Two people can enter the same birth details into two Chinese astrology calculators and get different charts back. Neither tool is broken. They are applying different answers to two questions the tradition has argued about for centuries — when a day begins, and what to do with a leap month — and almost nothing tells the user which answer is being used.
The first question: when does a day start?
The traditional day is divided into twelve two-hour blocks, and the first of them, 子, straddles midnight — it runs from 23:00 to 01:00. That creates an obvious problem. If a block spans midnight, does the day change at 23:00 when the block begins, or at midnight in the middle of it?
Both answers are held, and the choice has a large effect on the resulting chart. A birth between 23:00 and midnight is assigned to one day under the first answer and the previous day under the second. The day pillar differs by one position in the sexagenary cycle, and because the day stem is the reference point from which much of the rest of a reading is derived, a great deal of the chart moves with it. The year and month pillars do not, being fixed by solar terms rather than by the day count.
This is old, and it is not a modern muddle
It is sometimes said that the split is a recent invention, produced by squeezing a twelve-block day into a twenty-four-hour clock. The texts say otherwise.
Zhao Youqin's Gexiang Xinshu, from around 1300, states it directly: the first half of the 子 block, before midnight, belongs to the previous day, and the second half, after midnight, belongs to the current one. That is the split, described seven hundred years ago — and in a work of astronomy and mathematics rather than a divination manual.
The argument is older still. The astronomical treatise of the New Book of Tang records a seventh-century dispute over whether the day begins at the start of the 子 block or at its midpoint, with the court astronomer arguing for the midpoint against a colleague who had used the start.
And the convention moved. Work published in an astronomy journal on ancient Chinese and Japanese timekeeping records that the first double hour began at 23:00 in earlier practice, and at midnight from the Song dynasty. This is not a matter of one correct answer and one error. It is a convention that has changed, with both its older and newer forms surviving in use.
One modern argument that does not hold
A claim in circulation holds that the seventh-century passage is being misread — that the term rendered "midpoint of 子" actually refers to the winter solstice, so the whole textual case for the midnight boundary collapses.
It does not survive the passage. The term is set directly against "the start of 子" in the same sentence, and elsewhere it is paired with a term for celestial degree in a construction where both are plainly time-of-day measures. Whatever else is arguable here, that particular reading is not available.
The second question: what is a leap month?
The lunisolar calendar inserts a thirteenth month every few years. It is not given a number of its own — it is a second instance of the month before it. Which raises the question of what a chart should do with a birth inside one.
The principal classical manual of Purple Star astrology answers plainly. Its section on establishing the body and life palaces states that someone born in a leap first month should have those palaces established within the second month, and that all leap months follow this example. Leap month counts as the following month. This is the only place in the transmitted text that addresses the question.
Other texts disagree. A Republican-era manual gives a split — before the fifteenth day, treat it as the preceding month; from the fifteenth, as the following one — while also offering the "count it as the next month" rule as an alternative in the same book. A Qing imperial compilation splits it too, but at a different place: not the fifteenth day but the solar-term crossing that falls within the month.
So there is disagreement on whether to split the month, and for those who do, a second disagreement on where to place the split.
What this means if you are using a calculator
Two practical things follow.
If your birth time falls between 23:00 and midnight, the day convention applies to you, and it is worth finding out which answer your tool uses before comparing its output with anything else — including ours, since our bazi chart calculator and Zi Wei Dou Shu chart each take a position here too. A difference between two tools is not evidence that one is wrong.
If you were born in a leap month — roughly one year in three has one, so this is not rare — the same applies to the month convention. And be aware that a tool may take a position without stating it, which is the common case rather than the exception.
The question of which convention is correct has been open for centuries and is not settled by arithmetic. What we can do is point to which historical texts say what.
What this guide is not doing
Everything above is about how charts are constructed — which day, which month, under which rule. It says nothing about what a chart means, whether one convention yields better readings than another, or whether any of it predicts anything. Those are questions about the practice, and this is a description of a documented disagreement within it.
We have also deliberately not attributed positions to living practitioners. Some comparisons name living practitioners and attribute methods to them, often based on forum posts or blogs. The disagreement is real and textually documented, but we are not going to attach names to positions on that kind of sourcing. The point stands without it.
Where this comes from, and what will date
The Yuan-dynasty and Tang passages were read from digitised editions of the texts themselves. The timekeeping history is from a paper in a peer-reviewed astronomy journal. The leap-month rule of the principal Purple Star manual was read from the transmitted text; the Republican and Qing positions come from a calendrical specialist quoting them, and have not been verified against digitised originals, which is stated here rather than glossed.
The Hong Kong Observatory publishes the authoritative conversion tables for all this, and while it defines the 子 block as 23:00 to 01:00, it says nothing about where the day boundary falls. The astronomical authority declines to adjudicate, a point worth remembering if anyone claims it has.
None of this dates. The texts are fixed and the disagreement is centuries old. What changes is which convention a given piece of software has implemented, and that is worth re-checking rather than assuming.
