How we calculate
A Saju chart is arithmetic before it is interpretation. This page sets out the rules Bokki follows, where schools disagree and which side we took, how we checked the results, and what we can't promise.
Last reviewed 15 November 2026
The four pillars
Each pillar is a pair from two old counting cycles: ten heavenly stems and twelve earthly branches. Together they make a cycle of sixty.
- Year pillar. The Saju year begins at Lìchūn (立春, “start of spring”), when the sun reaches 315° of ecliptic longitude — around 4 February. We use the instant itself, so two people born the same day can fall in different years if one was born before it and one after.
- Month pillar. Each month begins at one of twelve solar terms (the 節), again at the instant the sun reaches its longitude. The month stem follows from the year stem by a fixed rule.
- Day pillar. Days have been counted on the sixty-day cycle without interruption. We compute it from the Julian Day Number; 1 January 2000 is 戊午. With a birthplace the day turns at midnight by the sun there, otherwise at midnight by the clock.
- Hour pillar. The day is divided into twelve double-hours starting at 23:00. The hour stem follows from the day stem by a fixed rule.
Which time counts
The hour pillar belongs to the sun, not to the clock. Clock time is a legal convention: a whole time zone shares one time even though the sun crosses it over the course of an hour or more.
When you give a birth time and place, Bokki converts clock time to true solar time in three steps. First it looks up the clock offset actually in force at that place on that date — daylight saving, wartime clocks and old standard times included — from the IANA time zone database. Then it corrects for longitude, four minutes per degree. Then it adds the equation of time, the seasonal quarter-hour by which the real sun runs ahead of or behind the average one.
The day and hour pillars are read from that true solar time. The year and month pillars are different: a solar term is a single moment for the whole planet, so we compare it with the absolute instant of your birth rather than with local solar time.
If you give a time but no place, we use the clock time as it is and say so on your chart. If you give no time, the chart has three pillars and we leave out anything that needs the hour.
Births near midnight
Between 23:00 and midnight the tradition is split. Bokki follows the “late Zi hour” convention (야자시) — one of the common options in Korean almanacs and the default of the calendar library we build on: the day pillar changes at solar midnight (midnight by the clock when no birthplace is given), but the hour stem of the 23:00 hour is taken from the following day. Charts made under the other conventions will differ for births in that hour, and only in that hour.
Lunar birthdays
A lunar date is converted to the solar calendar first; the pillars are then calculated as usual. Korea and China compute the same lunar calendar from different meridians, so their dates disagree on some days and occasionally on which month is the leap month (2012 and 2017, for example). The form asks which calendar your family used.
For Korean lunar dates before 1912 the printed almanac of the time followed Chinese dates, and we have not been able to verify every date in that period. For 1912 onward our Korean dates match the published comparison we tested against.
What is read from the pillars
Everything after the eight characters is interpretation built on conventions, and the conventions vary. We state ours plainly.
- Five-element balance. Each stem counts once. Each branch is split among the stems hidden inside it, with the main one weighted most. There is no single standard for these weights; ours is a common teaching ratio.
- Strength of the day master. Judged from the season of birth, the roots among the branches and the support among the stems. No school publishes a universal formula; treat “strong” and “gentle” as a description, not a score.
- The helpful element. Chosen first by balance — support a lightly supported day master, drain an overly supported one — and then adjusted for season, so a midwinter chart is given warmth.
- Ten-year luck cycles. The direction depends on the year stem and gender, as the classical rule has it; the starting age comes from the distance in days between birth and the nearest solar term, three days to a year.
- Today and the luck clock. Each day and each double-hour has its own stem. We name the relationship between that stem's element and your day master, and mark it a tailwind when it is your helpful element.
How we checked
Bokki's engine is built on an open-source calendar library, with our own layer for time correction. To check it we wrote a second calculator that shares no code with the first: day pillars from the Julian Day Number, hour pillars from the stem rule, and solar terms from the sun's position using a published astronomical algorithm.
We compared the two on 240,000 random birth moments between 1900 and 2100, on every minute within two hours of each Lìchūn from 1950 to 2030, and on births within 45 minutes of every solar term from 1900 to 2100. Outside a band of 20 minutes around a solar term — the stated accuracy of the independent algorithm — there were no disagreements. Inside that band the comparison can't decide either way.
Against a higher-precision ephemeris, the minute at which our engine changes month was within about two minutes of the true instant. So a birth within a couple of minutes of a solar term can land on either side, here and in any almanac.
We also ran eighteen worked births collected from independent calculators, including Korean daylight-saving years and the 1954–1961 period when Korea kept UTC+8:30. Every pillar that does not depend on a choice of convention matched.
What we can't promise
A chart is only as exact as the birth time on record, and many recorded times are rounded. If your time is near the edge of a double-hour, read both hour pillars as possible.
The city list currently covers about 430 places — one or more per time zone. If yours is missing, pick the nearest or enter the longitude yourself; each degree of longitude is four minutes.
Dates before 1900 are not supported.
Saju is a traditional practice, not a science. An exact calculation doesn't make a reading a prediction.
Where traditions differ, and what Bokki does
If your chart here differs from one made elsewhere, the cause is almost always one of these.
| Question | Variants in use | Bokki |
|---|---|---|
| When the year changes | Lìchūn instant; 1 January; Lunar New Year; winter solstice | Lìchūn, at the exact instant |
| Time correction | None; a fixed −30 minutes for Korea; full true solar time | Full true solar time: longitude and equation of time |
| Daylight saving and old time zones | Ignored; converted back to standard time | Converted, using the offset in force that day |
| Births from 23:00 to midnight | Day changes at 23:00; day changes at midnight with next day's hour stem; day changes at midnight with the same day's hour stem | Midnight, with the next day's hour stem |
| Lunar calendar | Korean (135°E); Chinese (120°E) | You choose |
| Hidden stems and element weights | Tables differ between classical texts | Main hidden stem weighted most |
| Start of the luck cycles | Rounded to whole years; counted to the month | Three days to a year, as the calendar library computes it |
Sources
Historical clock offsets and daylight-saving rules, read through the browser's built-in copy.
The open-source calendar library the engine is built on: solar terms, the sexagenary cycle and luck cycles.
Jean Meeus, Astronomical Algorithms, 2nd ed., ch. 25
The solar-position algorithm used for the independent check.
Park, Mihn & Ahn, “Operation and calculation of the Korean lunar calendar”, PKAS 32:407–420 (2017)
Why Korean and Chinese lunar dates differ, with the list of differing dates we tested against.
Hong Kong Observatory, The 24 Solar Terms
Definition of the solar terms by the sun's ecliptic longitude.
The classical text quoted on the day-master cards. Editions differ in wording; we quote the edition transcribed on Wikisource.
Found a chart you think is wrong? Tell us the date, time and place and what you expected.