I’m working on converting recurring decimals into fractions, and I think I understand the basic idea, but I keep getting tangled when there’s a non-repeating part at the start.
For a simple one like x = 0.\overline{3}, I write 10x = 3.\overline{3}, so 10x − x = 3. That seems clean because the repeating tails line up and cancel. So far, so good.
Where I get stuck is something like x = 0.1\overline{8} (so 0.18888…). I tried:
– 10x = 1.\overline{8}
– 100x = 18.\overline{8}
If I do 100x − 10x, the decimals look like they cancel and I get 90x = 17. But I’m not 100% sure why 90 is the correct coefficient here (and not 99 or 9), or whether subtracting 100x − x would be wrong. I also wondered if I should split it first as 0.1 + 0.0\overline{8}, but that feels like I’m moving the repeating block to the wrong place. What’s the reliable way to set this up?
Another example giving me trouble is y = 2.\overline{054}. Here’s what I wrote:
– 1000y = 2054.\overline{054}
– y = 2.\overline{054}
Subtracting should give 999y = 2052 (or is it 2052.\overline{0}? I’m second-guessing the alignment). I also tried mixing different multiples (like 1000y − 2y) after thinking about shifting the non-repeating part, but then I get a different integer on the right. Clearly I’m not lining things up consistently.
For z = 0.\overline{36}, I wrote 100z = 36.\overline{36}, so 99z = 36, which feels straightforward. But what if I accidentally pick a longer block than necessary? For example, if I multiply by 1000 and also by 10, I get 1000z − 10z = 990z, and the subtraction 363.\overline{6} − 3.\overline{6} looks like it gives 360. That seems to still work, just with bigger numbers. Is this always guaranteed to produce the same final fraction, even if I choose a longer repeating block than needed? How do I know the repeating block I chose is actually minimal?
Could someone please spell out the precise, step-by-step rule for which powers of 10 to use before subtracting when there’s a non-repeating start followed by a repeating block? I’d really like a principle I can apply mechanically so the tails always cancel cleanly without off-by-one mistakes.
Follow-up: Is there a general formula in terms of m = length of the non-repeating part and n = length of the repeating block that I can use to check my setup? And once I get a fraction, is there a quick way to predict the length of the repeating part when converting back to a decimal (or at least verify that the repeating block I picked was minimal)?
















3 Responses
Here’s the reliable, mechanical rule I use: first “park” the decimal right at the start of the repeating block, then “jump” one full block and subtract. Concretely, let m be the number of non‑repeating digits after the point and let n be the length of the repeating block. Step 1: multiply x by 10^m so the repeat begins immediately after the decimal; call that u = 10^m x. Step 2: multiply u by 10^n so one full copy of the repeating block shifts past the point. Step 3: subtract to cancel the tails: (10^n − 1)u is an integer, so x = (integer difference) / [10^m(10^n − 1)]. This also explains your coefficients: for x = 0.1overline{8}, m = 1 and n = 1, so u = 10x = 1.overline{8}, then 10u − u = 9u = 17, hence x = 17/[10·9] = 17/90 (splitting as 0.1 + 0.0overline{8} gives 1/10 + 8/90 = 17/90 too, so that’s fine). For y = 2.overline{054}, m = 0 and n = 3, so 1000y − y = 999y = 2054 − 2 = 2052, giving y = 2052/999 = 76/37 after reducing; your alignment there was right. For z = 0.overline{36}, using 100z − z gives 99z = 36, but using a longer block like 1000z − 10z gives 990z = 360; both reduce to the same fraction because you’ve just multiplied numerator and denominator by the same power of 10-so yes, any longer block that’s a multiple of the true period still works. To check minimality of the block you picked: see if your block is itself made by repeating a shorter pattern (e.g., “141414” isn’t minimal because it’s “14” repeated), or use the fraction: for a reduced fraction p/q, strip all 2s and 5s from q to get q′; the length of the repeating part is the smallest n with 10^n ≡ 1 mod q′ (and the non‑repeating length is max of the exponents of 2 and 5 you stripped). Summary formula you can trust: if the digits look like [integer part][m non‑repeating digits][n repeating digits], let A be the integer formed by the first two brackets together and B be the integer formed by including one copy of the repeating block; then x = (B − A) / [10^m(10^n − 1)].
I always picture these like sliding two clear overlays until the repeating stripes line up: first shove past the non-repeating “intro,” then line up one full repeat with the next so the tails cancel. I sometimes second‑guess whether it should be 99x or 9x, but the mechanical rule that keeps me honest is: if m is the number of non‑repeating digits and n is the length of the repeating block, use the pair 10^(m+n) and 10^m; subtracting gives (10^(m+n) − 10^m)x equals an integer. In fact, if A is the integer made from the first m+n digits after the decimal (ignore the decimal) and B is the integer from the first m digits, then for a number with no integer part you get x = (A − B) / (10^m(10^n − 1)); if there’s an integer part I, just do this for the fractional part and add I back at the end. So your examples: x = 0.1̅8 has m = 1, n = 1, so 100x − 10x = 17 ⇒ x = 17/90 (splitting as 0.1 + 0.0̅8 also works: 1/10 + (8/9)/10 = 17/90); y = 2.̅054 has m = 0, n = 3, so 1000y − y = 2052 ⇒ y = 2052/999 = 76/37; z = 0.̅36 has m = 0, n = 2, so 100z − z = 36 ⇒ z = 36/99 = 4/11. If you “overshoot” with a longer block (say 1000z and 10z), you just pick up a common factor: 1000z − 10z = 990z = 360, and 360/990 reduces to the same 4/11 because 10^(k+n) − 10^k = 10^k(10^n − 1). For the follow‑up: given a reduced fraction p/q, the non‑repeating length m is the maximum of the exponents of 2 and 5 in q (that’s how many trailing zeros you need to clear), and the repeating length n is the smallest n with 10^n ≡ 1 mod q′, where q′ is q with all 2s and 5s removed (that “multiplicative order” tells you the minimal block). Does it help if you try one more, like 3.04̅27 (m = 2, n = 2), and see if (10^(2+2) − 10^2)y lines up cleanly for you? Nice walkthrough here too: https://www.khanacademy.org/math/arithmetic/arith-decimals/arith-review-repeating-decimals/a/rewrite-repeating-decimals-as-fractions
I’m pretty sure the clean rule is: if the nonrepeating part has length m and the repeating block has length n, let X = 10^m x so the repetend starts at the decimal point; then (10^n − 1)X is an integer, giving x = ((uv as an integer) − (u as an integer)) / (10^m(10^n − 1))-so 0.1 8̅ → (18−1)/(10·9)=17/90, 2.054̅ → 2+54/999=2052/999, and using any longer multiple of n still cancels and reduces to the same fraction.
Conversely, for a reduced a/b write b=2^α5^βr with gcd(r,10)=1: the nonrepeating length is max(α,β) and the repeating length is the multiplicative order of 10 mod r, which (I think you’ll like this!) also certifies your chosen repetend is minimal.