Wordle, Nerdle, Connections, and where Chain fits

There are more daily puzzles than anyone has time for, and they are not interchangeable. Wordle asks you to guess a word and tells you how close you were after every try. Nerdle and Mathler do the same thing with arithmetic. Connections asks you to sort sixteen things into four groups and tells you almost nothing until you commit. Chain gives you a start number, a target, and a hand of moves, and says nothing at all until your last move lands. Below is what each of them actually costs and pays, including the places where Chain is the worse choice.

What makes a daily format work

Three things, and they are the same three every time.

One puzzle for everyone. The point of a daily is that the person next to you on the bus solved the same thing. That is why every format here fixes the puzzle to the calendar date rather than generating a fresh random one per player. It also means you cannot grind: when today is done, it is done, and that limit is a feature.

A low, honest time cost. A daily has to fit in a coffee. Wordle's six guesses and Nerdle's six guesses are hard caps. Countdown-style arithmetic puzzles are usually capped by a clock. Chain caps by the number of moves — three, four or five — but it does not cap the number of tries, so it is the format here most able to eat twenty minutes if you let it. Chain's answer to that is the difficulty curve: Monday and Tuesday are three-move puzzles and take a minute or two.

A result you can paste. Not a screenshot — text. Wordle's emoji grid works because it shows how the solve went without giving away the word, and it survives being pasted into any chat. Everything since has copied the idea, us included.

Wordle: the shape everything else borrowed

Josh Wardle made it public in October 2021, the emoji-grid share spread it in late December, and The New York Times bought it on 31 January 2022. Six guesses, one five-letter word, the same word for the whole world that day.

What Wordle does better than anything on this page is the on-ramp. There is nothing to learn. You know the alphabet, you type a word, the colours tell you what to do next. No rules page, no worked example, no arithmetic. Every guess is also useful — even a bad guess eliminates letters — so you are never stuck staring at a board with no move that teaches you anything. That is a real design achievement and it is the reason Wordle, not its imitators, is the one your family plays.

Nerdle and Mathler: equations with Wordle's feedback

Nerdle launched in January 2022 from Richard Mann and his family, built with Marcus Tettmar. You guess a hidden eight-character calculation in six tries; green means right character in the right slot, purple means right character in the wrong slot, grey means not in the equation at all. Mathler runs the same loop from the other end: it tells you the answer up front and hides the calculation that produces it, with boards of different lengths for different appetites.

These are the closest relatives Chain has, and they are honestly better than Chain at one specific thing: every guess pays you back. You learn that there is a 7 somewhere, or that the division sign is in slot four, and you carry that into the next attempt. Chain has nothing like it. A failed Chain gives you exactly one bit of information — this set of tiles does not reach the target — and you have to derive the rest yourself.

The trade is that Nerdle and Mathler are partly positional deduction rather than arithmetic. A good chunk of the skill is reasoning about which characters sit in which slots, the same skill Wordle trains. If what you want is to do arithmetic, some of the work is elsewhere.

Connections: the format with the best aha

The New York Times released Connections in beta on 12 June 2023, edited by Wyna Liu. Sixteen words, four groups of four, four mistakes allowed.

Connections is the hardest of these to be confident in, on purpose. The overlaps are deliberate: several words plausibly belong to two categories, and the puzzle is built so that the obvious group is often the wrong one. When it lands, the moment of seeing the category is the biggest payoff on this list — better than finishing a Wordle, which usually just feels like arriving.

It is also the least predictable in time. A Connections can be over in ninety seconds or it can sit there while you burn all four mistakes. And it depends on shared cultural reference, which makes it uneven if English is not your first language or you did not grow up where the editor did. That is a real limitation, not a criticism: no arithmetic puzzle has that problem.

Where Chain sits

Chain gives you a start number between 10 and 99, a target between 100 and 999, and a hand of five tiles (six on hard days). Each tile is one move: +n, −n, ×n, ÷n — exact divisions only — or swap digits, which reverses the digits of your running value. You reach the target in exactly the number of moves shown, using each tile at most once, and exactly one subset of tiles gets there. The rules page has the details and four worked puzzles.

Two things follow from that design.

First, the puzzle is the set, not the order. Once you have the right tiles, most orderings work, so there is no fiddly execution phase after the insight. That makes Chain closer to Connections than to Wordle: you are looking for a grouping, and either you see it or you do not.

Second, the decoys are constructed, not random. Every spare tile is legal from a position you can actually reach, and every spare tile is part of at least one full-length chain that lands within ten of the target. That is why a near miss in Chain is informative — 170 instead of 160 usually means one wrong tile, not a wrong order, and the strategy guide is mostly about reading those misses.

Difficulty follows the weekday for everyone: three moves Monday and Tuesday, four Wednesday, Thursday and Sunday, five Friday and Saturday.

What the other formats do better

Being straight about this:

• Wordle's on-ramp. Chain needs you to read a rules page. That is a real cost and we have not found a way around it.
• Nerdle's per-guess feedback. Chain tells you nothing until the chain ends. If you like a puzzle that converges, Chain will feel like a wall on a Friday.
• Connections' cultural reach. Chain will never produce the specific pleasure of recognising a category. It is arithmetic, and arithmetic does not have jokes in it.
• Reach. All three have enormous player bases, which means solvers, discussion threads, and someone to compare with. Chain does not.

The flip side is narrow and we think it is worth something: Chain is the same puzzle whatever language you read, it has no vocabulary to learn, and the work it asks for is arithmetic you can do in your head while walking.

Playing more than one

They stack well because they use different parts of your attention. A common set is a word puzzle, a grouping puzzle, and one arithmetic puzzle, which comes to roughly ten minutes. If you already do Nerdle or Mathler, Chain is not a duplicate — Nerdle asks which characters, Chain asks which operations — but it is the closest overlap on this page, so try a Monday puzzle before committing to a habit.

Chain's share text is three plain lines with the date, your tries and your time, and a link. No emoji grid, because there is no per-guess pattern to draw. Past puzzles stay playable in the archive, which is the one place we deliberately break the "one puzzle a day" rule: if you find Chain late, there is no reason to make you wait a week to get good at it.