An antonym matching game tests whether you can pick a direct lexical opposite rather than a vaguely related word, and the Antonym Match game does this across five fixed rounds for exactly 1,000 points. Each round shows one prompt word with its part of speech and a short sense note, then offers four candidate buttons; only one of them expresses the direct opposite for the displayed meaning in the same grammatical category. A correct pick adds 200 points, so the maximum run lands on 1,000. Every winning pair is verified against Princeton WordNet 3.1 using its explicit exclamation-mark antonym pointer between the relevant adjective or verb synsets, and is then cross-checked on the corresponding Cambridge Thesaurus page. The game is recreational vocabulary practice, not an adaptive test or reading-level tool, and it makes no claim about intelligence or language certification. The five rounds, the four candidates per round, and the candidate order are fixed and reproducible, so position alone does not reveal the answer and your score path is the same on every visit.
Why an antonym matching game is harder than it sounds
English antonymy is precise. Two words can appear on opposite sides of a debate, an argument, or a single sentence without being lexical antonyms, and a movement word is not automatically the opposite of a temperature word. A related probability term is not necessarily the direct opposite of capability, and a movement term is not the same as the opposite of a heat property just because both belong to a loose "change" cluster in casual conversation. Distractors in the game are built to expose that gap.
Words are also polysemous, which means the direct opposite shifts with the sense in use. A common verb can describe believing a statement, receiving an object, admitting a person, or agreeing to an invitation, and the right opposite depends on which of those meanings is on screen. An adjective can refer to physical temperature, emotional tone, current popularity, or electrical state. The Antonym Match interface names the intended sense before you select, so the relationship you are asked to identify is bounded and inspectable rather than left to guessing from context. That sense-bounding is what separates a direct lexical opposite from a general contrast, and it is the rule the distractor candidates deliberately test.
How Antonym Match verifies each direct opposite
The five rounds are not crowd-rated, AI-inferred, or curated from a similarity score. Each winning pair must satisfy two independent checks. The first is a literal exclamation-mark antonym pointer in Princeton WordNet 3.1, sitting between the recorded source and target offsets inside data.adj or data.verb for the displayed grammatical category. The second is a presence check on the corresponding Cambridge Thesaurus page, which returns HTTP 200 and contains the matching opposite in its downloaded body.
| Round | Prompt | Part of speech | Primary lexical check | Cross-check |
|---|---|---|---|---|
| 1 | hot | adjective | WordNet 3.1 data.adj antonym pointer | Cambridge Thesaurus page for hot |
| 2 | accept | verb | WordNet 3.1 data.verb antonym pointer | Cambridge Thesaurus page for accept |
| 3 | increase | verb | WordNet 3.1 data.verb antonym pointer | Cambridge Thesaurus page for increase |
| 4 | legal | adjective | WordNet 3.1 data.adj antonym pointer | Cambridge Thesaurus page for legal |
| 5 | possible | adjective | WordNet 3.1 data.adj antonym pointer | Cambridge Thesaurus entry referenced in fixture |
A private nine-case external fixture stores five direct relationships and four same-part-of-speech or related-sense boundaries that do not carry a qualifying pointer. Those non-antonym cases act as a guard against treating AI-similarity judgments as proof, and they confirm that round correctness is judged by literal lexical evidence rather than plausibility alone. Focused tests keep the external evidence independent from product evaluation, so the source-checked facts stand on their own.
Distractor candidates share plausible surface properties, and several sit in the same part of speech as the prompt, but none of them carries the required direct-opposite relationship in the disclosed sense. The answer is not exposed before selection: all four candidate buttons share the same markup, and no aria label, data attribute, CSS class name, hidden text, or status message marks the winning option. The highlighted style indicates only your keyboard cursor. Candidate order is fixed for reproducible testing but varies by round, and the correct position cycles across all four slots rather than repeating one answer location, so position alone does not reveal the right pick.
How to play the five rounds
- Read the prompt word, its part of speech, and the sense note. The sense note is what binds the round to a single lexical relationship; a candidate that is the correct opposite for a different sense is still wrong here.
- Inspect all four candidates. They are presented as full word strings; one carries the direct antonym pointer for the shown grammatical category and sense, and the other three are bounded non-antonyms that look related but lack the pointer for the disclosed meaning.
- Tap a candidate, or use the arrow keys to highlight a choice and press Enter to submit. Touch and pointer input submits immediately on tap; keyboard input moves a visible cursor through the four buttons and confirms with Enter.
- Earn 200 points for each correct match. Score updates only on a correct submission, and wrong picks do not subtract points or change the round.
- Clear all five rounds for the maximum 1,000 points. The run ends in either completion or a two-distinct-error dead-lock, both of which are terminal states where further selection or keyboard movement cannot alter the score, level, cursor, or history.
Controls: touch, mouse, and keyboard
Touch and pointer interaction is direct: tap a candidate to submit it. Each candidate is a native button at least 44 pixels tall, so it satisfies a comfortable touch target, and the text is allowed to wrap inside the button without breaking the layout. The mobile layout stays in a single bounded column before expanding to two columns on wider screens, and minimum-width plus overflow protection prevent any long candidate from creating horizontal page overflow on narrow viewports.
| Input | Action |
|---|---|
| Tap or click a candidate | Submit that candidate immediately |
| Arrow Left or Arrow Up | Move highlight to previous choice (wraps) |
| Arrow Right or Arrow Down | Move highlight to next choice (wraps) |
| Enter | Submit the highlighted choice |
| Escape, Escape (within window) | Boss key cover; press twice more to return |
| Restart | Reset to round 1, first highlighted candidate, zero score, and empty mistake list |
Keyboard interaction covers the entire puzzle. Left or Up moves to the previous choice, and Right or Down moves to the next; movement wraps around the four candidates so pressing Down on the fourth highlights the first, and pressing Up on the first highlights the fourth. Enter submits the highlighted choice. When a native button already has focus, its native Enter click is allowed to handle the action so the global shortcut cannot submit the same choice twice.
Scoring, mistakes, restart, and boss key
Each correct match awards 200 points, so five rounds total exactly 1,000. Wrong choices carry deterministic signatures made from the round and the candidate index, which means that repeating the same wrong selection does not count twice. A first distinct mistake leaves the round open and reports that the choice is not the direct opposite for the stated sense, while a second different mistake dead-locks the run until you Restart. Invalid negative, out-of-range, or fractional candidate indices are atomic no-ops, and unknown cursor directions do nothing, so stray input cannot corrupt the score or the cursor.
Completion and dead-lock are both terminal: once you reach one of those states, selection and keyboard movement cannot alter the score, level, cursor, or history. Restart restores round one, the first highlighted candidate, zero score, and an empty mistake list, so each run is reproducible from the same starting state. The shared GameShell stores the best completed score locally when browser storage is available, and it supplies a double-Escape boss key: press Escape twice within the timing window to replace the puzzle with a spreadsheet-like cover, and twice more to return. Boss mode does not alter the lexical fixture, the score, or the active candidate, so it is purely a visual disguise.
Everything runs locally in the browser. The game sends no selections to a server, requests no account, uses no paid service, and adds no package dependency, which means the same five fixed rounds are available offline once the page has loaded. Because the rounds, candidate orders, and score path are fixed and reproducible, the game does not adapt to your answers and it makes no adaptive vocabulary assessment claim. Completing the run proves only that these five source-checked direct-opposite relationships were identified, which is exactly what an antonym matching game is meant to test.
If you're weighing options, Battleships Puzzle Tips and Tricks: 6 Moves to 940 Points covers this in detail.