Connections review: the puzzle that turned “I guessed too early” into a national pastime

The first thing to know about Connections is that it does not look hard. Sixteen words in a 4×4 grid. Four hidden categories of four words each. Find the categories. You get four mistakes. That’s the whole game. The puzzle takes maybe 90 seconds to read the rules and roughly 8 minutes, on average, to actually solve.

The second thing to know is that the New York Times released Connections in June 2023 expecting it to be a quiet supplement to Wordle. Eighteen months later it is, by daily-active-user count, the second-most-played puzzle on the NYT Games platform — ahead of Spelling Bee, ahead of the full Crossword, second only to Wordle itself. The NYT product team, in a moment of refreshing honesty I appreciated, has admitted to multiple outlets that they did not see this coming.

I have been playing Connections every day since launch. I have notes.

Why it works

Connections is, in essence, a vocabulary problem disguised as a categorization problem. The brilliance is in the disguise. Every grid contains:

• One yellow category — always the easiest, always the trap. “Fruits.” “Colors.” “Things on a desk.”
• One green category — a step harder. Often a category that requires noticing something specific, like “words that follow ‘sun’” or “things on a baseball diamond.”
• One blue category — noticeably tricky. Usually a category where the four words have a less obvious shared property.
• One purple category — the hardest, almost always the one that uses the words you thought belonged in another category. Wordplay. Hidden meanings. Things you didn’t see at first.

The structural trick is that the puzzle’s difficulty comes not from the difficulty of any individual category but from the overlap between them. The yellow category, almost every day, contains at least one word that also fits the purple category. If you guess yellow first using that overlap word, you lose a guess and you don’t notice purple until it’s too late.

That single design decision — categories that share at least one plausible member — is the difference between a clever puzzle and a great one. Most days, the puzzle is unwinnable if you guess the obvious yellow category first.

What’s great

1. The difficulty curve is genuinely fair. After 18 months of play, my solve rate is about 78%. That’s a satisfying middle ground. Easier than the NYT Saturday crossword. Harder than the Spelling Bee. The exact gradient that keeps a daily puzzle compelling without being demoralizing.

2. The shareable result format is genuinely elegant. Connections inherited Wordle’s spoiler-free emoji-grid sharing format and improved on it. A finished Connections grid in iMessage — the four colored bands — reads instantly to anyone who plays. “Mixed yellow before purple” communicates the same thing as a long verbal post-mortem.

3. The themes feel written. A surprising fraction of Connections grids are themed. “All the categories somehow involve ducks.” “All the categories are something you can have on toast.” The puzzle has clearly been edited by someone with a sense of play, not just a vocabulary database.

4. The streak isn’t weaponized. Connections shows you a streak count, but — unlike Wordle, where breaking a long streak feels like grief — a Connections streak quietly accepts that you’ll lose sometimes. The puzzle is hard. The streak is a record, not a guilt trip.

What’s frustrating

1. Some days the categories are too cute. Once a month or so, the puzzle leans heavily on a category that requires you to see, say, that four words can all precede “BOX” — but only if you also know the British slang term “BOX” for something else. Rare, but it produces puzzles where the loss feels unfair rather than instructive.

2. The deshuffle button is too easy to misuse. Connections lets you randomly shuffle the grid to look at it from a fresh angle. It’s a useful feature. It is also, for some players (this player), a tic that produces 20 shuffles per puzzle and meaningfully impedes solving. A small UX fix — maybe a 3-second cooldown between shuffles — would force people to actually think between presses.

3. There’s no “why” on the result page. When you finish (or lose), the app shows you the four categories. It does not, for some inexplicable reason, show you the logic behind the harder ones. If purple was “words that mean small in different languages,” the puzzle should be willing to tell you that. Right now it just shows the words.

The four-pile method

If there’s one piece of strategy that has improved my Connections game more than any other, it’s this: do not submit anything until you can sort all sixteen words into exactly four piles.

Most losses come from guessing a yellow category early because three words obviously fit. The fourth word “kind of” fits, and you submit, and you’re wrong because the fourth word actually belonged to purple, and now you’ve burned a guess and — worse — you don’t know which of your other guesses are also wrong.

The four-pile method forces you to commit to a hypothesis about all sixteen words before you submit any of them. Our full strategy guide here walks through it in detail.

The verdict, after 18 months

Connections is the most consistently entertaining daily puzzle the New York Times has shipped since the Mini. It is harder than Wordle, faster than the full Crossword, and structurally more interesting than either. It is also, in a way that’s rare in modern phone games, fundamentally fair: when you lose, you can almost always see exactly where you went wrong.

If you have an NYT Games subscription and somehow haven’t tried it yet, it should be your tomorrow morning. If you don’t have a subscription, Connections alone might be worth the $50/year. We’ve ranked it #5 in our 2026 best-apps list and stand behind that ranking. The four puzzles ahead of it are all individually excellent. None of them are this clever.