How the word list works

Get Unscrambler is a letter-inventory search, not a claim that every returned word is legal in every game. The solver answers a precise question: which entries in its bundled lists can be built from the letters and wildcards supplied?

That distinction matters. Letter accuracy can be tested objectively, while dictionary acceptance depends on the puzzle, classroom, competition, language variety, and edition in use. We show candidates and explain our matching process; the rules for your activity remain the final authority.

THE MATCHING PIPELINE

From typed letters to ranked results

1. 1

   Normalize the rack

   Letters are converted to a consistent case. Spaces, punctuation, and unsupported characters are removed. Question marks are preserved as wildcard tiles, and input is limited to fifteen positions.

2. 2

   Build a frequency count

   The browser records how many copies of A through Z are available. Repeated letters remain repeated evidence; they are never collapsed into a set of unique characters.

3. 3

   Test every candidate locally

   Each candidate between two letters and the rack length is checked against the frequency count. When a candidate needs a missing copy, one wildcard is consumed. A result fails as soon as it needs more wildcards than the rack provides.

4. 4

   Rank long results first

   Matches are ordered by length. Within the same length, entries in the size-60 Core subset are prioritized, followed by alphabetical order for stable results. Core reflects an ESDB size tier; it is not a measured frequency ranking.

5. 5

   Apply your confirmed clues

   Exact length, starting letters, ending letters, contained letters, and word category are applied after matching. Filters change which valid candidates are visible; they do not add letters to the rack.

SOURCE AND COVERAGE

Pinned source, selection, and limits

Both generated files come from the English Speller Database / SCOWL repository, pinned to release rel-2026.02.25 at commit 7e99edab8e32f9f9ea2b15f249ca8d4d67237410. We generate size 80 for the broad list and size 60 for the Core subset. The source describes size 80 as its broad current-use/word-game layer; this is not a promise that every word is accepted by a particular game. The upstream project documents dialect and commonness metadata, but its size tiers are not a measured frequency ranking for our site. See the upstream ESDB/SCOWL documentation, the pinned source snapshot, and the complete source notice and license text.

We request American and British spellings (source spellings A and B; regions US and GB), include variant levels through 5, omit abbreviations, and omit categorized special lists. We then keep only lowercase ASCII alphabetic forms of 2–15 letters. That intentionally excludes proper-name capitalization, abbreviations, punctuation, apostrophes, accented spellings, and open or hyphenated compounds; we do not strip punctuation or lowercase names to manufacture entries. Both outputs are deduplicated and bytewise sorted. The size-60 Core list is a subset of the size-80 list. No unidentified old entries are mixed in.

The generated broad list has 190,518 entries (1,933,112 bytes); Core has 82,126 (782,284 bytes). SHA-256 checksums and extraction settings are in the word-data manifest. The exact reproducible command is pnpm generate:word-data. Upstream notices are distributed with the assets and reproduced verbatim in the linked notice. These lists are not complete English, not human-curated word-by-word, and not official Scrabble, Wordle, or other game dictionaries.

For example, QI is in the size-80 All words list but not the smaller size-60 Core list. QADI and QAT are in the broad list. The site's Q-without-U and no-standard-vowels collections are generated from the selected output and can therefore change when the pinned upstream source changes.

SEARCH URL POLICY

Useful filters without a page for every combination

The site publishes a finite set of curated guides and word lists. The solver can take a starting letter, ending, contained letters, or exact length from a small number of contextual links, but those query-string states remain part of the root tool and keep the root canonical. Arbitrary filter combinations are not generated as indexable landing pages and are not placed in the sitemap. This keeps browser navigation and useful clue handoffs while avoiding an unlimited crawlable URL set.

WHAT THE LIST CONTAINS

Two bundled word files

words.txt contains 190,518 lowercase alphabetic entries generated at ESDB size 80. common.txt contains 82,126 entries generated at size 60 and is fully contained in the broad list. The UI calls this smaller subset “Core words.” It is neither a frequency ranking nor a guarantee of everyday familiarity.

US and GB regional variants are requested, so forms such as COLOR/COLOUR and CENTER/CENTRE can both appear. Some entries remain uncommon or dictionary-dependent. A result means only that the word occurs in this pinned reference list and matches the rack—not that a named game accepts it.

Proper nouns, hyphenated expressions, multiword phrases, and punctuation-dependent forms are outside the current single-word workflow. The solver also does not provide definitions, pronunciations, or game-specific scores. Its job is candidate generation from a controlled inventory.

WORKED INVENTORY CHECKS

How matching decisions are made

PRIVACY & PERFORMANCE

The rack is solved in your browser

The two word-list files are downloaded as site assets. Matching and filtering then happen on the visitor's device, so the letters entered for a normal solve are not submitted to Get Unscrambler for processing. Favorites, solve count, and daily-puzzle progress may be stored locally in that browser.

This design keeps the core search responsive after the lists load and limits the information needed by the solver. Hosting, analytics, advertising, and contact-message processing are explained separately in the Privacy Policy.

REVIEW & CORRECTIONS

How we investigate a reported word

First, we reproduce the exact rack and filters. Second, we compare the candidate's letter frequency with the rack. Third, we check whether the issue concerns matching, ranking, the bundled list, or a game-specific acceptance rule. We correct mechanical matching errors and misleading explanations; we do not claim that one general list can replace every official dictionary.

To report a result, include the entered letters, every active filter, the expected or questionable word, and the relevant dictionary or game when one is involved. Send it through the Contact page.

QUESTIONS

Word list and solver FAQ