Can Wordle Have Double Letters? Rules & Examples
Yes. An answer can contain the same letter more than once. Each green or yellow tile uses one copy; any extra guessed copies are gray.
On this page
Yes. An answer can contain a letter more than once, and a guess can repeat it too. Each green or yellow tile accounts for one copy in the answer. If a guess has more copies than the answer, the extra copies are gray—even when another copy of that letter is green or yellow.
The easiest way to read a repeated letter is to count its positive tiles, rather than treating every gray tile as “this letter is absent.” This reference walks through the rule and provides downloadable examples and a reproducible check of the disclosed algorithm.
Three rules for double letters
- Reserve every exact-position match first. Those tiles are green and use up their matching answer positions.
- Read the remaining guess positions from left to right. Award yellow only if an unused copy remains elsewhere in the answer.
- A guessed copy with no match left is gray. It does not cancel a green or yellow copy in that same row.
These are the allocation rules used by this reference. University algorithm notes describe the same two-stage procedure; their links are given below. We do not inspect or test the hidden implementation of the live NYT game.
A gray duplicate can tell you an exact count
For a particular letter, count the green and yellow tiles in one row. Together, they account for the smaller of the number of copies in the guess and in the target.
If the row also contains a gray copy of that same letter, the positive copies give its exact count in the target under this rule. For example, one green E plus one yellow E plus one gray E establishes exactly two Es. If no copy was gray, the positives give a minimum; there could be another copy you did not guess.
This is why removing every word containing a gray letter can break a solver. The row constrains both letter positions and the allowed number of copies.
Worked examples
These are deliberately chosen demonstration pairs. They are not today’s answer, an answer archive, or a claim that every string is accepted by the live game.
E: green · E: gray · R: yellow · I: gray · E: green
The two exact Es use both Es. The R is yellow; the extra E is gray.
A: green · L: yellow · L: gray · E: gray · E: green
Only one L remains to match. Of the extra Es, the final green E takes priority.
S: green · A: yellow · S: gray · S: gray · Y: gray
The first S is green; extra copies of S are gray. A is yellow.
H: green · E: green · L: green · L: green · O: green
Repeated letters can all be green.
Why greens must come first
With EERIE against ERASE, a left-to-right procedure that immediately spends letters can accidentally award the second E a yellow tile. It then has no E left for the final exact match. Reserving all greens before processing yellows avoids this mistake.
Why an “includes” check is not enough
Checking only whether the target contains a guessed letter ignores how many copies are available. For ALLEE against APPLE, both Ls appear to pass that test, but the target has only one L. The second guessed L cannot use the same target copy again.
Use the solver to apply the entire row consistently. Its broad purpose is filtering candidates; this page documents the count rule and the reference data behind these examples.
The reproducible check behind this reference
Version 1.0.0 checks every equality pattern of the ten ordered positions in a five-letter guess plus a five-letter target. Two letter arrangements belong to the same pattern when renaming their letters consistently makes the arrangements identical. The generator produces one canonical representative of each pattern.
There are 115,975 such patterns. This corpus includes arrangements with no repeats, several repeated letters, excess guessed copies, multiple exact matches, and mixtures of exact and misplaced matches. It uses synthetic symbols; it does not need a dictionary or copy a daily-answer list.
| Check | v1 result |
|---|---|
| Equality patterns evaluated | 115,975 |
| Disagreements between two independent implementations | 0 |
| Per-letter count checks | 405,623 |
| Patterns where the naive membership method gives different feedback | 55,550 |
One implementation keeps a counter of unused target letters. The other tracks individual unused target positions. The script compares their entire output for every pattern, and separately checks that each letter receives the correct total of green and yellow tiles.
The 55,550 figure is a count within this synthetic pattern space. It is not a percentage of real Wordle games, a measured error rate of another website, or the probability that a player encounters a duplicate. Dictionary words are not uniformly distributed across these patterns.
Download and reproduce
Download the worked cases, computed results, method, source list, limitations and reproduce-v1.py. Run python3 reproduce-v1.py with Python 3. It uses the standard library and makes no network calls. It regenerates the canonical patterns, checks both implementations, and writes the results and examples.
Sources and boundaries
The University of Central Florida assignment describes the two-pass scoring procedure and the order for assigning remaining matches. We use that scoring section, not the assignment’s separately simplified dictionary-filter rule.
Dan Ridley-Ellis’s university-hosted algorithm notes also explain why repeated guessed copies cannot all claim a single available letter. Brooke Chalmers’s implementation writeup provides an independent developer’s account of the same problem.
Those sources establish the procedure and its practical context. The canonical pattern corpus, verification code and aggregate results on this page were produced by Wordle Solver. This does not mean the sources reviewed or endorsed our tests.
The scope is five ordered single-byte letters and this disclosed feedback rule. It does not cover dictionary acceptance, hard-mode guess restrictions, optimal starting words, other alphabets, or the live game’s internal code. Full equality-pattern coverage is conditional on those assumptions.
Version 1.0.0 · 23 October 2026. Published by Wordle Solver. Not affiliated with The New York Times. Found a mistake? Send a correction with the guess, target, expected colors, and your reasoning. A changed rule or result will receive a new dated version.