ORX-LAB-002 · OPEN RESEARCH
Two calculators, one birth time, different charts — how far apart do conventions push the result?
Calendar, Time-Zone, and Solar-Time Boundary Effects in Three Birth-Chart Systems: A Computational Divergence Study
The short answer
Even with the birth time exactly right, calculators disagree when they apply different conventions. Anyone born in Korea between 1954 and 1961 is the clearest case: the country ran on UTC+08:30 back then, so computing with today’s standard time changes something in 97.7% of the sample.
- 30 minis how far apparent solar time is displaced for those births. Median and maximum are identical — the whole era shifts by the same amount.
- 23.6%of those charts get a different hour pillar. The Zi Wei ming palace moves with it, since both sit on the same two-hour boundary.
- 38.9%of present-day charts change if the solar-time correction is skipped. That one has nothing to do with era — it is purely a convention choice.
- 0.0%changes in the present-day control cell. That zero is what makes the other numbers trustworthy — if it were not zero, the harness would be wrong.
Era × convention
The birth data is held fixed and only the convention changes. “Historical standard time ignored” is a calculator that always assumes UTC+09:00; “solar time ignored” is one that skips the longitude correction.
| 시대 | 바꾼 관례 | 뭐라도 바뀜 | 진태양시 차이 | 시주 | 일주 | 명궁 | 상승점 | MC |
|---|---|---|---|---|---|---|---|---|
| 구표준시 | 표준시 이력 무시 | 97.7% | 30분 | 23.6% | 5.2% | 23.6% | 22.8% | 26.9% |
| 구표준시 | 진태양시 무시 | 24.0% | 6분 (최대 64분) | 20.8% | 3.1% | 20.8% | – | – |
| 서머타임 | 표준시 이력 무시 | 37.5% | 0분 (최대 60분) | 9.4% | 0.0% | 9.4% | 18.6% | 16.2% |
| 서머타임 | 진태양시 무시 | 53.5% | 34분 (최대 94분) | 48.3% | 8.3% | 48.3% | – | – |
| 현행 | 표준시 이력 무시 | 0.0% | 0분 | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% |
| 현행 | 진태양시 무시 | 47.2% | 32분 (최대 34분) | 38.9% | 8.3% | 38.9% | – | – |
The old standard time fit Korea’s longitude better
Korea’s 1954–61 offset of UTC+08:30 puts the standard meridian at 127.5°E. Seoul sits at 126.98°E, so the solar-time correction is only a median of 6 minutes — one fifth of the 32 minutes it takes under today’s 135°E. For that era, getting the standard time wrong is costly, but skipping the solar correction barely matters.
Method
- Sample — three Korean cities × UTC+08:30 (1954-03-21 ~ 1961-08-09) (1956, 1959); UTC+10:00 (여름) (1987, 1988); UTC+09:00 (1980, 2000), 12 dates per year, 12 times per day, for 2,592 births in total.
- Why 127-minute spacing — a Chinese hour is exactly two hours, so sampling every two hours locks the phase relative to the boundary. Our first attempt did exactly that and the hour pillar never changed, because the closest sample sat 38 minutes from any boundary. A study about boundaries was avoiding them.
- Conventions are reproduced through the input — no flag was added to the engine. Instead we compute what input a calculator using that convention would effectively be given, and feed it to the same engine. Both sides run identical code, so the difference can only come from the convention.
- Cells we did not measure — the solar-time convention is left blank for Western astrology. Bazi and Zi Wei read apparent solar time while the astrology path reads UTC, and shifting the input moves UTC too, so the comparison does not hold. Writing 0% there would read as “measured, no change”, which is false.
- Reproducibility — the sampling frame is fixed in code and the computation path has no randomness, no clock reads and no network calls. The same command always produces the same table.
For what happens when the birth time itself is wrong, see the birth-time sensitivity study. This one measures what diverges even when the time is right.
What we measured
- How often results change when historical standard time is ignored, by era
- How often the hour pillar and ming palace change without the solar-time correction
- How many minutes apparent solar time is displaced by each convention
- A control cell in the present-day era, where nothing should change
What this study does not claim
- We do not judge which calculator is correct. We measure only that conventions diverge.
- This is not a study of predictive power or of whether divination works.
- The solar-time convention was not measured for Western astrology; the comparison does not hold there and the cells are left blank.
- The sample is an evenly spread grid of dates and times, not a real birth population.
- Standard-time histories outside Korea are not covered.
Method and data
All three systems were computed by the same production engines this site uses, so no cross-engine difference can leak into the comparison. The calculation path contains no randomness, no wall-clock time and no external calls: the same command produces the same table.
Citing this study
The dataset is published in a separate repository so that a citation keeps working even if this site changes. That repository carries a CITATION.cff file, so GitHub renders a ready-made citation under “Cite this repository”.