Most Roman numeral errors are not random. They come from a small set of misunderstandings about repetition, subtraction, and place value, and each misunderstanding produces a characteristic invalid form that the converter rejects with an error. This guide catalogues those forms — the ones the converter rejects, the ones people defend anyway, and the near-misses that look right but are not — with the correct form beside each one. If the converter has ever rejected your input, the reason is almost certainly in this guide, and the fix is usually a single pair substitution.

Repetition errors: IIII, XXXX, VV, LL, DD

The repetition rule is simple: a 1-symbol (I, X, C, M) may repeat at most three times, and a 5-symbol (V, L, D) may never repeat. Every repetition error violates one half of that rule. IIII violates the first half: four Is where IV belongs. XXXX violates it at the tens place, where XL belongs. VV violates the second half: two Vs where X belongs, since V plus V is 10 and 10 already has its own symbol. LL violates it where C belongs, and DD where M belongs. The converter rejects all of these, and the fix in each case is to replace the repetition with the symbol or pair that already means that value.

The IIII case deserves special attention because it is the one repetition error people defend. Clock faces traditionally show IIII for 4, opposite VIII for 8, for visual balance on the dial. That tradition is real, old, and widely followed by clockmakers — and it is still not standard Roman form. The converter follows the standard, so it rejects IIII, and a reader who writes IIII in a document, an exam, or an engraving outside a clock face risks being marked wrong. The distinction to remember is context: on a clock face, IIII is a deliberate design tradition; everywhere else, 4 is IV. The same reasoning covers IIII on some historical inscriptions, where carvers occasionally used additive forms before subtractive notation became standard. Historical additive forms are evidence about history, not about the modern standard the converter implements.

Subtraction errors: IL, IC, ID, IM, XM

Subtraction errors come from over-generalising the subtractive idea: if I before X means 9, why should I before L not mean 49? The answer is the “next two symbols” rule: a 1-symbol subtracts only from the next two symbols above it in the order I, V, X, L, C, D, M. I subtracts from V (IV, 4) and X (IX, 9), but not from L, C, D, or M, so IL (49), IC (99), ID (499), and IM (999) are all invalid. X subtracts from L (XL, 40) and C (XC, 90), but not from D or M, so XD (490) and XM (990) are invalid. C subtracts from D (CD, 400) and M (CM, 900), and there is nothing above M for it to subtract from in the standard system.

The correct forms show why the restriction exists: each place value is handled by its own pair, so 49 is XLIX (40 + 9), 99 is XCIX (90 + 9), 499 is CDXCIX (400 + 90 + 9), and 999 is CMXCIX (900 + 90 + 9). Each place — hundreds, tens, ones — contributes its own correct pair, and the numeral is the join of those contributions. A shortcut pair like IL tries to handle two places at once (the 50s place and the ones place in a single subtraction), and the system forbids that because it would break the place-by-place uniqueness that makes the notation reliable. Whenever you are tempted by a shortcut pair, split the number by place value instead, and the correct pairs will present themselves.

Double subtraction: IIX, XXL, CCD

Double subtraction tries to subtract twice in one place: IIX for 8 (10 − 1 − 1), XXL for 30 (50 − 10 − 10), CCD for 300 (500 − 100 − 100). All are invalid in strict form. The rule is that each place value uses at most one subtraction, and that subtraction is one of the six pairs. Eight is VIII (5 + 1 + 1 + 1), not IIX, because the ones place is built additively from 5 once the value exceeds 5, with no second subtraction. Thirty is XXX (10 + 10 + 10), not XXL, because the tens place below 40 is built additively. Three hundred is CCC (100 + 100 + 100), not CCD, for the same reason at the hundreds place. The pattern is consistent: subtraction appears only for the values 4 and 9 at each place (4, 9, 40, 90, 400, 900), and every other value at that place is built additively.

XXL deserves a note because it circulates as a clothing-size joke and occasionally as a claimed numeral for 30. As a size label it is a modern commercial code, not a Roman numeral, and treating it as one confuses two systems that happen to share letters. As a numeral it fails twice: it subtracts twice, and it subtracts X from L across a gap the rules do not allow as a double operation. The correct form for 30 is XXX, and the correct response to XXL in a conversion context is to recognise it as a size label or an error, not as a numeral to convert. The converter rejects it, correctly, and the fix is XXX.

Place-value errors: reading the wrong place

Place-value errors are subtler than repetition or subtraction errors, because the resulting numeral is valid in form but wrong in value. The classic case is misreading a pair’s place: reading XC (90) as if it were at the hundreds place, or reading CM (900) as if it were at the tens place. A second case is mis-splitting a numeral into places: reading MCMXCIV as M (1,000) + CM (900) + XC (90) + IV (4) is correct, but reading it as M (1,000) + C (100) + M (1,000) + X (10) + C (100) + I (1) + V (5) — ignoring the subtractions — gives 2,216, which is wrong by 222. The left-to-right look-ahead method from the companion guide prevents both errors, because it forces the reader to check, at each symbol, whether the next symbol is larger before adding.

Writing errors mirror reading errors. Writing 1,994 as MCMXCIIII (adding an extra I), as MCMXCVI (writing 6 instead of 4 in the ones place), or as MCMXIV (dropping the XC tens place entirely) all produce valid-looking or near-valid strings with wrong values. The defence against all of them is the chips-sum check: add the breakdown values and confirm they total the intended number. MCMXCIV breaks down as 1,000 + 900 + 90 + 4 = 1,994. Any variant whose breakdown does not sum to 1,994 is wrong, and the place where the sum diverges identifies the faulty place. That single check, applied to every conversion that matters, catches nearly every place-value error before it reaches print.

Lookalikes: V11, V111, and digit confusion

A distinct family of errors comes from mixing Roman letters with Arabic digits. Strings like V11 or V111 appear in searches because a reader has seen VII or VIII in a font where the Is look like the digit 1, or because a keyboard or a copy-paste has substituted digits for letters. V11 is not a Roman numeral: it contains Arabic digits, which have no Roman value, and the converter rejects it as invalid input. The intended numeral behind V11 is usually VII (7): V (5) followed by II (1 + 1). Behind V111 it is usually VIII (8): V (5) followed by III (1 + 1 + 1). The fix is to replace every digit 1 with the letter I and re-read: VII is 7, VIII is 8, and the confusion dissolves.

The same digit confusion affects other letters in poor fonts or handwriting: I and 1, V and a check mark, X and a multiplication sign, L and the digit 1 in some typefaces. When a numeral looks wrong, the first check should be character identity: is every character actually a Roman letter, or has a digit or symbol crept in? The converter’s error message for invalid input is the signal to perform that check, not to force the input through. Paste the numeral into a plain-text editor, confirm each character, and re-enter it. Most “converter is broken” reports in this topic resolve at this step, with the converter vindicated and a digit replaced by a letter.

Case, spacing, and punctuation

Case is not an error in this converter: it accepts lowercase input and reads it as the corresponding uppercase symbols, so “mmxxvi” converts as MMXXVI and “mcmxciv” as MCMXCIV. That leniency is a converter behaviour, not a Roman rule — classical inscriptions used capitals, and modern standard form is uppercase — but it means a lowercase entry is never the reason for a rejection. Spacing and punctuation are a different matter: spaces inside a numeral, dots between parts, or hyphens can change how the input is parsed, and the safest habit is to enter a single numeral with no internal spaces or punctuation. For dates, where parts must be separated, use the converter’s date modes with their explicit part fields rather than improvising separators inside a single field.

A related habit worth adopting is to avoid letter-spacing a numeral for display and then copying the spaced version back as input. Engravers letter-space Roman numerals for visual effect — M M X X V I with wide spacing looks monumental — but the spaced string is a display form, not an input form, and pasting it into a converter may produce an error or a misreading depending on how spaces are handled. Keep a clean, unspaced copy of every numeral that matters, and treat spaced versions as display-only. The same caution applies to numerals copied from PDFs or images, where spacing and character substitution are common.

A fixing routine for any rejected numeral

When the converter rejects an input, work through this routine in order. First, check character identity: every character a Roman letter, no digits, no stray punctuation. Second, check repetition: no 1-symbol four times, no 5-symbol twice. Third, check subtraction: every smaller-before-larger occurrence must be one of the six pairs, and no place may contain two subtractions. Fourth, check place value: split the numeral by place, convert each place, and confirm the places sum to the intended number. Fifth, if the intended number is known, regenerate the numeral from the number by the greedy method and compare: the regenerated form is the correct form, and any difference from your input localises the error. In practice, steps one to three resolve the large majority of rejections, and step five resolves the rest.

Two worked fixes illustrate the routine. Suppose the input is IC for an intended 99. Character identity passes (both letters). Repetition passes (no repeats). Subtraction fails: IC is not one of the six pairs, because I subtracts only from V and X. The fix by place value is XCIX (90 + 9). Suppose the input is XXXX for an intended 40. Character identity passes, repetition fails (X four times), and the fix is XL. Suppose the input is VV for an intended 10. Repetition fails (V repeated), and the fix is X. Each fix replaces the faulty construction with the symbol or pair that already means that value, which is the general principle: strict Roman form never needs a new construction, because every value already has one.

Validating by round trip

The strongest validation available is the round trip: convert a number to Roman form, convert the result back to a number, and confirm the original number returns. A round trip that fails identifies an error in one direction or the other, and the direction can be isolated by checking the breakdown chips: if the chips sum to the original number, the forward conversion is correct and the backward reading is at fault, and vice versa. For lists, batch mode makes round-tripping cheap: convert the list, convert the results back, and diff the two lists. Any line that differs is a line to fix by the routine above. For single values that matter — a date to be engraved, a year for a cornerstone, a chapter number for print — a round trip plus a chips-sum check is the minimum diligence before committing the numeral to a permanent medium. Stone, unlike a web page, does not accept corrections gracefully.

Frequently asked questions

Why does the converter reject IIII?

Because strict Roman form allows a 1-symbol to repeat at most three times, and 4 has its own subtractive pair, IV. IIII on clock faces is a traditional design exception, not standard form.

Is XXL a valid Roman numeral?

No. XXL subtracts twice in one place, which strict form forbids, and as a size label it is a modern commercial code, not a numeral. The correct form for 30 is XXX.

What does V11 mean in Roman numerals?

Nothing: V11 mixes a Roman letter with Arabic digits, which have no Roman value. The intended numeral is usually VII (7), with the digit 1 mistaken for the letter I. Replace digits with letters and re-read.

Can I write 49 as IL?

No. I subtracts only from V and X, so IL is invalid. The correct form for 49 is XLIX (40 plus 9), with each place value contributing its own pair.

Why is VV wrong for 10?

Because 5-symbols never repeat. V plus V is 10, and 10 already has its own symbol, X. The same rule makes LL wrong for 100 (use C) and DD wrong for 1,000 (use M).

How do I fix a numeral the converter rejects?

Check character identity, then repetition, then subtraction against the six pairs, then place value by splitting the numeral and summing the breakdown. If the intended number is known, regenerate the numeral from it by the greedy method and compare: the regenerated form is the correct form.