Transparent methods
How the words and movie puzzles work
The site separates licensed source data, offline preparation, tested matching logic, and small browser interfaces so each part can be checked independently.
Word source and commonness
Version 1 uses American English output from the English Speller Database (ESDB, previously SCOWL), at source sizes 35, 50, and 60. Membership in the smaller output becomes the “common” tier; size 50 becomes “familiar”; the remaining size-60 entries form the broader list.
The offline pipeline removes entries outside 2–15 alphabetic characters, normalises diacritics for matching, removes duplicates, splits files by word length, and records SHA-256 checksums in a versioned manifest.
Letters, positions, and wildcards
A fixed-position pattern compares each known character directly. Included letters use multiplicity-aware counts, so requiring two copies needs two copies in a result. Excluded letters reject a word when any copy appears.
The unscrambler counts every available letter and lets each ? wildcard cover one missing letter. Exact anagrams require identical letter counts after spaces and punctuation are removed.
Movie selection and daily schedule
Movie metadata is imported manually from Wikidata, cached, normalised, validated, and committed as a static dataset. English title, release year, genre, director, stable Wikidata ID, and sitelink count are required. Adult-content genre labels and malformed records are excluded.
The daily game uses America/New_York calendar dates and a deterministic date-to-puzzle mapping. Streak logic compares date keys rather than subtracting fixed 24-hour intervals, which avoids daylight-saving-time errors.
Why there are no movie images
The game is intentionally based on title letters and factual metadata. Avoiding posters, frames, clips, audio, and logos reduces copyright and performance risk while keeping the puzzle accessible.
Updates and limitations
Word, movie, and video-game datasets are versioned and updated manually after validation. Spelling lists cannot represent every specialist term or every third-party game rule. Wikidata coverage and sitelink counts change over time, and difficulty remains an explainable estimate rather than an objective fact. Title records receive automated checks and periodic editorial review.