The 33-Year Lunar-Solar Cycle

People searching for the 33-year cycle are usually after a date. When was the last one, when is the next? Both questions can be answered, though not on the terms they are asked. No particular day marks a changeover. What exists is a drift, and it has come full circle at a different moment for every date you might follow.

Most people look for it as the "lunar-solar cycle"; the technical term is lunisolar, meaning a calendar that binds lunar months to the solar year. Both refer to the same thing, the relationship between the lunar month and the solar year. As you will see, neither label really fits what the 33 years actually describe.

The short answer

A purely lunar calendar runs away from the solar year. Twelve lunar months last 354.37 days, a solar year 365.24. So every year the months slip forward by 10.88 days.[1] Eventually they have travelled all the way through the year and arrive back at the start. That takes 33.59 lunar years, which is 32.59 solar years, or just under 32 years and seven months.

The famous number 33 therefore refers to lunar years rather than solar ones, and it is rounded. The US Naval Observatory phrases it carefully and correctly: the twelve-month cycle "regresses through the seasons over a period of about 33 years".[2] That "about" carries weight, because 33.59 is not a whole number. The alignment consequently never repeats exactly, only ever approximately.

It becomes concrete once you pin down a date. Take the start of Ramadan, which fell on February 18, 2026, and the series looks like this:[3]

The gap swings between 32 and 33 years, averaging 32.6. That is precisely why no fixed figure works: one pass does not fit neatly into whole years.

The real Islamic cycle is 30, not 33

The Islamic calendar does have a fixed leap cycle, but it runs to 30 years. Friedrich Karl Ginzel sets it out in detail in his 1906 standard work on chronology: a 30-year cycle of astronomical lunar years amounts to 10,620 days plus 11 days, so 10,631 days in total, or 30 civil years and 11 leap days.[4] According to the Arab astronomers, the leap years are the 2nd, 5th, 7th, 10th, 13th, 16th, 18th, 21st, 24th, 26th and 29th of the cycle.[5]

That cycle is remarkably accurate, missing the true lunar motion by about one hundredth of a day per 30 years. What it binds the calendar to, though, is the Moon, not the Sun. So the Islamic calendar keeps drifting through the seasons no matter how precisely its leap cycle works. Anyone keeping a purely lunar calendar accepts that drift deliberately.

The other 33: a Persian solar calendar

A second 33 exists, and it compounds the confusion because it is a real leap cycle. It comes from the Persian calendar reform of 1079, the work of a commission of eight mathematicians "among whom Omar Khayyam is pre-eminent," as Ginzel puts it.[6] This Jalali calendar, however, is a purely solar one. Its 33 has nothing to do with the Moon.

Even there the matter is less settled than popular accounts suggest. Ginzel reports that 33-year leap periods alternated with 37-year and 29-year ones, then adds a sentence you rarely see quoted: which years within the cycles counted as leap years, and how many, "remains uncertain."[7] The tidy 33-year cycle with leap days in years 2, 6, 10, 14, 18, 22, 26 and 30, the one circulating through the literature, traces back to a computational assumption made by Wilhelm Matzka in 1844. Ginzel names it explicitly as the premise of a conversion formula rather than an attested rule.[8]

Why 33 years fail as a lunisolar cycle

A lunisolar calendar does exactly what a purely lunar one declines to do: it binds the Moon to the Sun so the months stay in step with the seasons. For that it needs a span in which a whole number of lunar months corresponds almost exactly to a whole number of solar years. The 33 meets that condition poorly:[9]

After 19 years the Moon and Sun stand almost perfectly together again, off by a little over two hours. After 33 years, four and a half days are missing. A 33-year scheme is therefore three times less accurate than the octaeteris and roughly fifty times less accurate than the Metonic cycle. No calendar-maker would have chosen it, and historically none did. The arithmetic confirms what the sources already show: the 33 is a drift figure and a solar-calendar figure, but never a lunisolar leap figure.

One more confusion lurks nearby. Ginzel describes an eight-year cycle used in Turkish calendars, made up of 2,835 days, which fills exactly 405 weeks.[10] Those eight years bind the lunar calendar to the week, not to the Sun. Despite sharing a number, it has nothing in common with the Greek and Germanic octaeteris, which brings 99 lunar months together with eight solar years.

What about the Germanic calendar?

The 33 plays no part there. The Germanic lunar calendar works with the octaeteris, eight years carrying three leap months, and after Christianisation the 19-year Easter cycle took its place. Where medieval sources name longer spans, they speak of 19 or 304 years, never 33.

No source supports a Germanic reading of the 33. It belongs to Islamic and Persian chronology, where it is well documented. For how differently cultures have solved the underlying problem that Moon and Sun refuse to line up, see the survey on the history of lunar calendars.