar sits nearly three hours from solar noon on a Beijing clock.

curl -X POST https://api.astroway.info/v1/chinese/true-solar-time \
  -H "X-Api-Key: aw_live_..." \
  -H "Content-Type: application/json" \
  -d '{ "date": "1995-06-15", "time": "09:20:00", "timezoneOffset": 8, "longitude": 75.9 }'

The correction reaches the day and the hour, and stops there

Li Chun and the twelve 節 are instants on the world clock: the Sun reaches 315° once, and every birth on Earth is either before it or after it. A sundial reading states where the observer stands, not when that instant fell. Feeding a corrected clock into the year and month boundaries therefore asks the wrong question of them.

The cost of getting this wrong is concrete: a Beijing birth at 23:00 on 4 February 1988, seventeen minutes after Li Chun, corrects back to 22:31 and would be handed the previous solar year, the previous month, a different day master and a different set of ten gods. We made that mistake and fixed it before release.

4. What to do about daylight saving

Our answer: only the caller knows.

timezoneOffset is the offset the clock was actually running on at the birth moment. A 1988 Beijing birth ran on UTC+9, not +8, because China kept summer time from 1986 to 1991. No arithmetic on our side recovers that from the date alone, so we do not guess.

A run against an external dataset

Jade Almanac published a CC0 comparison of eight implementations across thirteen probes built specifically on these four axes: jadealmanac.com/bazi-calculator/conventions. The dataset measures conventions, not correctness, and says so itself.

We ran POST /v1/bazi/four-pillars through all thirteen. The result reproduces from a script in the repository and is frozen by a test.

Probe	What it isolates	Our answer
1985-02-10	year boundary	Yi-Chou, because Li Chun has passed
1990-06-15 23:30	day rollover	day Ren-Zi: the day has already turned
1988-02-04 23:00	Li Chun and 23:00 together	Wu-Chen Jia-Yin Geng-Yin Bing-Zi
1995-06-15 09:20 Kashgar	true solar time	uncorrected Yi-Si, corrected Gui-Mao

One honest note about whom we agree with. The dataset’s publisher writes that its own engine rides on tyme4ts, and lunar-javascript shares that author. So the top three rows of the table are one lineage, not three independent votes. Among the rows independent of it, feng-shui-institute.org has full coverage, and since true solar time shipped we reproduce its answer on all three probes where it applies the correction.

That leaves exactly one genuine school disagreement between us: the 23:00 day rollover.