About Wordtallyo

Most word games hand you positions. You guess, and the board colours in which letters were right and where. That is a pleasant puzzle, and it is largely solved by the fourth guess.

Wordtallyo gives you two numbers and nothing else: how many letters landed in the right place, and how many are in the word but somewhere else. Which ones is never revealed. The deduction is entirely yours.

• Easy — 4 letters, 10 guesses. Short words leak the least: only 15 tallies exist for four letters, against 28 for six. Easy gets the most room, and no score at all — just the word.
• Regular — 5 letters, 8 guesses. The standard board. Unique letters in every answer, so a repeated letter in your guess is always a wasted slot.
• Hard — 6 letters, 8 guesses. Answers here can repeat a letter, and the tally will not tell you when they do.

Guess limits are not guesses. They were measured by running two solvers over every answer in every pool and taking the ninetieth percentile of a solver that searches the whole dictionary, plus one.

That measurement turned up something counterintuitive worth stating plainly: the four-letter board is the hardest. Four letters admit only fifteen distinct tallies where six admit twenty-eight, so each guess carries about a third less information. Easy is called Easy because the words are short and there are more guesses, not because it asks less of you.

One run per puzzle, per player. There are no restarts and no second attempts, and the server keeps the answer to itself until your run ends.

There is a ceiling on that, and it is worth being honest about: someone can solve the board in a private window and then replay a perfect run on their real account. No client-side check fixes it. The mitigation is social rather than technical — no leaderboard here will ever carry a prize. Ranking is for pride, and pride is not worth cheating for.

Assist mode counts the words still consistent with your feedback. That is a solver, so any run using it is flagged and ranked on its own board. It runs entirely in your browser and never sees the answer.

The palette is blue and amber rather than green and yellow. Green/yellow reads as a clone of another game, and it is the worst possible pair for the roughly eight percent of men with red-green colour deficiency in a game where colour carries meaning.

Every state also carries a shape, so nothing depends on colour alone. The board is fully keyboard operable: type to guess, arrow keys to move between marked letters, Space to cycle a mark. Each guess is announced to screen readers with its result.

• ENABLE — the list of valid guesses (public domain)
• Princeton WordNet 3.0 — definitions, and the base-form check that keeps inflections out of the answers (WordNet licence)
• Norvig web-corpus unigrams — one half of the familiarity ranking (MIT)
• OpenSubtitles frequency lists — the other half, so answers skew toward words people actually say (CC BY-SA 4.0)
• LDNOOBW — the blocklist that keeps slurs out of the answer pools (CC0)

Answers are drawn from a curated pool: a word only qualifies if WordNet knows it as a base-form lemma with a usable, non-proper-noun definition, and it sits inside a frequency ceiling blended from both corpora. Proper nouns, plurals of words already in the pool, participles, and British spellings whose American form is also a word are all excluded.

The notebook has the measurements behind the claims on this page — every legal opening guess ranked, what a tally costs you against coloured tiles, and why the easy board is the hardest. New here? Start with the rules in two minutes.