Loading...
Loading...
Contributors:
Dataset URI:
10.5281/zenodo.23535961Which first guess solves a five-letter word game fastest? Instead of counting letters, this dataset plays every game: a fixed solver started from 113 opening words and played each of 1,660 common English five-letter answers to the end — 187,580 complete games — and records how many guesses each took. It adds first-guess statistics (entropy, expected remaining answers, largest remaining group) for all 8,636 five-letter words of the ENABLE word list.Findings. TRACE is the best of the 113 openers played: 3.32 guesses on average, 64.8 percent of games solved in three guesses or fewer, ahead of CRATE, CRANE and SLATE. The most popular vowel-heavy openers come last: AUDIO ranks 113 of 113 (3.55 guesses), ABOUT 112 and ADIEU 111. Over 100 games AUDIO costs about 23 more guesses than TRACE, because after AUDIO 129 candidate answers remain on average, after TRACE only 48. The most informative first guess is not the fastest solve: RAISE splits the answers best after one guess (5.92 bits) but ranks 39th in full play-outs. No opener failed to solve any answer within six guesses. Words with many look-alikes that differ in one letter are harder: 3.30 guesses on average over the ten best openers for answers with at most one look-alike, 3.55 for answers in a family of five or more. The largest families are _IGHT (9 words), _OUND (8), _ATCH (7), _ERRY (7) and SHA_E (7). The hardest single answers, JOKER, GIVER, JOKED, POKER and YUMMY, sit in small families but contain letters the best openers do not test.Method. Answers are ENABLE five-letter words with a wordfreq Zipf frequency of at least 3.2, excluding every word that ends in a single S (plurals, verb forms and about two dozen others such as FOCUS) and offensive words — our own approximation of common words. No publisher's answer list was used. Feedback follows the game's rules for repeated letters and is self-tested against 30,000 random pairs. The solver plays normal mode and picks the maximum-entropy guess over the remaining answers (ties: possible answer, more frequent word, alphabetical). The script that produces every file in one run is included.Limits. The answer set is frequency-based and contains some words that are mostly names in everyday text. The solver is a strong greedy strategy, not the proven optimum. All answers are weighted equally.The dataset was published by WordBreakfast, a site with daily word-game hints and word tools. The results are explained for players at wordbreakfast.com/best-wordle-starting-words. Wordle is a trademark of The New York Times Company. This dataset is independent and not affiliated with or endorsed by it.
Keyword(s):
entropy; five-letter words; information theory; open data; puzzle solver.
MeSH Term(s):
Grants and Funding:
No Grants or Funding Data Available
Associated Publication(s):
No Associated Publication Data Available
Supplementary Data:
Related Dataset(s):
No Related Dataset Data Available
The National Library of Medicine (NLM) provides information about biomedical datasets. Inclusion on an NLM Web page does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: NLM Web Policies (nih.gov)