How Does CloseWord Know Two Words Are Close?

CloseWord compares meanings, not letters: it checks your guess against a hidden word and returns where that guess ranks among the game’s vocabulary for semantic closeness. CloseWord is a one-player semantic word game about finding one hidden word through ranked feedback.

The setup is deliberately bare. There are no cards, no board squares, no opponent, no hand to manage, and no points total to protect. You begin with an empty list of guesses and a hidden answer. Each legal guess is a word the game recognises. After you submit it, CloseWord gives that word a rank against the whole vocabulary.

A turn is therefore small but useful: type a word, read its rank, decide whether that meaning points nearer to the target or away from it. The win condition is exact, not interpretive. You win by guessing the hidden word. If you arrived here searching how does contexto work, the broad idea will feel familiar: both are forms of word similarity game where the feedback is semantic rather than spelling-based.

What is CloseWord actually comparing when it says one word is “close” to another?

CloseWord is comparing semantic closeness: how near two words are in meaning, not how many letters they share, how they sound, or whether they have the same length. A guess such as doctor can be closer to hospital than hostel, even though hostel looks more similar on the page.

The visible rule is the rank. After a legal guess, CloseWord places that word among the game’s vocabulary according to closeness to the hidden answer. A very near guess gets a better rank than a vague or unrelated one. The hidden word itself is the target; when you guess it, the search is over.

The underlying method belongs to the same family as other semantic comparison tools: words are treated by their patterns of meaning and use, then compared by distance. The exact model matters less in play than the shape of the output. You do not get a definition, a synonym list, or a hot-and-cold bar. You get an ordered clue.

This is why CloseWord behaves differently from a crossword, an anagram, or a spelling puzzle. The word bank may pull you towards money, rivers, finance, buildings, or verbs depending on the target and the vocabulary around it. Meaning is messy. The ranking makes that mess playable by turning it into one number per guess.

That number is feedback, not a score in the usual card-game sense. In Klondike Solitaire (Turn 1), moves expose cards and build foundations. In CloseWord, guesses expose a semantic direction. You are not trying to earn points; you are trying to reduce uncertainty until only the answer fits.

What do you see at the start of a game, and what counts as a legal guess?

At the start, CloseWord gives you a hidden word and an input for guesses. The answer is not shown. There is no grid to fill, no deck to shuffle, and no visible list of possible solutions. The game is for one player, and the state of the game is built from the guesses you have already made.

A legal guess is a word that CloseWord recognises in its vocabulary. If the word is not in that vocabulary, it cannot be ranked usefully against the answer. The exact boundary of a vocabulary always matters in a semantic word game: names, slang, rare inflections, and compounds may not behave like everyday dictionary words. The honest practical rule is simple. If the game accepts it, treat the returned rank as information. If it rejects it, choose a clearer word.

CloseWord does not ask you to match word length, place letters, or avoid repeating a colour. A short word can be useful. A long word can be useless. The only question is whether the meaning of the guess sits near the hidden word.

Your first screen is plain by design. You are not reading a rulebook on the page; you are creating your own clue list. The first accepted guess gives you a point of reference. The second gives you a comparison. After that, the game starts to feel less like guessing in the dark and more like triangulation.

What happens on a turn, from first guess to feedback?

1. You choose a starting word. Pick something broad enough to test a region of meaning: food, work, animal, music, city. The first guess is not meant to be brilliant. It is meant to produce the first ranked clue.
2. You submit the guess. CloseWord checks whether the word is recognised in its vocabulary. A recognised word can be compared with the hidden answer. An unrecognised word is not useful feedback, because it has no place in the game’s ranking.
3. The game returns a rank. That rank tells you where your guess sits when compared with the whole vocabulary for closeness to the hidden word. A better rank means closer meaning. A worse rank means the guess is farther away semantically.
4. You compare it with earlier guesses. One number is only a hint. Two numbers form a direction. If teacher ranks much better than office, the answer may sit nearer education than general work. If school then improves again, your next guesses should test classroom, pupils, lessons, exams, and related words.
5. You keep narrowing. Each turn adds one more measured point. The game does not end because you have made a certain number of moves or reached a score target. It ends when your submitted guess is the hidden word.

The turn is short, but the thinking between turns is the game. Bad guesses still help when they rule out a meaning cluster.

How do you work out the hidden word without seeing it?

Treat the ranks as a map made from your own guesses. You are not being told the answer’s letters, so letter tactics do little. You are testing fields of meaning, then splitting the promising field into smaller parts.

• Start broad. Use words that cover large areas: nature, sport, money, family, law, weather, medicine, transport. Broad guesses quickly tell you which parts of meaning are cold.
• Follow improvements. If one word ranks better than another, ask what changed. Did you move from object to action, from place to person, from emotion to behaviour?
• Test neighbours, not copies. If doctor is close, try nurse, patient, clinic, medicine, and surgery. A synonym may be too narrow too early; a neighbour can reveal the true angle.
• Use opposites carefully. Opposites often live near each other in meaning because they appear in the same conversations. Hot and cold are far in sense but close as concepts.
• Watch for category traps. A good rank for dog might mean animal, pet, loyalty, barking, breed, or even police work. Do not assume the first category you see is the target.

The strongest play is comparison. Write mentally in pairs: this was closer than that; this branch got worse; this related word suddenly jumped. CloseWord rewards players who can abandon a neat theory after one better clue.

A useful guess is not always close. Sometimes it is the word that proves your current idea is wrong.

Why does this word similarity game stay interesting after the first win?

CloseWord stays interesting because the puzzle changes with the hidden word, but the tool you use stays the same. Every game asks the same action from you: guess a word and read the rank. The thinking behind that action keeps changing.

Some answers sit in tidy clusters. If the target is near a profession, an object, or a familiar place, the path can feel clean: job, workplace, tools, person, exact word. Other answers are slippery. Abstract words, words with several senses, and words tied to culture or emotion can produce rankings that feel surprising until you see the answer.

That surprise is part of the value. A semantic word game shows you that vocabulary is not stored in neat schoolroom boxes. Words connect through use, association, topic, contrast, and habit. A guess can be close because it is a synonym, but it can also be close because people discuss the two ideas together.

It also avoids the stale part of many word puzzles. Once you know a trick in an anagram, you can apply it almost mechanically. CloseWord has tactics, but no guaranteed route. A strong opening word may be useless on the next answer. A poor-looking detour may find the winning cluster.

The game is small enough to play in minutes, yet uncertain enough to make you argue with your own guesses. Each rank is precise; the reason behind it still needs human judgement.

Frequently asked questions

Final thoughts

The best way to play CloseWord is to stop treating the rank as a verdict on whether you are clever. It is a measurement. Measurements are useful even when they are disappointing.

If your first five guesses are poor, you have still drawn the edges of the problem. You know several meanings that probably are not central. That is progress, though it feels less satisfying than a sudden near miss. Keep one eye on categories and the other on relationships: object and action, person and place, cause and effect, literal and figurative meaning.

Do not be too loyal to a beautiful theory. Semantic feedback can tempt you into a story: this must be about hospitals, this must be about schools, this must be about sport. Then one guess breaks the story. Believe the new rank. The game is fairer when you let the evidence embarrass you quickly.

There is also pleasure in noticing why the final answer made sense. The hidden word often explains the strange guesses that came before it. That little click is the point. You began with a blank field, threw meanings into it one at a time, and ended with a word that had been shaping the whole space from the start.