Wordle Answer Statistics: What 1,994 Verified Answers Reveal

One Wordle answer a day feels like a coin toss with extra steps. Line up 1,994 of them and the coin turns out to be weighted. One letter accounts for 10.83% of all 9,970 letters on the board. S starts 300 answers on its own. X has never started one at all. Close to a third of the answers use a letter twice, and thirty-two words in this list have already been the answer more than once.

This page is the count. It covers the whole verified answer list from 19 June 2021 to 5 December 2026, 1,994 words in total, and every figure below comes from that one list. There are no reader surveys here, no guesses-per-puzzle logs, and no difficulty ratings borrowed from a publisher. The only constructed number on the page is the hardness score in section 10, and that section says plainly what it is and what it is not.

If you want the words themselves rather than their shape, the Wordle answer archive lists them by date. If you want today's puzzle nudged rather than spoiled, start with Wordle hints today. What follows is the reference layer underneath both: the frequencies you are implicitly betting on every time you type five letters.

How We Counted 1,994 Answers

The source is the same verified answer feed that powers our archive page, and it contains 1,994 dated entries running from 19 June 2021 to 5 December 2026. The list is the scheduled answer sequence rather than a scrape of social posts, which matters for one reason: a scraped list depends on somebody remembering what they played, while a scheduled list does not. It also means the later dates are the answers the schedule has already fixed, not predictions.

Nothing was sampled and nothing was deduplicated. Every entry was counted as it stands, including the words that appear more than once, because quietly collapsing duplicates would erase the finding in section 9. The list has 1,994 entries and 1,962 distinct words, and both numbers belong on the page.

What each column actually measures

• Letter frequency counts every letter of every answer as one instance. Five letters per answer across 1,994 answers gives 9,970 letters, so a count of 1,080 for E means 1,080 filled letter slots in the list, not 1,080 words containing an E. Those are very different claims and only the first one is being made.
• First letter is one observation per answer, 1,994 observations in total. Every answer contributes exactly one starting letter, so these counts sum to the number of answers.
• Last letter is counted the same way, once per answer, with the top ten reported.
• Repeated letters uses a simple test: does any letter appear at least twice in the answer? That gives 619 answers, or 31.0% of the list. Inside those 619 answers sit 633 double-letter occurrences and 14 triples. The occurrence count runs ahead of the answer count because some answers repeat more than one letter, and a triple is not the same event as a double.
• Vowel counts uses A, E, I, O and U, and nothing else. Y is treated as a consonant on every line of this page, which is the standard choice but not a neutral one, as section 6 explains.
• Rare letters counts answers that contain at least one J, Q, X or Z. That is a different unit from the letter table, and the two are reconciled in section 7.
• Common endings takes the final three letters of each answer, groups identical strings, and ranks them by how many answers share them.
• Hardness is a proxy score built from three inputs: rare letters, repeated letters, and a low vowel count. Section 10 explains the construction and its limits.

Three things this data is not

It is not the guess dictionary. Wordle accepts a far larger list of valid guesses than the set of words that can become the answer. Every table here describes answer words only. A letter that looks rare on this page can still be common in the wider guess list, and a word you have typed successfully many times may never be able to win a puzzle.

It is not a record of player behaviour. There are no completion rates, no average guess counts and no failure percentages in this dataset. Nothing here tells you how hard a word felt to people. The hardness table in section 10 is an attempt to approximate that from spelling alone, and it is labelled as an approximation for exactly this reason.

It is not an official rating. The publisher of Wordle does not publish per-word difficulty. Any page that shows you a precise difficulty number for a specific answer is either using a proxy like ours or making the number up. Ours is at least transparent about which three inputs it used.

Letter Frequency: E Leads, and It Is Not Close

Count every letter in the list and E comes out on top with 1,080 appearances, 10.83% of all 9,970 letters. A is second at 848 (8.51%), R third at 764 (7.66%), then O at 670 and T at 649. The full order is E, A, R, O, T, L, I, S, N, C, U, H, D, Y, P, M, G, B, F, K, W, V, Z, X, Q, J.

The first five letters are the backbone. E, A, R, O and T between them account for a large share of the board, and every one of them sits in the top five for a reason that has nothing to do with luck: five-letter English leans on them for vowels, for the -ER endings, and for the consonant clusters that hold a short word together. The bottom of the table is the opposite story. Z appears 38 times (0.38%), X 34 (0.34%), Q 28 (0.28%) and J 23 (0.23%).

Reading the table without overreading it

The natural mistake is to treat a frequency table as an opener ranking. It is not one. The table tells you which letters are likely to be somewhere in the answer. It says nothing about where, and position is most of the game. A letter can be the most common in the list and still be a poor first letter, which is precisely what happens to E: 10.83% of all letters, but only 3.16% of first letters. Section 3 has that table, and the gap between those two numbers is one of the more useful facts on this page.

What the table does justify is coverage. You get six guesses and five letters per guess, so the letters you name early are the ones whose status you learn early. Choosing from the top of this list maximises the chance that a letter turns green or yellow before your supply of guesses runs out. That is the whole logic behind the famous openers. CRANE is C, R, A, N and E, all of them in the top nine. SLATE is S, L, A, T and E, also all in the top nine. RAISE is R, A, I, S and E, again all in the top nine. They are not magic words. They are bundles of common letters, and the bundle is the entire trick.

There is a real cost to that logic, though, and it shows up in the middle of the table. Letters ranked tenth to fifteenth, C, U, H, D, Y and P, are common enough to matter and rare enough to be missed by an opener built purely from the top five. A guess that names five top-five letters can leave you with three greens and no idea about the consonants that actually finish the word. That is why the most useful second guess is usually not a second copy of the same idea. It is a word built from the next band down, which is where our ranked guide to Wordle starting words spends most of its time.

One more caution, because this is the number people quote out of context: 10.83% for E is a share of answer letters. It is not the share of English words that contain an E, and it is certainly not a probability that today's answer contains one. An answer contains five letters out of 9,970, and the frequency table is a statement about the pool, not about any single draw from it.

First Letters: S Opens 300 Answers

S starts 300 of the 1,994 answers, or 15.05%. C is a distant second at 171 (8.58%) and B third at 150 (7.52%). Then the slide begins: T at 132, A at 126, P at 123, F at 109, G at 105, M at 98, D at 94. By the time you reach the bottom of the table the counts are tiny. Z starts 2 answers (0.10%), Y starts 8 (0.40%), J starts 15 (0.75%), and K and Q start 21 each (1.05%).

And then there is X, which starts nothing. Not one of the 1,994 answers begins with X. That is not a rounding artefact or a gap in the data; it is a property of the answer list, and it holds across the entire date range.

What a grey S actually eliminates

When your opening guess starts with S and the S comes back grey, you have not learned that S is unlikely. You have learned that this answer is not one of the 300 words that begin with S. The other answers in the list are still live, and the rest of your board is unchanged. Players routinely over-read that grey tile, decide the puzzle is unusual, and start guessing exotic words when the answer is sitting in the most ordinary part of the list.

The same logic applies in reverse to the top of the table. S, C, B, T, A and P are the six letters that start at least 123 answers each, which is why a single well-chosen opener that tests three or four of them is genuinely informative about position one. An opener that tests none of them is not. This is the argument for starting words like SLATE and CRANE over prettier words: not that they contain common letters, but that they place a common letter where common letters actually go.

Two more readings are worth pulling out. First, the spread between the top and the middle is much larger than the spread in the letter table. S starts 300 answers while H starts 57, and nothing in the letter table falls off that sharply. Position concentrates the distribution. Second, the tail of this table is long, and it is where a lot of guesses are wasted: H starts 57 answers, N starts 35, V starts 34. If your board has ruled out the top six, the candidate set has collapsed enough that guessing by feel becomes expensive.

A caution that applies to this table more than any other on the page: it describes answers, not guesses. A word like XENON is a perfectly legal guess and a reasonable way to test X, E, N and O at once. It simply cannot be the answer, and the table is the reason you can know that before you type it.

Last Letters: E Ends 395 Answers

The end of a word is a different distribution from the start of one, and it is the half most players ignore. E finishes 395 answers, 19.81% of the list. Y finishes 285 (14.29%), T finishes 225 (11.28%), R finishes 169 (8.48%) and L finishes 139 (6.97%). Behind them sit H at 122, N at 113, D at 99, K at 93 and A at 54.

The top three final letters are doing most of the work. E, Y and T between them close the largest block of the list, and each one gets there through a different mechanism. E benefits from the silent final E and from the -ATE and -ONE families, which appear in the ending table at 13 and 11 answers. Y benefits from the -LLY group, the joint most common three-letter ending at 16, plus the many adjectives that close on Y. T benefits from -OUT at 11 and from the many five-letter nouns that close on -ANT, -ENT or -IST.

Why the ending matters more than the opening

An opening letter is a guess about where a word begins, and you get that information almost for free because most strong openers already test common starting letters. The ending is where the last guess is usually decided, and by then you have no guesses left to spend on the question. That asymmetry is the argument for knowing this table: not to open with it, but to recognise it when four tiles are already green.

Suppose your board reads four greens and one blank, and the four green letters end in E. You are now choosing from the largest ending group in the list. If the blank is the final letter and the four greens end in Y, you are in the second largest. The table gives you a prior about which final letter to try first, and a prior is exactly the right thing to use when you have no other information left. It is not an answer. If the four greens rule out every E-ending word you can think of, the prior is worthless and you should trust the letters instead of the table.

There is a second, subtler use. When you are choosing between two candidate answers that differ only in their last letter, the table is a tiebreaker with real evidence behind it. E beats Y, Y beats T, T beats R, and so on down the list. That is a better tiebreak than which word you happen to like more, and it costs nothing to apply.

One honest limit: this table is unconditional. It tells you what endings look like across the whole list, not what endings are likely given the three or four letters you have already confirmed. Conditioning beats base rates every time in this game, and a player who uses the ending table to override confirmed letters will lose more puzzles than they win. Use it when the board is silent, and ignore it when the board has spoken.

Repeated Letters: 619 Answers Break the Five-Distinct-Letters Assumption

619 of the 1,994 answers contain at least one letter twice or more, which is 31.0% of the list. Inside those answers there are 633 double-letter occurrences and 14 triple-letter occurrences. The occurrence count runs ahead of the answer count, so a handful of answers repeat more than one letter; a triple is counted as its own event rather than as three doubles.

That single number, 31.0%, is the most practically useful figure on this page after the E frequency, because it contradicts the habit almost every player has. You look at the board, you see four coloured tiles and a grey, and you start listing five-letter words with five different letters. Nearly a third of the time, that instinct is aiming at the wrong shape.

The duplicate-grey rule, worked through

Repeated letters are not just a vocabulary problem. They change what a grey tile means, and getting that wrong is one of the few ways to genuinely misread a Wordle board. Take SPEED, which has two Es at positions three and four.

• If the E at position three is green and the E at position four is grey, the answer contains exactly one E. The grey tile is telling you the second E has nowhere to go.
• If both Es are grey, the answer contains no E at all. Two greys on the same letter is a stronger statement than one.
• If one E is yellow and the other grey, the answer contains exactly one E, and it is not in the yellow position.
• If one E is green and the other yellow, the answer contains at least two Es. The yellow one has to be somewhere other than the green position.

Each of those readings is a counting statement, not a feeling. The reason players get it wrong is that they treat every grey tile as an independent "not here" mark, when a grey on a duplicated letter is really a statement about how many copies exist. Count the coloured tiles per letter before you conclude the board is broken. It almost never is.

There is a mirror-image error on the other side of the same coin: avoiding a word because it repeats a letter you have only seen once. If you have seen a single yellow E, the answer contains at least one E and may well contain two. A double-E candidate is not disqualified by a single yellow. The 633 doubles in this dataset are the evidence that those candidates are worth keeping on the list.

Vowel Counts: Two Is the Normal Shape

Count A, E, I, O and U in each answer and the distribution is sharply peaked. Two vowels is the most common shape by a wide margin: 1,158 answers, 58.1% of the list. One vowel accounts for 620 answers (31.1%), three vowels for 205 (10.3%), none for 8 (0.4%) and four for 3 (0.2%).

The practical takeaway is about opener design rather than answer selection. A four-vowel guess like ADIEU or AUDIO is popular because it tests four vowel slots in one move, and as an information play that is defensible. But four vowels is a shape that occurs 3 times in 1,994 answers. Two-vowel guesses match the majority shape and leave room for consonants, which is where the actual uncertainty usually is once the vowels are known.

The zero-vowel answers deserve a mention because they are the ones that break people. Eight of the 1,994 contain no A, E, I, O or U at all, and three contain four vowels. Both extremes exist. If your opening guess names four vowels and every one comes back grey, the zero-vowel bucket is live and worth considering instead of assuming you have misread the board.

Rare Letters: J, Q, X and Z

Four letters sit far below everything else, and they are the ones that make a puzzle feel hard when they turn up. J appears in 23 answers (1.2%), Q in 28 (1.4%), X in 34 (1.7%) and Z in 32 (1.6%).

Two counts that look like a contradiction

Z appears 38 times in the letter table but in only 32 answers here. Both numbers are correct, and the difference is the unit. The letter table counts instances: every Z tile in the list, wherever it sits. This table counts answers: every word that contains at least one Z. Thirty-two answers hold 38 Z letters between them, which is possible because a word can carry two copies. A word like JAZZY does exactly that.

The same reconciliation applies to the other three. J shows 23 in both tables, Q shows 28 in both, X shows 34 in both, which is consistent with no answer in the list holding two copies of those letters. Z is the only one of the four where the instance count and the answer count diverge, and it diverges in the direction you would expect from a letter that can double.

X has a second distinction. It appears in 34 answers and has never started one, which means all 34 of those Xs sit somewhere in positions two to five. That is a hard constraint you can use: an X in the first position of your guess is always wrong, and an X anywhere else is a genuine signal.

The strategic reading is about cost. Naming a rare letter costs a slot that could have tested S, E, T or R, and the payoff arrives only when the letter is actually present, which happens in 1.2% to 1.7% of answers. That is a bad trade as a default and a good trade as a response. Once the board shows a J or a Z in yellow, the candidate set collapses fast and finding its exact position becomes the highest-value thing you can do. Spend the slot then, not before.

Common Endings: LLY and ING Lead

Group the final three letters of every answer and a small set of families emerges. LLY and ING tie at the top with 16 answers each, then TCH and TER at 15, ATE and VER at 13, ILL at 12, and OUT, ONE and ASH at 11 each.

These families matter because they are candidate generators. When four tiles are green and the blank is near the end, you are usually not searching the whole dictionary. You are choosing between a handful of words that share a suffix, and knowing which suffixes are productive tells you which handful to check first. The -TCH group alone contains a dense cluster of five-letter words built on a short vowel plus TCH, and the -ILL and -ASH groups work the same way.

The ceiling on this method is worth stating, because it is lower than it looks. These ten families are the top of a long tail, and most of the 1,994 answers end in something that is not on this list at all. The table is a checklist to run through, not a prediction to lean on. If none of the ten fits your four greens, the correct conclusion is that your answer is in the tail, not that you have made a mistake.

Answers Used Twice: The "Never Repeats" Rule Is Not True

The list contains 1,994 entries and 1,962 distinct words. Thirty-two words appear more than once, and the arithmetic closes neatly: 1,962 plus 32 equals 1,994, which means each of those thirty-two appears exactly twice rather than one word appearing three times.

This is the section that contradicts received wisdom. "Wordle never repeats an answer" is repeated so often that players treat it as a rule of the game, and use it to cross candidates off. In this dataset it is simply false. Thirty-two words have already been the answer twice inside the covered date range.

• ACUTE
• AGREE
• AWAKE
• BATON
• BLURT
• BREAK
• BRING
• CIGAR
• CLICK
• CLOCK
• COMET
• FEIGN
• FIRST
• FLUME
• FOCUS
• FORTH
• GRADE
• GRIPE
• IVORY
• LINEN
• MAJOR
• MOTOR
• PROVE
• SANDY
• SMILE
• SPICY
• SQUAD
• STAND
• STINK
• STOUT
• THUMB
• TRACE

We are not going to guess at the mechanism behind the repeats. The list may have been extended, the schedule may have been rebuilt, or the curation may simply allow a word to return after a long gap. What we can show is the list itself, with the duplicates intact, and the honest conclusion is narrower than a theory: as a rule for this answer list, "never repeats" does not hold.

That has a direct consequence for how you use an archive. If you are choosing between two candidates and one of them already appears in the archive, that is not a reason to drop it. ACUTE, AGREE, AWAKE, BATON, BLURT, BREAK and BRING have each been the answer twice. Treating a previous appearance as disqualifying would have eliminated the correct answer on those days. The answer archive is still the right tool, but use it to check what a word is, not to eliminate what it can be.

Hardest Answers, by Our Own Proxy

Before the table, the caveat, because it belongs in front rather than buried under it. The score below is our construction. It is not an official difficulty rating, it is not derived from player results, and the publisher of Wordle does not release per-word difficulty data. It takes three inputs, rare letters (J, Q, X, Z), repeated letters, and a low vowel count, and reports them as a single score where higher means harder by that definition. It knows nothing about meaning, familiarity, or how easy a word is to recall under time pressure.

You can watch the formula work down the list. JAZZY scores 12 and is the only entry in double figures: one J, two Zs, a Y, and a single vowel in the middle. TIZZY, FUZZY, FIZZY and DIZZY all score 9 on the same skeleton, a rare consonant plus a doubled letter plus one vowel. PIZZA scores 8 with its double Z and two vowels. Then the list steps down to 6 for JOLLY, JIFFY and JELLY, which trade the Z for a J and keep the double and the single vowel, and to 5 for the Q-words and double-O shapes, WOOZY, QUILL, QUEUE, QUELL, QUEER and QUEEN.

That coherence is the point and also the problem. A constructed index will always find words that match its own definition, so a table full of Zs and Js and doubled letters is not a discovery about Wordle. It is the definition of the index, printed out. The useful question is not whether JAZZY is the hardest answer, because under this formula it trivially is. The useful question is whether the formula captures anything a player would recognise, and on that it is only partly successful.

What it misses: a word can be hard because its meaning is obscure rather than its spelling, and this score would give it the same number as an easy word with identical letters. A word can be hard because a common letter is doing something unexpected, and the score does not know about that. A word can be easy because it is famous, and fame is invisible here. What the score does capture is spelling awkwardness, and that is a real component of difficulty even if it is not the whole of it. Read the table as a list of unusual spellings, which is what it is.

What This Means for Players

Five conclusions survive contact with the numbers, and each one changes a decision you make during an actual puzzle.

1. Open with common letters in common positions, not just common letters. The letter table says E, A, R, O and T. The first-letter table says S, C, B, T, A and P. The two lists overlap at A and T, and that overlap is where the strongest openers live. S, T, A, R and E is SLATE. C, R, A, N and E is CRANE. Both put a common letter in the position where common letters actually occur. If you want the ranked version with the reasoning behind each entry, the best starting words guide covers the trade-offs in more detail than a frequency table can.

2. Budget for a repeat before you need one. 31.0% of answers use a letter twice, and there are 633 doubles and 14 triples in the list. If you reach your fourth guess with five distinct letters and nothing fits, the problem is probably the shape of your candidates rather than your vocabulary. Try a double before you try an obscure word.

3. Assume two vowels, and stop spending guesses on vowels. 1,158 answers have exactly two, 620 have one, and only 205 have three. A four-vowel opener is an information play, not an answer guess. Once you know the vowels, the remaining uncertainty is almost always in the consonants, which is where the next guess should go.

4. When four tiles are green, generate by ending family. LLY, ING, TCH, TER, ATE, VER, ILL, OUT, ONE and ASH are the ten most productive three-letter endings. Run them as a checklist before you guess blind. If none of them fits, you are in the long tail and the table has stopped helping, which is itself useful to know.

5. Treat "never repeats" as folklore and check the archive instead. Thirty-two words in this list have already been the answer twice. If a candidate of yours appears in the archive, that is not evidence against it. It is evidence that the rule you were taught is not a rule. The archive is the place to confirm a word, and the Wordle solver is the place to test a board against the full candidate list when the ending families run out.

None of this replaces the feedback loop. Every table on this page is a statement about the answer pool before the puzzle starts. Once you have coloured tiles, the board outranks the statistics, and a player who follows a frequency table past three greens is using the data wrong. The tables are there to make the first two guesses and the last one better. The middle belongs to the letters you have actually confirmed.