How we check words

Every word list on this site makes a promise: that it has been checked rather than scraped. This page explains exactly what that means, which dictionary we use, how our figures are calculated and where the limits of our claims begin.

You should not have to take the accuracy of a word list on trust. The method should be clear enough for you to judge it yourself.

The dictionary behind the game

Find the Word validates every entry against a single fixed list derived from ENABLE — the Enhanced North American Benchmark Lexicon, a public-domain word list widely used in word games and language software.

Our version has been filtered to entries of three letters or more and prepared for use inside the game.

The measured figures for this version are:

• 172,727 entries in total.
• The shortest accepted length is 3 letters, with 972 entries.
• The longest accepted entry contains 28 letters.
• The most common length is 8 letters, with 28,420 entries.
• Every retained entry consists only of lowercase letters from a to z.
• No retained entry contains a capital letter, space, hyphen, apostrophe, number or accented character.

These figures count dictionary entries rather than independent word families. Inflected forms are counted separately, so aah, aahs, aahed and aahing are four different entries.

The list also contains archaic, dialectal, informal and highly specialised vocabulary. Words such as zek, wae, kae and pfft are all present.

The source list is designed to exclude proper nouns, hyphenated compounds, contractions and entries that require punctuation, capitalisation or accented characters. We also verified that every entry retained in our game file consists only of lowercase English letters.

Some spellings can have more than one meaning. A form commonly recognised as a name may also have a separate lowercase dictionary sense. The game checks the spelling against the lexicon; it cannot determine which meaning a player intended.

How the figures are calculated

All counts on this site are generated programmatically from the same plain-text dictionary file used by the game.

Each line in the file represents one accepted entry. From that file, we can calculate:

• the number of entries beginning with each letter;
• the number of entries at every word length;
• the number of words beginning with a particular two-letter combination;
• complete subsets, such as all three-letter Z-words or all Q-words without a U.

Opening-letter totals are calculated from the first character of each entry. Word-length totals are calculated from the number of characters on each line.

Because the calculations use the same fixed file as the game, an article describing the Find the Word dictionary should produce the same result every time unless the dictionary itself is updated.

The game accepts no two-letter words

This deserves its own section because it surprises experienced players and changes the strategy of difficult rounds.

Our copy of the lexicon was filtered to entries of three letters or more. As a result, every two-letter word is rejected by Find the Word, even when another word game accepts it.

Much traditional word-game advice recommends memorising valuable two-letter forms such as qi, za, xi, jo, ka, ax and ox. In games that use the appropriate lexicon, that can be excellent advice.

None of those words work here. The shortest possible answer in Find the Word contains three letters.

That means the shortest escape route is one letter longer than it is in many tile-based games, making the scarcest opening letters noticeably harder.

Three-letter escape routes for difficult letters

These are the complete three-letter sets accepted by the current game dictionary for six strategically difficult opening letters.

Q — 2 entries

qat, qua

X — 1 entry

xis

Z — 12 entries

zag, zap, zax, zed, zee, zek, zig, zin, zip, zit, zoa, zoo

J — 23 entries

jab, jag, jam, jar, jaw, jay, jee, jet, jeu, jew, jib, jig, jin, job, joe, jog, jot, jow, joy, jug, jun, jus, jut

V — 21 entries

vac, van, var, vas, vat, vau, vav, vaw, vee, veg, vet, vex, via, vie, vig, vim, vis, voe, vow, vox, vug

K — 27 entries

kab, kae, kaf, kas, kat, kay, kea, kef, keg, ken, kep, kex, key, khi, kid, kif, kin, kip, kir, kit, koa, kob, koi, kop, kor, kos, kue

W is more generous than the groups above, with 38 three-letter entries in the same dictionary snapshot.

The most severe case is X. If an X round appears in Find the Word, xis is the only three-letter answer available in the entire game dictionary.

How rare each opening letter actually is

We counted every entry according to its first letter. These are the real totals used in the game rather than estimates based on general English frequency.

Because inflections and plurals are counted separately, these figures represent dictionary entries rather than unique roots. For example, related forms such as zone, zones, zoned and zoning count separately when present.

• Most generous: S with 18,985 entries, C with 16,229, P with 14,563 and A with 10,440.
• Middle ground: W with 3,692 and V with 2,735.
• Genuinely difficult: K with 1,715, J with 1,375, Q with 822, Y with 544 and Z with 532.
• Hardest by a clear distance: X, with only 134 entries.

X accounts for approximately 0.078% of the entire dictionary. There is roughly one X-entry for every 142 entries beginning with S.

Two findings run against common expectations.

Z is not the hardest letter in this dictionary. X has roughly one quarter as many entries. Y is also scarcer than J, even though Y feels like a much more ordinary letter until a player is asked to begin a word with it.

The Q-words without a U

These are all the Q-entries in the current dictionary that do not contain a U immediately after the opening letter:

qaid, qaids, qanat, qanats, qat, qats, qindar, qindarka, qindars, qintar, qintars, qiviut, qiviuts, qoph, qophs, qwerty, qwertys

There are seventeen entries in total.

Qwerty is included, which is the kind of result that can easily be missed when a list is built from memory rather than checked against the actual file.

Two famous word-game forms are absent. Qi is rejected because Find the Word requires at least three letters. Qis is not included in this version of the ENABLE-derived dictionary.

What “valid” means on this site

Word-game dictionaries do not agree on every entry. ENABLE is not identical to the official North American competition lexicon or the international Collins word list. Each is maintained separately and may accept words that another rejects.

For that reason, we avoid phrases such as “valid everywhere.” No single word list can support that claim.

Instead, we use the following language:

• Accepted by our game means the word has been checked against the 172,727-entry file described on this page.
• Varies by dictionary means acceptance differs between established word-game lexicons and should be confirmed against the rules of the game being played.
• Not accepted here means Find the Word rejects the entry, even when another game or dictionary may accept it.

“Accepted by our game” is the only claim we can make with complete certainty, so it is the one we lead with.

How our word lists are built

1. Start with observed play patterns.

   We first identify the words and word families that are useful under time pressure, based on playing and testing the game. This is why our articles focus on practical recall routes rather than presenting thousands of entries in alphabetical order.

2. Run every entry against the game dictionary.

   Each suggested word is checked against the fixed 172,727-entry file. Anything rejected is removed or explicitly identified as accepted elsewhere but not by Find the Word.

3. Describe the level of certainty.

   Words are labelled using the acceptance language explained above. We distinguish between words verified in our game and words whose status varies across competing lexicons.

4. Keep only what serves the article.

   We generally remove highly specialised entries unless they are short, strategically useful, unusually memorable or necessary to explain a pattern in the data.

5. Check the final page again.

   Before publication, the article is reviewed for rejected entries, duplicated words, inconsistent labels, incorrect totals and claims that go beyond what our dictionary can establish.

What we do not do

We do not publish exhaustive lists for their own sake. Word-finder tools are better suited to returning hundreds or thousands of matching entries. A human reader cannot memorise all 822 Q-words for use in a sixty-second round.

We do not replace dictionaries. Definitions and short glosses are included as memory aids and are written in our own words. For complete definitions, pronunciation, usage history and grammatical information, readers should consult an established dictionary.

We do not claim that curated lists are complete. A curated list is deliberately narrower than the underlying dictionary. Its purpose is to select useful or memorable entries rather than reproduce the entire file.

We do not imply that our findings apply to every word game. Comparing tens of thousands of entries across several independently maintained lexicons is outside the scope of this site. When an article discusses a different game or dictionary, the distinction is stated clearly.

We do not hide the limits of the data. A word’s presence in the dictionary proves that the spelling is accepted by Find the Word. It does not prove that the word is common, modern, suitable for every context or accepted by every competing lexicon.

Etymology and language history

The articles in Spelling & Etymology make historical claims about subjects such as borrowed vocabulary, sound changes, the printing press, the Great Vowel Shift and the origins of individual words.

Those claims are checked against recognised dictionaries, etymological references and works on the history of English. Each completed historical article includes its own sources or further-reading section.

Where specialists genuinely disagree, the uncertainty is stated rather than replaced with the neatest explanation.

The history of zebra is one example. The word reached English through Portuguese, but its earlier origin remains disputed. A disputed answer is less satisfying than a clean one, but it is more accurate.

Corrections

Errors can be reported through the contact page. Every message is read.

Typographical mistakes are corrected when found. Substantive corrections — such as an incorrect accepted word, a wrong total or a false historical claim — are noted at the foot of the affected article with the date of correction.

We do not quietly remove a meaningful error and pretend it was never published.

A site about words that refuses to acknowledge a mistake about words is not worth reading.

Who writes and checks the site

Everything on The Verbarium is written and checked by Kleoniq Kosova, who also built Find the Word.

There is no editorial team and there are no outside contributors. The site represents the work of one person, which is worth knowing when you decide how much confidence to place in what you read.

You can read more about the purpose of the publication on the About page.