The four pillars are supposed to be arithmetic. Year, month, day and hour each get one heavenly stem and one earthly branch, and given a birth moment there should be exactly one right answer. So the first time you see two apps disagree, it looks like one of them has a bug.
Almost always, it doesn't. The disagreement comes from the calendar, not from the astrology. Below are the six rules that account for nearly every mismatch, roughly in order of how many people they affect.
OneThe year does not start on January 1 — or on Lunar New Year
In four-pillar practice the year pillar changes at 立春 (Lìchūn), the solar term that begins spring. It falls around February 3–5, and it is an astronomical instant, not a date — it has an hour and a minute.
This surprises people twice. Once because January births belong to the previous year's pillar, and again because Lunar New Year — the date most people assume — is a different thing entirely and can land weeks away.
Who this hits: anyone born between January 1 and roughly February 5. If you were born in that window and two apps disagree about your animal sign, this is almost certainly why.TwoMonths begin at solar terms, not on the 1st
The month pillar does not follow calendar months, and it does not follow lunar months either. It follows the twelve 節 (jié) solar terms — Lìchūn, Jīngzhé, Qīngmíng, and so on. Each one is an astronomical instant, and the month pillar advances the moment it passes.
These instants drift by a day or so from year to year, so a calculator that hard-codes "the month changes on the 4th" will be right most years and wrong in the others. A calculator that uses lunar months will be wrong far more often.
Who this hits: anyone born within a day or two of a solar term — that is roughly one birth in fifteen. The month pillar is the one most often reported as "the app changed my chart."ThreeThe hour pillar's stem depends on the day's stem
The hour branch is the easy half: two-hour blocks, 子 for 23:00–01:00, 丑 for 01:00–03:00, and so on. Many people assume the stem works the same way — a fixed table from clock time.
It doesn't. The hour stem is derived from the day stem, by the rule traditionally called 五鼠遁 (the five-rat method). Two people born at exactly 09:30, on different days, generally do not share an hour stem.
This is the single most common implementation shortcut. It is also invisible: the chart looks complete and plausible, so nobody notices until they compare against another tool.
Who this hits: everyone, if the app took the shortcut. Half your hour pillar is simply wrong.FourClock time is not solar time
Time zones are political. Korea runs on UTC+9, which is the solar time of the 135°E meridian — a line that passes through Japan. Seoul sits near 127°E. That gap is about 32 minutes: when the clock in Seoul says noon, the sun gets there roughly half an hour later.
Some calculators apply this longitude correction, some don't, and a few let you choose. All three are positions people defend in practice.
Who this hits: anyone born within ~30 minutes of a two-hour boundary — near 01:00, 03:00, 05:00 and so on. For them the correction moves the hour pillar by a whole branch.FiveThe 23:00 problem is a real disagreement, not a bug
The 子 hour spans 23:00 to 01:00 — so it straddles midnight. If you were born at 23:30, does your day pillar belong to that day, or to the next one?
Schools genuinely differ here. One tradition advances the day pillar at 23:00 along with the hour; another keeps the day pillar until midnight. The distinction is old enough to have its own names (早子時 / 夜子時).
This one is worth calling out separately, because it is the only item on this list where "the other app is wrong" is not a fair thing to say. Both answers have a lineage behind them.
Who this hits: births between 23:00 and midnight — and it changes the day pillar, which is the pillar most readings lean on hardest.SixKorea had daylight saving time, and people forget
Korea observed DST in several stretches last century — around 1948–1951, 1955–1960, and 1987–1988. Births inside those windows are recorded in clock time that runs an hour ahead of standard time.
An app that ignores DST shifts those births by a full hour. Combined with rule four, a birth near a boundary can land two branches away from where it belongs.
Who this hits: anyone born in those windows — and, if the family recorded the time from memory rather than a document, possibly no one can be sure at all.A practical test you can run in five minutes
You don't need to trust any single app. Pick a birth moment where the rules above are maximally annoying and see how a calculator behaves:
- Enter a birth date in late January. If the app gives you that calendar year's animal instead of the previous year's, it is ignoring rule one.
- Keep everything identical and change only the date by one day. If the hour stem never changes across different days, it took the rule-three shortcut.
- Enter a time like 23:30 and then 00:30. If the day pillar is identical for both, check the app's documentation — it has made a choice on rule five and should say which.
- Enter 05:10 with and without any "true solar time" option the app offers. If the hour branch doesn't move, the option isn't doing anything.
An app that handles all four is not necessarily right about your reading. But it is doing the arithmetic it claims to be doing, which is the part that can actually be checked.
The short version
| Rule | What splits | Pillar affected |
|---|---|---|
| Lìchūn year start | Jan 1 vs Lunar New Year vs Lìchūn | Year |
| Solar-term months | Calendar / lunar / true solar terms | Month |
| Five-rat hour stem | Derived from day stem, or hard-coded | Hour |
| Longitude correction | ~32 min in Korea, applied or not | Hour |
| 23:00 day boundary | Genuine doctrinal split | Day |
| Historic DST | 1948–51, 1955–60, 1987–88 | Hour |
None of this makes four-pillar astrology more or less true. It just means that when two apps disagree, the interesting question is not "which one knows my fate" but "which one is using the right calendar" — and that question has a checkable answer.
What Saju Today does
This page comes from the studio that makes one of these apps, so it would be unfair not to state its own answers to all six. Hold it to the same test:
| Year start | Lìchūn, computed as an instant from solar longitude — not a fixed date. |
| Months | The twelve solar terms, each located per year by search, not from a lookup table. |
| Hour stem | Derived from the day stem (five-rat rule). |
| Solar time | Applied to the hour pillar by default (mean solar time from your longitude). |
| 23:00 | Midnight boundary by default — the day pillar does not advance at 23:00. The 23:00 convention exists in the engine as a separate named preset; the app ships the midnight one. |
| Historic DST | The birth offset is resolved from the IANA time-zone database at your birth instant, so Korea's 1948–51 / 1955–60 / 1987–88 periods are carried automatically. |
One honest caveat on the last row: charts saved before the app supported time zones fall back to a flat UTC+9, which is correct for most Korean births but does not carry DST. Re-entering the birth details fixes it.
The pillars themselves are computed in your browser, in plain JavaScript — producing them needs no server round-trip. It is a free daily reading, so you can check your own four pillars against any other tool and see for yourself.
Open Saju Today →Saju Today is an entertainment app. Every reading in it is generated by an AI language model (Google Gemini) applying traditional Saju concepts — for reflection and entertainment, not a prediction of the future, and not medical, psychological, legal, or financial advice. This page describes calendar and time conventions used in four-pillar practice; it is not a claim that those conventions predict anything.