WordleSolver is a constraint filter, not an answer predictor. It checks every word against the rows and optional filters you entered, removes contradictions, and orders the survivors by how well their distinct letters cover the remaining set.
From colored tiles to a candidate list
The current tool can search four-, five-, six-, and seven-letter word lists. For the standard five-letter game, it starts with 3,424 entries. Each complete or partial clue row is applied in turn, so a word must satisfy the first row, the second row, and every later row to remain.
- Select the list. Word length chooses the source vocabulary.
- Apply tile positions. Green positions must match; yellow positions must not.
- Apply letter counts. Colored copies set a minimum and gray extras can set a maximum.
- Apply optional filters. A pattern, required letters, and exclusions narrow the set again.
- Score the survivors. Candidate-letter coverage determines display order.
Everything happens as you type or change a tile. There is no “search” request and no remote answer service. That is why the result count updates immediately.
A 5-letter Wordle solver takes the green, yellow, and gray tiles from your grid, filters every valid five-letter word down to the ones that still fit, then ranks the guess that would eliminate the most remaining answers. You enter your clues after each guess and work from the shortlist it returns.
Using a 5-letter Wordle solver step by step
Standard Wordle is always five letters, so a “5-letter Wordle solver” and a plain Wordle solver are the same tool. The solver on this site is the one to use for the daily puzzle; the 4-letter, 6-letter, and 7-letter versions are for spin-offs. The loop takes a few seconds per turn once you know which field is which.
- Play a first guess in Wordle and note the colors. Use your normal opener. The solver needs real feedback first, so guess one is always yours.
- Enter green letters in their exact positions. A green tile is a confirmed letter in a confirmed spot, so the solver locks that position.
- Enter yellow letters and mark the position they are not in. Yellow means the letter is in the answer but not where you tried it: add it as required, and rule out that one slot.
- Enter gray letters to exclude them, with one exception: if the same letter is green or yellow elsewhere, it is a limited repeat, not an exclusion, so leave it out of the exclude field.
- Read the ranked shortlist and pick your next guess. Take the top word unless another tests two uncertain letters at once. The top word is the best test, not a prediction of the answer.
- Repeat after each guess. Add the new colors to what you already entered; the shortlist shrinks each turn, usually to one or two words by guess three or four.
A worked run: open with CRANE and get a green E in position five, a yellow R, and gray C, A, and N. Enter that, and the solver returns five-letter words that end in E, contain R somewhere other than position two, and use none of C, A, or N, ranked by how evenly each splits what is left. Guess the top pick, and the two or three new colors it produces usually cut the list to a single word by the next turn.
When the solver returns too many or too few words
A result that looks wrong is almost always an input problem, not a solver problem.
| Result | Most likely cause | Fix |
|---|---|---|
| No words at all | A duplicate letter entered as gray when it is green or yellow elsewhere, or contradictory clues from two guesses | Remove the letter from the exclude field; re-check each tile color against your grid |
| One impossible word | A green letter placed in the wrong box | Move the green letter to the position it actually occupies |
| Dozens of words after guess two | Yellow letters entered without marking the position to exclude, or clues from only one guess | Add the excluded positions for each yellow; enter a second guess first |
| A hundred or more words | Letters typed into the wrong field, or only gray letters entered | Put greens and yellows in their own fields, not just exclusions |
| The answer you know is missing | It is not in the solver's word list, or a clue is stricter than you think | Loosen the most recent clue; accept that rare words may be absent |
Using a solver is a personal call. Many players open one only when a streak is on the line or a puzzle has stalled at guess four, then play the next day unaided. This site's solver runs entirely in your browser, needs no account, and never sees or stores your game.
How green, yellow, and gray become rules
| Tile evidence | Constraint used by the solver |
|---|---|
| Green A in position 3 | Every candidate must have A in position 3 |
| Yellow R in position 2 | Every candidate needs R, but not in position 2 |
| Gray N with no colored N | The candidate cannot contain N |
| Blank tile | No rule is added for that position |
A yellow tile does two jobs. It requires the letter somewhere in the word and blocks it from that specific slot. If another row puts the same yellow letter in a different slot, both positions are blocked. This row-level approach is more precise than a single general “letters included” field.
The optional pattern filter accepts a letter for a known position and an underscore or question mark for an unknown one. The “must include” filter respects repeat counts, so entering SS requires two S copies. The “exclude” filter removes any word containing a listed letter.
Duplicate letters need minimum and maximum counts
For each letter in a clue row, the solver counts green and yellow copies. That positive count becomes the minimum. If the same row also contains a gray copy, the minimum becomes the maximum as well.
The green P requires one copy in position one. The gray P caps the word at one P. POINT can pass that count rule; PUPPY cannot.
Our automated tests also check the inverse case: two positive E tiles require at least two E copies. THREE passes; LATER does not. These examples reflect the actual filter, not a manually curated result.
For a player-focused explanation of these mixed colors, use the Wordle double-letter guide.
How candidates and smart guesses are ranked
After filtering, the solver counts how many remaining words contain A, B, C, and each other letter. A letter is counted once per candidate even when it appears twice. This measures coverage: how much of the current candidate set a letter can touch.
A word receives the sum of those frequencies for its distinct letters, plus a small reward for using more unique letters. Candidate answers are sorted with that score. The smart next-guess row scores the full word list against the same remaining-letter frequencies, so it can suggest a useful probe that is not itself a candidate.
Important distinction: this is not Shannon entropy, expected remaining set size, or a learned prediction model. It is a fast, transparent unique-letter coverage heuristic.
What the percentages mean
The “useful letters now” panel divides a letter's candidate count by the total number of candidates. If E appears in 40 of 50 remaining words, the panel shows 80%. It means 80% of the current list contains at least one E, not that the answer has an 80% chance of containing E.
Why a suggested guess may break hard mode
Smart guesses are drawn from the full list and optimized for coverage. They do not pass through the clue filter first. That is useful for normal-mode probes, but hard-mode players should verify every suggested word against their revealed hints.
Word data, privacy, and honest limitations
The five-letter corpus is intentionally broader than a claimed daily-answer list. That improves pattern and word-finder coverage, but it can surface an inflection, plural, regional word, or unusual candidate. The New York Times controls its own accepted guesses and daily selection and can update them.
The browser receives the word lists with the static page. React filters them in memory, and the current clue interface has no network call or storage step. Copying candidates uses the browser clipboard only when you press the copy button.
Our tests cover green, yellow, and gray positions; a gray duplicate as a maximum; two positive duplicates as a minimum; repeated include filters; rack wildcards; and standard word-game scoring. A passing test does not prove that the corpus matches another game's private list, so the site states that limitation wherever it matters.
Inspect and use the methodology
- Open WordleSolver and use “Try an example” to see the candidate set change.
- About our word data for corpus scope and product purpose.
- Editorial policy for corrections, calculations, and source standards.
- The New York Times Wordle for the live game.