The Ninth Month Began on Two Different Days

Two national astronomical offices published the same instant this year and gave it two different calendar dates. The National Astronomical Observatory of Japan's calendar summary for 2026, the Rekiyoko — the official table of solar terms and moon phases, released in February 2025 — lists October's new moon at 0:50 on October 11, Japan Standard Time. The Hong Kong Observatory's 2026 table of Gregorian and lunar dates puts the first day of the ninth lunar month on October 10. Neither office has made a mistake, and they are not describing different events. 23:50 on the 10th in Beijing is 00:50 on the 11th in Tokyo, and a lunar month begins on the day the new moon falls on.

So the ninth month of this year began on October 10 in Hong Kong and on October 11 in Tokyo, and it has been one day apart ever since. Korean calendars follow the Korea Astronomy and Space Science Institute, which reckons in Korea Standard Time — the same meridian as Japan — and they agree with Tokyo: today, Wednesday, October 21, is the 11th day of the ninth lunar month in Seoul and the 12th in Hong Kong. Vietnamese almanacs keep UTC+7 and side with Hong Kong, giving today as the 12th. All four print the same day pillar.

One rule, four clocks

The rule is the first of the three the lunar calendar is made of, and it is short: a lunar month begins on the day containing the new moon. Not the hour of it, not some rounded-off portion of a day — the calendar date the instant lands on. The first of the month is that whole day, and every date after it is simply how far you are into the month. The comparison that rule asks for has no middle ground: the instant is either before midnight or after it. Which is why it needs a clock, and why the clock has to be declared by somebody.

China and Hong Kong declare UTC+8. Japan and Korea declare UTC+9. Vietnam declares UTC+7. A one-hour difference in the declared clock is enough to move a month boundary whenever the new moon falls in the last hour of a Beijing day, and this October it fell in the last ten minutes of one. The asymmetry is worth seeing plainly: the same instant sat ten minutes from a boundary on one meridian and fifty minutes clear of one on the next. In Hong Kong this was a close call. In Tokyo it was not close to anything.

Which layers of an almanac move, and which do not

This is the part worth taking away, because the answer is clean. Almost nothing moves. Today is the Yang Earth Dragon day of the Yang Earth Dog month of the Yang Fire Horse year in Hong Kong, in Hanoi, in Seoul and in Tokyo alike — the Korean calendars print the day's pillar as mujin and the Vietnamese ones as Mau Thin, which are both the Yang Earth Dragon. The day pillar is a count from a fixed epoch across calendar dates, and all four countries number this Wednesday, October 21, so the count lands in the same place everywhere.

Everything derived from those pillars therefore lands in the same place too. For today our own almanac gives the day officer as Break, the day's god as the Green Dragon, the sound element as Wood of the Great Forest, the lunar mansion as the Basket (Water Leopard), the clash on the Dog, and six of the twelve hours favorable: Tiger 3–5, Dragon 7–9, Snake 9–11, Monkey 15–17, Rooster 17–19 and Pig 21–23. A Vietnamese almanac for the same date lists its Yellow Path hours as the Tiger, Dragon, Snake, Monkey, Rooster and Pig — the same six, in the same ranges, across a border where the lunar date itself disagrees by a day.

The solar terms are instants as well, so the month pillar can split the same way. It does not this autumn: Frost's Descent falls at 17:38 Beijing time on October 23 and the Start of Winter at 17:52 on November 7, both in the middle of an evening on any of these meridians. The month pillar therefore turns on the same day everywhere, and with it the cycle of day officers and date selection, which is built on the day officers rather than on lunar dates. A date chosen for an undertaking in Hong Kong is the same date in Seoul.

Two things do move. The lunar date line itself, which is what the whole of this post is about. And Yang Gong's Thirteen Taboos — the one layer in our own almanac keyed to a lunar date rather than to a pillar. It is thirteen fixed dates, of which this month's is the 25th of the ninth month. By Hong Kong's count that is November 3, and our code fires it there. By Seoul's and Tokyo's count it is November 4, where our code prints the 26th of the ninth month and no portent at all.

The Double Ninth, on two days

The Double Ninth is the ninth day of the ninth month — nine days counted inclusively from the first, which means it moves with the month boundary and nothing else. By the Hong Kong Observatory's table, the ninth of the ninth fell on Sunday, October 18. Korean calendars mark the festival, Jungyangjeol, on Monday, October 19. The festival whose entire claim to significance is a number was observed on two different days, one day apart, in two countries that take it from the same text.

We wrote about that ten-minute margin on this site on October 12, and we got the framing wrong in one respect worth correcting. We said that had the new moon arrived ten minutes later, the first of the month would have been the 11th and the Double Ninth would have fallen on October 19 — and we offered the two days' readings side by side as a comparison between the real day and the day it nearly was. October 18 comes out Level, Black Path, with the Dipper's Handle firing; October 19 comes out Settle, Yellow Path, carrying Month Virtue instead. That comparison was not between a real day and a hypothetical one. It was between Hong Kong's festival and Korea's.

“

New moon: 11 October, 0:50

None of this is new and it is not a curiosity. Park Seok-jae, a researcher at the Korea Astronomy and Space Science Institute, set the mechanism out in a newspaper column in 2014 and gave the 2012 case: the new moon that opened the fifth lunar month fell at 00:02 on June 20 in Korea and 23:02 on the evening of the 19th in China. The same column records who objected hardest when the two calendars came apart that year — date-selection practitioners. That is exactly the right audience to have noticed, because they are the people for whom a lunar date is an input rather than a label.

Our honest read

The useful rule that falls out of this is narrow and worth holding onto: anything counted from the first and the fifteenth of the lunar month is local, and anything counted in the stem-branch cycle is not. A festival date, a monthly observance, a taboo list keyed to a lunar day — those are statements about a particular country's calendar, and carrying one across a border can shift it by a day without any visible sign that it has moved. A day pillar, an officer, an hour verdict, a chosen date for an undertaking: those travel. The failure to guard against is small and specific, and it is taking a lunar date off one country's calendar and feeding it into a rule written for another's.

The split closes on November 9. Hong Kong's ninth month runs thirty days and Japan's and Korea's run twenty-nine, so all of them print the first of the tenth month on the same Monday, and the new moon that does it falls at 16:02 Japan time — hours clear of any boundary, on every meridian involved. Thirty days of divergence, opened by ten minutes and closed by arithmetic.

Which is the whole argument this site keeps making, arriving from a direction we did not expect. The lunar calendar is not a table somebody wrote down once; it is a calculation, and a calculation has inputs, and one of the inputs is a choice about where you are standing. That choice is almost never printed. If you want one reason to distrust a lunar date that arrives with no method attached, it is that nothing on its face tells you which of these thirty days you are holding — and this month, for anybody reading between Hanoi and Tokyo, that is a real question with two correct answers.