How seed quality works

A valid letter set is not automatically a good puzzle. Bloomword scores thousands of candidates, rejects unsafe sets, and reserves the shared Daily for the strongest results.

The five-part score

Each candidate receives a quality score from 0 to 100 using five weighted ingredients:

• Headline familiarity (25%). Is the full anagram a word most players are likely to recognize?
• Approachability (25%). Can an everyday vocabulary reach the number of finds needed to rank?
• Target richness (20%). Does the set have a satisfying number of familiar answers for its size?
• Vowel balance (15%). Is there enough variety to avoid a consonant slog or an all-vowel giveaway?
• Letter friendliness (15%). Do difficult letters such as Q, X, Z, or J create useful words rather than dead weight?

The weighted result is reproducible: the same dictionary and inputs produce the same score.

Safety is a gate, not bonus points

Bloomword also applies a curated profanity and sensitivity blocklist. A blocked word is removed from playable solutions and cannot become a Daily headline or target. Safety does not merely lower a candidate's score; a hit rejects it from Daily consideration.

Only the top pool becomes the Daily

A generated seed must first meet the basic word-count and full-anagram rules. The shared Daily then draws only from candidates scoring at least 80. Practice can use the broader generated pool because it is private, repeatable play rather than the one puzzle presented to everyone.

Automated scoring is a filter, not an excuse to stop looking. Candidate calendars can still be inspected and individual days vetoed when a technically valid puzzle reads poorly to a person.

Next: read how we pick the Daily or how scoring works.