355/113 Is Good Because the Next Term Is 292
Published 10/21/2026 · 3 min read · Everyday calculators
Lena Hoffmann — Science & education writer at OneKitly
Mathematics · Physics
Checked against 2 sources
A continued fraction rewrites a number as a whole part plus one over a whole part plus one over another, and the sequence of whole parts is its signature: 415/93 is [4; 2, 6, 7], and truncating that expansion early gives 4/1, then 9/2, then 58/13, then the exact 415/93. Those truncations are called convergents, and they are provably the best rational approximations available for their size — no fraction with a smaller denominator gets closer. For π the expansion opens [3; 7, 15, 1, 292, …], whose convergents are 3, 22/7, 333/106 and 355/113. A large term signals that the convergent before it is unusually good, and 292 is very large: 355/113 is off by only 85 parts in a billion.
π expands as [3; 7, 15, 1, 292, …], and the convergent just before that 292 is 355/113 — accurate to 85 parts in a billion, from a fraction with three digits on top.
The golden ratio is the hardest number to approximate
If large terms mean good approximations, the number with no large terms at all must be the worst case — and that number is the golden ratio, whose expansion is [1; 1, 1, 1, 1, …] forever. Its convergents are ratios of consecutive Fibonacci numbers, 1/1, 2/1, 3/2, 5/3, 8/5, 13/8, and each one improves on the last by the smallest possible margin. That is not a curiosity: a sunflower packing seeds at an angle related to the golden ratio never lets successive seeds line up, precisely because no simple fraction approximates the angle well. Being badly approximable is the useful property there.
A repeating pattern means an irrational of a particular kind
The square root of two expands as [1; 2, 2, 2, 2, …], repeating a single term forever, and every square root of a non-square whole number behaves this way — the pattern eventually cycles. That is a theorem, and it runs both ways: a continued fraction whose terms are eventually periodic is always the root of a quadratic equation with whole-number coefficients, and one that never settles into a pattern is something else. π's expansion has no pattern anybody has found, which is consistent with what is known about it. The finite case is simpler still: an expansion that stops is a rational number, and 415/93 stopping after four terms is the whole content of that statement.
| Convergent | Value | Relative error |
|---|---|---|
| 3/1 | 3 | 4.5 % |
| 22/7 | 3.142857 | 0.040 % |
| 333/106 | 3.141509 | 0.0026 % |
| 355/113 | 3.14159292 | 0.0000085 % |
Frequently asked questions
- Why is 22/7 so famous if 355/113 is far better?
- Because it is short enough to remember and accurate enough for most of what people did with it. 22/7 is wrong by four hundredths of a per cent, which is invisible in any measurement made with a ruler and adequate for most workshop arithmetic; 355/113 is wrong by less than a ten-millionth, which nobody needs unless the calculation is being carried to many digits. The convergents form a ladder of accuracy against memorability, and 22/7 sits at the rung where the trade is most favourable for mental arithmetic. Where more precision is needed there is usually also a calculator.
- Where are continued fractions used in practice?
- Wherever a ratio has to be built out of whole things. Gear trains approximate an arbitrary speed ratio with integer tooth counts, and the convergents give the smallest gears that get close enough; calendar rules approximate the length of a year with whole days and leap rules; musical temperaments approximate an octave with a whole number of steps. The same machinery also appears inside algorithms for reducing fractions and for reconstructing a rational number from a decimal, which is what lets a computer algebra system recognise that 0.142857142857 is a seventh rather than a long decimal.
Articles you may find interesting
All guides →Related tools
Sources
Spotted a mistake in this article?