Word game statistics: what 3,858 answers actually look like

An original measurement of every answer list used on this site, across five board sizes. Last recalculated 22 August 2026.

Why this page exists. Almost every "best Wordle starting word" article on the internet quotes the same handful of numbers, usually derived from general English text rather than from the list the game actually draws from. Those are different things. This page measures our answer lists directly - all 3,858 of them - so the numbers below describe the words you will genuinely be asked to guess here.

The headline finding: repeated letters scale with length

The most consistent piece of Wordle advice is "assume every letter is different". Measured across board sizes, that advice has a shelf life:

BoardAnswers with a repeated letterAnswers with a double (adjacent) letter
3 letters8.4%5.4%
4 letters19.7%13.2%
5 letters28.8%12.3%
6 letters52.0%17.2%
7 letters67.5%18.8%

At three letters, assuming no repeats is right 91.6% of the time. At seven letters the same assumption is wrong more often than it is right. The jump between five and six letters is the steepest - it nearly doubles - which is exactly where most players carry their five-letter instincts across and start losing rounds they should win.

How big each answer list is

Every board separates answers (words that can be the hidden target) from accepted guesses (words you are allowed to type). The gap is deliberate: you can spend a guess on an obscure word purely to test letters, but you will never lose to one.

BoardPossible answersAccepted guessesRatio
3 letters2992,1261 : 7.1
4 letters7757,1791 : 9.3
5 letters86115,9091 : 18.5
6 letters97729,8701 : 30.6
7 letters94641,9941 : 44.4

The ratio matters strategically. On the 7-letter board there are 44 accepted guesses for every possible answer, so "wasting" a guess on a word that cannot win costs you very little and buys a lot of information. On the 3-letter board, where the ratio is 7 to 1, that trade is much worse and you should be trying to win from guess two.

Letter frequency by board size

These figures are the share of answers containing each letter at least once, which is the number that matters when you are deciding what to test. They are not counts of total letter occurrences.

Rank3 letters4 letters5 letters6 letters7 letters
1A 34.1%E 43.7%E 51.2%E 68.6%E 66.2%
2T 22.7%A 35.4%A 41.2%R 46.1%I 52.7%
3E 22.4%O 27.6%R 40.0%A 41.2%R 48.4%
4O 21.7%L 24.8%T 31.7%T 36.5%A 46.8%
5I 20.7%T 24.0%O 31.0%N 35.1%N 46.7%
6P 16.1%R 22.3%I 27.8%I 35.1%T 42.7%
7N 16.1%I 22.2%L 26.9%O 30.9%O 34.2%
8D 14.7%N 20.4%S 25.3%L 30.4%L 31.8%
9R 13.7%S 19.5%N 23.9%S 27.3%S 29.7%
10Y 11.7%D 17.8%D 20.1%D 26.8%C 27.9%

Three things stand out. A leads at three letters but E takes over from four onward - short words lean on A because of the huge number of consonant-A-consonant forms. E peaks at six letters (68.6%), not seven, because six-letter words are dense with -ER and -ED endings. And I jumps to second place at seven letters (52.7%), which is almost entirely the -ING suffix at work.

First and last letters

Openings are far less predictable than endings on every board. The most likely first letter never exceeds 14.3%, while the most likely last letter reaches 22.2%.

Most likely first letter

Board1st2nd3rd
3 lettersA 8.4%B 8.0%T 8.0%
4 lettersS 9.9%T 7.5%B 7.2%
5 lettersS 14.3%C 8.5%B 8.0%
6 lettersS 12.5%C 8.4%P 7.3%
7 lettersC 11.9%S 9.7%P 9.5%

Most likely last letter

Board1st2nd3rd
3 lettersT 14.4%N 11.0%D 9.7%
4 lettersE 22.2%T 11.7%D 9.0%
5 lettersE 19.9%T 11.8%Y 10.1%
6 lettersE 17.5%R 15.1%D 13.7%
7 lettersE 18.6%G 16.5%D 13.4%

The 7-letter row is the outlier worth memorising: G is the second most likely final letter, at 16.5%. No other board has a consonant that high in final position, and it exists for one reason only.

The -ING effect

Common endings are worth testing directly rather than deducing. Measured as a share of each answer list:

Ending5 letters6 letters7 letters
-ING1.2%5.6%16.3%
-ED4.3%9.5%11.8%
-ER6.5%11.9%6.8%

At seven letters, 154 of the 946 answers end in -ING. Combined with -ED, more than a quarter of the list ends in a suffix you can test in a single guess. This is why solving the last three tiles first is often faster than working out the stem on the long boards, and it is the biggest single edge available anywhere on this site.

Vowel structure

How many vowels (A, E, I, O, U) a typical answer contains, by board:

Board1 vowel2 vowels3 vowels4 vowels
3 letters81.6%15.4%--
4 letters48.9%50.2%0.6%-
5 letters27.3%64.1%8.4%0.2%
6 letters5.4%64.6%29.1%0.8%
7 letters0.8%38.9%53.5%6.8%

The practical rule: below five letters, hunt for the vowel. From five letters up, assume there are at least two and stop testing vowels once you have found them - your remaining guesses are better spent on consonants.

Highest-coverage opening words

We ranked every answer in each list by how much of that list its distinct letters cover. This is a letter-coverage score, not a full game-tree solve, so treat it as a strong heuristic rather than a proof of optimality.

BoardTop single openers
3 lettersATE, EAT, TEA, ANT, APT
4 lettersLATE, TALE, EARL, REAL, RATE
5 lettersALERT, ALTER, LATER, ARISE, RAISE
6 lettersRETAIN, RENTAL, RETAIL, LINEAR, REASON
7 lettersCERTAIN, TRAINED, ARTICLE, SENATOR, ROUTINE

Best two-word openings

A stronger routine than any single word is a fixed pair that shares no letters, so the second guess wastes nothing:

BoardPairDistinct letters tested
3 lettersATE then ION6 (4 vowels)
4 lettersLATE then IRON8 (4 vowels)
5 lettersALERT then SONIC10 (4 vowels)
6 lettersRETAIN then SHOULD12 (all 5 vowels)
7 lettersLASTING then PRODUCE14 (all 5 vowels)

RETAIN then SHOULD is the standout of the whole set: two ordinary words, no shared letters, covering twelve distinct letters and every vowel. On the 7-letter board, CERTAIN then PROUDLY is also excellent - it overlaps only on R and covers thirteen letters - with the bonus that LASTING tests the -ING ending on move one.

How these numbers were produced

Every figure on this page was computed directly from the word bank files this site ships to your browser, so they describe the exact lists in play. Three notes on method, so you can judge the numbers fairly:

Word lists are compiled from open-licence sources credited on the about page. If you spot an error in any figure here, let us know and we will recheck it.

Questions

Why do your numbers differ from other Wordle statistics pages?

Most published figures are computed over general English text or a full dictionary. Ours are computed over the curated answer list this game actually draws from, which is a much smaller and more common-word-heavy set. Both are correct for what they measure; ours is the one that predicts what you will face here.

Does knowing this make the game less fun?

In our experience it does the opposite - it moves the challenge from guessing to reasoning. If you would rather play cold, skip the opener tables and just use the repeated-letter rates, which mostly stop you making mistakes rather than telling you answers.

Which board is hardest?

Seven letters, by a clear margin: the fewest structural certainties, the highest repeated-letter rate (67.5%) and the same six guesses. Six letters is the biggest step up in difficulty though, because the repeat rate nearly doubles from five while most players' habits do not adjust.

Do these figures change?

Only if we change the answer lists, which is rare. The date at the top of this page is when the figures were last recalculated.

Put this to work - play a round