Saju calculation benchmark
A Saju chart is a calculation with one right answer — the same birth date, time and place must produce the same eight characters, no matter who computes them or when. That makes verification simple. Below are the trap types where Four Pillars tools most often slip, published as concrete cases with our engine's computed values. Copy any case's input prompt, paste it into whichever AI you use, and compare its answer with this table yourself.
AI answers vary by model and time — verify yourself. We don't grade other services or publish their results; this page publishes only our own computed values and how to check them.
How to verify — 4 steps
- Pick a case. Choose one trap type from the table below — solar-term cutoff, True Solar Time, daylight saving, the late Rat hour, or an overseas birth.
- Copy the prompt. Use the case's 'Copy this input' button. The prompt contains only the birth input — never the answer.
- Paste it into any AI. Paste it into whichever AI chatbot you use and ask for the four pillars and the starting age of the first luck pillar.
- Compare with the table. Compare the AI's answer with the engine values in this table. If they differ, read the convention note first — True Solar Time and late-Rat-hour conventions differ between schools, and different is not the same as wrong.
The benchmark table — 9 cases
The stem–branch pairs below are not hardcoded — they are what the Gwiraedang engine actually computed when this page was built. Agreement between the published values and the engine output is locked by automated golden tests.
| Input | Trap type | Year | Month | Day | Hour | Luck starts (age) |
|---|---|---|---|---|---|---|
| Mar 3, 1998 · 11:05 · Seoul · F | True Solar Time — hour-branch boundary | 戊寅 무인 | 甲寅 갑인 | 己酉 기유 | 己巳 기사 | 9 |
Seoul sits at longitude 126.98°E, about −32 minutes from the KST meridian — 11:05 on the clock is 10:33 in solar time, which lands in the 巳 (Sa) hour. Without the correction, raw clock time gives the next branch, 午 (庚午). If two tools disagree on the hour pillar, check the True Solar Time setting first. | ||||||
| Feb 4, 2012 · 19:23 · Seoul · M | Lichun (start of spring) cutoff — minute-level | 壬辰 임진 | 壬寅 임인 | 乙未 을미 | 乙酉 을유 | 10 |
In 2012, Lichun (입춘, the solar term that starts the Saju year) fell at 19:22 KST on February 4 — our own astronomical calculation, within ±2 minutes of the KASI published time. A birth at 19:23 is one minute after the cutoff, so the year pillar is 壬辰 rather than 辛卯, and the month pillar is 壬寅. Solar-term boundaries are drawn by the minute, not by the date. | ||||||
| Jan 17, 1994 · 15:41 · Incheon · F | True Solar Time correction + branch boundary | 癸酉 계유 | 乙丑 을축 | 癸卯 계묘 | 庚申 경신 | 6 |
Incheon (126.7°E) gets about −33 minutes of True Solar Time correction — 15:41 becomes 15:08 in solar time. The 申 (Sin) hour runs 15:00–16:59, so the hour pillar is 庚申 with or without the correction. The correction doesn't always change the hour pillar — only when the birth sits near a boundary. | ||||||
| Lunar May 10, 1987 · 14:33 · Sokcho · F | Lunar conversion + Mangjong term + historical DST | 丁卯 정묘 | 乙巳 을사 | 丙戌 병술 | 乙未 을미 | 1 |
Lunar May 10, 1987 converts to June 6, 1987 on the solar calendar. That date sits on the Mangjong (망종) solar-term boundary, so the month pillar is judged as the 巳 month (乙巳). On top of that, Korea was on daylight saving time in the summer of 1987 — clock 14:33 is standard time 13:33 — and Sokcho's longitude (128.59°E) adds its own True Solar Time shift. Three traps stacked in one birth. | ||||||
| Jul 15, 1987 · 15:10 (clock) · Korea · F | Korean daylight saving time (1987–88) | 丁卯 정묘 | 丁未 정미 | 乙丑 을축 | 癸未 계미 | 8 |
Korea ran daylight saving time from May 10 to October 11, 1987 — clock 15:10 is standard time 14:10, the 未 (Mi) hour. Forget the DST and the hour pillar slides one branch to 申. This case isolates the −1 hour DST step: computed against the 135°E meridian, no True Solar Time correction. | ||||||
| Jul 15, 1988 · 15:10 (clock) · Korea · F | Korean daylight saving time (1987–88) | 戊辰 무진 | 己未 기미 | 辛未 신미 | 乙未 을미 | 3 |
1988 was also a DST year in Korea (the Seoul Olympics) — clock 15:10 is standard time 14:10, the 未 hour. Same trap as the 1987 case, but the year, month and day pillars are all different, which also filters out answers recited from memory rather than computed. 135°E meridian, no True Solar Time correction. | ||||||
| Jun 15, 2000 · 23:30 · Seoul · M | Late Rat hour (11 PM–midnight) — day-pillar convention | 庚辰 경진 | 壬午 임오 | 甲辰 갑진 | 乙亥 을해 | 7 |
For births between 11 PM and midnight, schools split on whether the day pillar belongs to that calendar day or the next (야자시 vs 정자시 — the 'late Rat hour' question). Gwiraedang uses the midnight-boundary convention: the day pillar stays on the calendar day (甲辰). For the hour pillar, Seoul's ~−32 min True Solar Time correction turns 23:30 into 22:58 solar time — the 亥 (Hae) hour (it would be 子 without the correction). If another tool's day pillar is one day off here, that's a convention difference, not an error. | ||||||
| Jul 15, 1958 · 10:30 · Korea · F (UTC+8:30 era) | 1954–61 Korean standard time (UTC+8:30) + old DST | 戊戌 무술 | 己未 기미 | 癸巳 계사 | 丙辰 병진 | 2 |
From August 1954 to August 1961, Korean standard time was UTC+8:30 (meridian 127.5°E) — 30 minutes off today's. And the summer of 1958 fell inside an old Korean DST window, so clock 10:30 is standard time 9:30. Gwiraedang follows the common manseryeok convention of computing against the 135° meridian, and flags births from this era on the result screen: the hour branch can sit about 30 minutes from a boundary. When tools disagree on the hour pillar for this era, this handling is usually why. | ||||||
| Jul 15, 1990 · 08:30 · New York · M | Overseas birth — US daylight saving (EDT) | 庚午 경오 | 癸未 계미 | 辛巳 신사 | 壬辰 임진 | 8 |
New York in July is on EDT (daylight saving). The engine converts the local clock time to UT exactly, then judges the hour pillar by True Solar Time at longitude −74.006°. The year and month pillars are cut by the solar-term boundaries at the UT instant of birth. Miss the timezone or DST history of an overseas birth and several pillars slip together. | ||||||
Convention differences — different ≠ wrong
True Solar Time. The hour pillar follows the sun over your birthplace, not the clock. In the first case (Mar 3, 1998 · 11:05 · Seoul · F), applying the correction gives the hour pillar 己巳 기사, while raw clock time gives 庚午 경오. By contrast, in the Jan 17, 1994 · 15:41 · Incheon · F case the hour pillar is 庚申 경신either way — the correction doesn't always change the hour pillar, only when the birth sits near a branch boundary. If a calculator that skips True Solar Time disagrees with ours on the hour pillar, that is a convention difference.
The late Rat hour — day-pillar convention. For births between 11 PM and midnight, schools split on whether the day pillar belongs to that calendar day or the next. Gwiraedang draws the day boundary at midnight, so the day pillar stays on the calendar day; calculators using the late-Rat-hour convention can differ by one day — neither side is wrong. We disclose this convention on screen for the affected case.
Korean standard time, 1954–1961.In this era Korean standard time was UTC+8:30 (meridian 127.5°E) — 30 minutes off today's. Gwiraedang follows the common manseryeok convention of computing against the 135° meridian, and flags births from this era on the result screen because the hour branch can sit about 30 minutes from a boundary. When tools disagree on the hour pillar for these births, this handling is usually the reason.
How this is verified
- All 9 cases in this table are locked by golden tests — if the published expected values and the engine output ever diverge, the tests fail and deployment is blocked.
- Solar-term times (Lichun, Mangjong and the rest) come from our own astronomical calculation — VSOP87 solar longitude with ΔT correction — matched to within ±2 minutes of the times published by the Korea Astronomy and Space Science Institute (KASI) and locked by regression tests.
- The engine itself carries a separate golden-test suite: solar-term cutoffs, exhaustive hour-branch boundaries, daylight saving, lunar conversion, overseas births, and the 1954–61 Korean standard time.
- Manual cross-check: we confirmed the expected values by hand against multiple published manseryeok calendars and KASI figures. We don't name or grade specific services.
- The chart is calculation; everything read on top of it is interpretation, offered as reference — Saju is not a scientifically proven prediction, but a language for self-reflection.
How were the values in this table verified?
Every value in the table is not hardcoded copy — it is the live output of Gwiraedang's manseryeok engine, computed when this page is rendered, and every case is locked by golden tests: if the engine changes, the tests break. Solar-term times come from our own astronomical calculation (VSOP87 solar longitude with ΔT correction), matched to within ±2 minutes of the times published by the Korea Astronomy and Space Science Institute (KASI), and we cross-checked the expected values by hand against multiple published manseryeok calendars. We don't name or grade other services.
If another calculator or an AI disagrees with this table, which one is wrong?
Check the convention first. Where schools genuinely split — whether True Solar Time is applied, how the 11 PM–midnight day pillar is judged, how the 1954–61 Korean standard time is handled — both answers can be right within their own convention. Different is not the same as wrong. But minute-level solar-term cutoffs and forgotten daylight saving are not convention questions; those are calculation errors. Each case's note tells you which kind of difference you're looking at.
If an AI matches this table, can I trust it for a Saju reading?
A correct chart is a good start. But AI answers vary by model and by the day — matching today doesn't guarantee matching tomorrow, so verify yourself. And everything above the chart is interpretation: the chart is calculation; the interpretation is reference, not prediction. Get the chart from a deterministic calculator, and treat any reading as a tendency to reflect on, never a verdict.
Compute your own chart with the same engine
Free, no sign-up. True Solar Time, daylight saving and overseas births handled automatically — every correction disclosed.
Free Saju calculatorRead next: Why ChatGPT gets your Saju wrong · The Four Pillars, explained · Saju glossary · Saju vs BaZi