Leap years come every four years. So what about 2100?
2000 had a February 29, but 2100 will not. Count how the 97 leap days are distributed across 400 years, and a calendar rule that seems full of exceptions becomes a single calculation.
Estimated time for the article text only. Allow extra time for figures, comparisons and interactive exhibits.
The Gregorian calendar adds a day every four years, omits that addition in years divisible by 100, and retains it in years divisible by 400. Distributing 97 leap days across 400 years produces a calendar with an average year of 365.2425 days.
2000 had a February 29. 2100 will not.
A leap year comes once every four years. That rule will get you through most calendars. Yet 2000 was a leap year, while 2100 will be a common year. Both are divisible by 4 and end in 00, but February has a different number of days in each. Does the rule change along the way?
The Gregorian calendar used today has a three-part rule with exceptions built in from the start. Add a day in years divisible by 4. Do not add it, however, in years divisible by 100. But in years divisible by 400, add it after all. The number 2000 is divisible by 400; 2100 is not. That is what separates the two years.
Before memorizing the exceptions, ask what these additions and subtractions are trying to keep in step. A calendar can grow only in whole days, while the cycle of the seasons does not fit into a whole number of days. Leap years distribute that fractional remainder across a long span of time.
READ THE COMPARISON / Make a discovery in the diagram
Start with 100 leap years in 400 years, then subtract 3.
Among the years 1–400, there are 100 multiples of 4. Make years 100, 200, and 300 common years again; year 400 remains a leap year.
Hatching: exclude 100, 200, 300 / Dark border: keep 400
| What is counted | Result |
|---|---|
| 365 days per common year × 400 years | 146,000 days |
| Additional leap days | 97 days |
| Total ÷ 400 years | Average: 365.2425 days |
Comparison conditionsThis diagram counts a 400-year cycle using Gregorian calendar rules. It does not mean that every region historically used this calendar from the year 1 CE, or that the lengths of the seasons remain constant forever.
See the evidence behind the diagram and the article’s sources
- National Astronomical Observatory of Japan — Which years are leap years? ↗ (opens in a new tab)
- Ephemeris Computation Office, National Astronomical Observatory of Japan — Calendar Wiki: Gregorian calendar ↗ (opens in a new tab)
- Ephemeris Computation Office, National Astronomical Observatory of Japan — Calendar Wiki: Seasons / The cycle of the seasons ↗ (opens in a new tab)
What cycle does a year measure?
Keeping a calendar in step with the seasons involves an average cycle called the tropical year. It lasts approximately 365.2422 days: the whole number 365, plus about 0.2422 of a day. In a calendar that advances by just 365 days each year, this small difference accumulates, gradually shifting the relationship between a given date and the seasons.
Simply remembering that "Earth completes one circuit in 365.2422 days" leaves out what marks the completion of that circuit. A sidereal year, measured against distant stars, is not the same as the tropical year associated with the seasons. Earth's rotation axis changes direction over long periods, so the seasonal reference points also move relative to the stars.
Nor is the actual interval between one spring equinox and the next always exactly 365.2422 days. Celestial motions vary, so we must distinguish an average tropical year from the interval in an individual year. Here we will focus on how a calendar of whole days can approach that average cycle, rather than on the finer changes from year to year.
Adding one day every four years adds a little too much
Collect four years' worth of the approximately 0.2422 days left over after 365, and you get about 0.9688 days—almost a full day. Adding a day every four years therefore brings the calendar closer to the cycle of the seasons than giving every year 365 days. The four-year total is 365 × 4 + 1, or 1,461 days, for an average of 365.25 days per year.
A whole day is slightly longer than 0.9688 days, though. Continuing to add one every four years adds about 0.0312 days too much per four-year period, or roughly 45 minutes, using the approximation here. The difference may barely register once, but it becomes hard to ignore when the same system is used for centuries.
The exceptions are not there to remove February 29 whenever someone feels like it. They reduce the small surplus that remains after the first, four-year adjustment. We add a whole day to approximate the fraction, then refine that relatively coarse adjustment over a longer cycle.
Put 97 leap days into a 400-year box
Count the 400 years from 2001 through 2400 as a single box. The four-year rule alone would give 100 leap days. Remove the four century years divisible by 100—2100, 2200, 2300, and 2400—and 96 remain. But 2400 is divisible by 400, so restore one leap day. The result is 100 − 4 + 1 = 97.
Divide that box into four drawers of 100 years each. There are 24 leap days in 2001–2100, another 24 in 2101–2200, and 24 in 2201–2300, followed by 25 in 2301–2400. Only the final drawer contains the extra day retained by the 400-year exception. Writing the counts as 24, 24, 24, 25 makes it easier to see where the exception falls.
The average over 400 years is 365 + 97 ÷ 400, or 365.2425 days per year. That is closer to the average cycle of the seasons than the 365.25 days produced by adding a day every four years. Over 400 years, just three additions that the ordinary four-year rule would make are skipped. Those three omissions are the purpose of the familiar 100-year and 400-year conditions.
Check the conditions in a different order to catch every exception
To decide whether a particular year is a leap year, begin with the condition that overrides the others. First, is it divisible by 400? If so, it is a leap year. If not, is it divisible by 100? If so, it is a common year. If neither applies, check whether it is divisible by 4: if it is, the year is a leap year; otherwise, it is a common year.
In this order, 2000 is identified as a leap year at the first step, 2100 as a common year at the second, and 2024 as a leap year at the third. Looking only at the century condition and saying that every year ending in 00 is a common year would get 2000 and 2400 wrong. Looking only at the four-year condition would get 2100 wrong instead.
As a result, the next leap year after 2096 is 2104. Skipping 2100 leaves an eight-year gap. That may be surprising if you think leap years are always four years apart, but it follows naturally from the same rule. Calendar exceptions make a real difference to the sequence of dates.
Even the weekdays return after 400 years
The total number of days in 400 years is 365 × 400 plus 97, giving 146,097. Divide that by the 7 days in a week, and the result is 20,871 weeks with no remainder. In the Gregorian calendar, both the pattern of leap years and the weekday of a given month and day repeat after 400 years.
If we take 2001–2400 as one complete cycle, for example, the following year, 2401, begins on the same weekday as 2001, and the leap years occupy corresponding positions. This does not mean that celestial bodies return to exactly the same arrangement after 400 years. It is an arithmetic property: the combination of days prescribed by the calendar also fits exactly into the seven-day cycle.
The number 97, the count of leap days, does not by itself reveal the relationship with weekdays. Add up all the days and divide by a different unit, the week, and another regularity appears. One year, four years, four hundred years, one week: periods we usually think of separately connect through a single division.
365.2425 days is not a permanent portrait of nature
The Gregorian calendar's average year of 365.2425 days is a number determined exactly by its rules. The tropical year against which we compare it is an astronomical quantity that changes slowly over time. The National Astronomical Observatory of Japan gives an approximate difference between the two of about 0.00031 days per year. Across several thousand years, that amounts to roughly one day, but this subtraction alone cannot determine the seasons in the distant future or dictate a calendar reform.
Actual intervals between spring equinoxes do not exactly match the mean tropical year, and there are also long-term changes. It would be premature to treat statements such as "this date will definitely shift in three thousand and some years" or "a new exception applies in the year 4000" as established rules. The current test uses the conditions 4, 100, and 400.
There is also a difference between applying these rules mathematically to a year in the past and establishing that people in a particular place actually used this calendar at the time. Reading dates in historical records requires checking the country or region and when the calendar was adopted. Knowing the arithmetic does not automatically let us read calendars from every era.
February 29 makes a long calculation visible
The extra 29 on a February calendar looks like a small change if you consider only that year. View 400 years in a single diagram, though, and you can see how 97 additions produce an average of 365.2425 days. We cannot write an extra 0.2425 days into each year, so whole days are assigned to designated years instead.
Asked about the difference between 2000 and 2100, you can answer simply by checking divisibility by 400. Go one step further and ask why that condition is needed, and you reach the fraction left between the cycle of the seasons and the calendar's boxes for dates. The exceptions do not obstruct the rule; they adjust it to bring it closer to a natural cycle.
The next time you see February 29, try picturing the century years in which that extra day was not added. Including those choices that never appear on the printed calendar, we balance the account of long stretches of time one day at a time.
Sources and verification notes
This article is an original synthesis of primary sources and explanations from specialist institutions.
Generative AI assists with research, structure and English translation. Claims are compared with sources and editorial checks are performed; independent specialist and native-English reviews have not yet been completed.
- National Astronomical Observatory of JapanWhich years are leap years? ↗ (opens in a new tab)
- Ephemeris Computation Office, National Astronomical Observatory of JapanCalendar Wiki: Gregorian calendar ↗ (opens in a new tab)
- Ephemeris Computation Office, National Astronomical Observatory of JapanCalendar Wiki: Seasons / The cycle of the seasons ↗ (opens in a new tab)
Sources checked:
The National Astronomical Observatory of Japan's leap-year FAQ and its explanations of the Gregorian calendar and seasonal cycles were opened and checked for the leap-year rules, the day count over 400 years, the difference between mean tropical and sidereal years, and the limits of long-range predictions. The number of leap years in 2001–2400, the counts for each century, and the weekday cycle were calculated independently from the rules. Calculations using a tropical year of about 365.2422 days are approximations, not predictions of the interval between spring equinoxes in a particular year.
Rights, status, and revision history
- Status
- Free public edition. Preparation and revision history is retained below.
- 0.1.0
- Rights to the text and diagrams
- The original Japanese text is an independently written summary. The SVG illustrations and conceptual diagrams were newly created. No external images, original source text, or audio recordings are reproduced.
- How sources are used
- Sources are referenced by links only. All rights to each source remain with its respective rights holders.
- · v0.1.0 — Checked primary sources and explanations from specialist institutions to create a private preview draft with original comparisons and diagrams. Connected it to a flagship feature. Formal publication has not been approved.
For another read, another day.
THE NEXT BOOK
A new question, on the next shelf.
These paths are editorial comparisons and associations. They do not claim that different subjects work by the same principle.
When a calendar changes, how do the dates connect?
Mechanically applying these rules to any year is not enough to interpret dates recorded by people in the past. Look at which date actually followed which during Japan's calendar reform, alongside when the new calendar was adopted.
Month 12, day 2—then New Year’s Day. Where did the missing dates go?From the fraction left in a year to differences between clocks and the Sun
Leap days bring the calendar's day count into line with the cycle of the seasons. Standard time separates longitude from the equation of time to explain why the Sun can occupy different positions when clocks show the same 12:00. These are related stories about time, but the quantities being adjusted are different.
When all of Japan says noon, why doesn't the Sun?Before counting fractions, match the units
To compare the number of days in a year, we distinguished tropical years, the calendar's average, and individual years. In the kitchen, too, the same number means different things depending on whether it represents volume or weight. Take a detour to check what counts as one unit.
One tablespoon, three weights: jo-hakuto sugar 9 g, water 15 g, salt 18 gFrom calendar exceptions to a game’s guarantee
Leap years require a complete rule, including its exceptions. A game’s 5% drop also needs conditions: independent attempts or a guarantee? Read the rule behind the number in another setting.
Does a 5% game drop become certain after 20 tries?