Methodology

How a guess is checked

A guess must be exactly five letters and must already be in one of our two lists: the 1,994-word answer pool or the 2,942 extra words accepted as guesses. A real English word that is in neither list is rejected. That is a property of the list, not a judgement about the word.

You can test any five-letter word on the valid words page.

How the tiles are scored

Each of the four boards hides a different word, and your guess is scored against each board separately. That is why the colours differ from board to board.

• Green: the letter is in that board’s word, in the right position.
• Yellow: the letter is in that board’s word, in a different position.
• Grey: the letter is not in that board’s word.

Repeated letters are scored by count. If a board’s word has one E and your guess has two, only one of your E tiles can turn green or yellow. The other turns grey, even though the word does contain an E. Greens are marked first, then yellows are given out from the letters still unaccounted for.

How the unlimited game picks words

Each round of Quordle Unlimited draws four words at random from the 1,994-word answer pool. There is no daily cap and no account. Start a new round whenever you like.

How the daily puzzle works

Today’s Quordle gives every player the same four words. It resets at midnight UK time (GMT in winter, BST in summer), so it changes at the same moment everywhere. If you play from outside the UK, the change happens at a different local time.

The words for each date are set in advance, not picked when you load the page. A date’s puzzle does not change after it has run.

Past daily puzzles are listed in the answers archive. If an archive entry ever differs from the puzzle you played, please report it.

How the solver works

The Quordle solver filters rather than guesses. You enter the letters you have confirmed and the ones you have ruled out, and it returns every word in our lists that still fits. It applies the repeated-letter rule above.

What you get is a shortlist, not an answer. When several candidates remain, the best next guess is usually the word that splits them into the most different outcomes, not the one most likely to be right.

Limits of the solver

• It only knows what you tell it. One wrongly marked tile gives a confident but wrong shortlist.
• Each board must be entered separately, because each board has its own clues.
• It is bounded by our word lists. If a word is not in them, filtering will not find it.

How the starting-word numbers are calculated

The starting word tool computes every figure it shows from our own 1,994-word answer pool, not from a generic English frequency table. The letter figures are the share of answer-pool words that contain each letter.

S ranks unusually high because our answer pool includes plurals, which the original Wordle answer list leaves out. That is a fact about our list, not general word-game advice.

For any opener you type, the tool reports vowels tested, answers covered, blank rate, an information score, average candidates left and four-board reach. The information score is entropy in plain terms: a word scores well when its possible colour patterns are spread across many different outcomes, and poorly when most answers would give nearly the same pattern.

Limits of these numbers

• Every figure depends on our answer pool. Change the list and the rankings change.
• There is no single best opener. A word chosen for letter coverage is not the word an information model picks, and neither is always the easiest to reason from mid-game.
• None of these numbers guarantees a win.

How the starting words guide differs

The best starting words guide is editorial. Its picks come from reasoning about letter frequency, letter position and avoiding repeated letters, not from a ranked scoring run. For computed numbers on a specific word, use the tool.

One figure in that guide comes from outside this site: a simulation study by Raw Data Studies that reports a 53.5% solve-in-seven rate for CHIMP, ROBED, SLANT. That is their work, not ours, and the Sources page explains what we could and could not check about it.

What none of this guarantees

Our tools narrow the possibilities. They do not win games. Four candidates with two guesses left is still a gamble, and a well-chosen opener can still meet an awkward word.