Most arithmetic has one right answer. Date arithmetic often has two defensible ones, and which you get depends on software nobody told you about.
Add one month to 31 January 2026 and the honest answer is that there is no 31 February. Some systems roll forward into March, some clamp back to 28 February. Both are reasonable. They are not the same date.
The leap year rule has three parts, not one
Most people know the divisible-by-four rule. Fewer know it has two exceptions stacked on top:
- Divisible by 4, and it is a leap year. 2024 was.
- Unless it is divisible by 100, and then it is not. 1900 was not a leap year.
- Unless it is also divisible by 400, and then it is. 2000 was.
The middle rule catches people out because most of us have never lived through a century year that broke the pattern. The one century year in living memory, 2000, was an exception to the exception and behaved like an ordinary leap year. The next time the rule bites is 2100, which will not be a leap year despite dividing neatly by four.
The leap year calculator reports which of the three rules applies to a given year, which is more useful than a yes or no on its own.
Why 29 February is a problem for anniversaries
Add one year to 29 February 2024 and the target date does not exist. There is no 29 February 2025. A system that rolls forward gives 1 March; a system that clamps gives 28 February.
This matters for anything dated: subscription renewals, insurance periods, notice periods and birthdays all have to pick a convention, and different providers pick differently. Someone born on 29 February has a legal birthday that varies by jurisdiction for exactly this reason.
Month lengths break the intuition of "a month later"
Adding one month to 31 August 2026 lands on 1 October, because September has 30 days. Adding one month to 31 January 2026 lands on 3 March, because February has only 28 that year. The same instruction, one month, produced a one day overshoot in the first case and a three day overshoot in the second.
That is not a fault in the arithmetic. It is what happens when a unit that varies in length, the month, is used as if it were fixed. The date calculator rolls forward in these cases, so it is worth knowing which convention any given system uses before relying on it for a deadline.
If the answer has to be exact, count in days
Days are the only calendar unit that does not change length. Anything with legal or financial consequences attached is safer expressed in days, which is why notice periods and cooling-off periods are usually written that way rather than in months.
When a period does have to be in months, the contract should say which convention applies at the end of a long month. Most disputes about a date are really disputes about a convention that was never stated.
For elapsed time in days, weeks and years at once, see the age calculator. For how many days a particular year or month actually holds, see the days in the year calculator. For the separate question of working days against calendar days, see the piece on business days.