Word Wheel

• Every word must contain the centre letter. A word made from the outer eight alone does not count, however long it is.
• Every word must be three letters or longer. Two-letter words do not score.
• Only the nine letters on the wheel may be used — no others, at all.
• Each letter may be used once per word, for each time it appears on the wheel. One T on the board means one T in your word; a wheel showing two E's will take two.
• Words score by length, from 1 point for a three up to 25 for the word that spends all nine letters.
• Proper nouns, hyphens and apostrophes are out, because they are not in the dictionary this game judges against.

The one-use rule is what makes a wheel a wheel rather than an anagram with a spare. Nine tiles spent once each is also a hard ceiling on word length: nothing longer than nine letters can ever be legal here, which is why the nine-letter word is the end of the search rather than a step in it.

Because the nine letters are a rack of tiles, not a palette. It is the rule that separates a word wheel from a letter-hive puzzle, where the same seven letters may be reused as often as a word needs, and it is a much bigger difference than it sounds: measured over the first 200 standard wheels, lifting the rule would add a median of 49 more findable familiar words to a wheel that holds 52 of them. It would very nearly double the puzzle.

What it takes away is mostly the easy half. Doubled letters are where English hides its cheap words — the -LL, -SS, -EE and -TT endings that a reuse rule hands you for free — so a wheel asks you to spell things out of what is actually in front of you. The compensation is that you can count: if there is one N on the board, every answer has at most one N in it, and a word you are unsure of can be checked by eye rather than by guessing.

The same rule also decides what a duplicate tile means. A wheel read off a word such as agreement puts two E's on the board, and then a word may spend two E's and no more. Try the reuse rule for yourself on Letter Hive, which is the same idea with the opposite answer.

The curve is steep on purpose. Three- and four-letter words are 61% of a standard wheel's word list and only 29% of its score, so they are worth taking when you see one and never worth grinding. The nine-letter word alone is worth 25 points — about 15% of the 167 points a median standard wheel holds, and as much as twelve four-letter words.

A hint costs 2 points. The first one uncovers the opening letter of the nine-letter word, the second uncovers two, and so on — but it stops at four of the nine and will not go further, and a hint that repeats one you have already bought is free. The cap is not meanness. The nine-letter word spends every tile exactly once, so its last letter is simply whichever tile the others did not use: uncover eight and the ninth is arithmetic, not a guess. Stopping halfway leaves five tiles whose order you still have to find.

They are shares of how many familiar words the wheel actually holds, so the targets move with the puzzle rather than being the same round number every day.

Reaching Very good is what counts as a win in the statistics panel. The targets are measured in words rather than points on purpose, and it is the opposite of the choice Letter Hive makes: a wheel's scoring is steep enough that one nine-letter word is 15% of everything the board is worth, so a score-based Very good would be 58 points on a median wheel and that one word would be 25 of them — 43% of the target from a single find, with the other forty answers left to split the rest. Any real word you find counts towards a target, including one the familiar list never counted.

Not from a random draw. Nine letters pulled out of a bag usually make a dead wheel, and worse, they make a wheel with no nine-letter answer on it — which is the one thing this puzzle promises. So each wheel is read off a real nine-letter word, taken with its repeats. There are 2,367 familiar nine-letter words that qualify, making 2,351 distinct sets of letters between them.

A source word must carry at least seven distinct letters. Nine-letter words with fewer are far poorer puzzles and the gap is measured, not assumed: across every source and every choice of centre, a five-distinct word yields a median of 18 findable familiar words where a nine-distinct word yields 72.

The wheel is then measured before it is dealt. If the pool of familiar words falls outside the size chosen for that board, the whole thing is thrown away and read off a different word, up to forty times — and on the rare wheel where forty draws all miss, the generator stops guessing and walks the source list until it finds one that fits. The size on the label is therefore a guarantee and not an average: every wheel across the first 600 puzzle numbers of each size, and across 20,000 of the random numbers a New wheel click produces, landed inside its window. 584 of the first 600 standard wheels were a different set of nine letters from each other.

Almost every wheel holds exactly one nine-letter answer: 7 of the first 600 standard wheels held two, and none held three. That is why the end-of-round panel can simply name it.

Yes, and more than most people expect. The centre is a filter applied to every single answer, so the same nine letters make a noticeably different puzzle depending on which one sits in the middle. Take the letters of creations — nine letters, all different — with only the centre changed:

Centring on S gives 192 words where centring on I gives 116, from an identical board — two-thirds more puzzle for changing nothing but the middle cell. That is why the generator measures the pool after choosing the centre rather than before, and it is why the first thing worth doing on a new wheel is to look at the centre letter and ask what normally sits beside it.

Two different lists do two different jobs, and confusing them is how a word game loses players. Your entries are judged against the full 172,823-word validation list, so an obscure word you happen to know still scores. The 25,322-word familiar list is what the wheel is built from and what the targets are measured against — it decides how much is on the table, not what is allowed.

The practical effect is that there is always more on a wheel than the count admits to. On a standard wheel the validation list accepts about 2.25 times as many words as the familiar pool, so a run of unusual finds pushes you towards a target without moving it.

Both lists come from ENABLE, which is public domain. Nothing here is drawn from any copyrighted or licensed tournament list, and the same two lists sit behind Word Hunt and Word Scramble, so a word that scores on one scores on the others.