There is a kind of puzzle game that is pleasant for an hour and never once makes you stop and think. And there is another kind where you stare at a board for four minutes and then say "oh".

The difference is usually not effort, taste, or how generous the designer was feeling. For a lot of genres it is a property of the rules, fixed before anyone starts designing levels, and no amount of tuning moves it.

We measure this for our own games before building them, because it decides whether we are making a calm game or a hard one — and it is much better to know that in advance than to discover it after shipping a difficulty ladder that does not mean anything.

Here is the test, and where our catalogue falls.

The test

Can a legal move make the board unsolvable?

That is it. If every move is safe, the puzzle cannot be about planning, because planning is only valuable when a choice can be wrong.

Some genres fail this test by construction. In the arrow-tap family — tap a piece whose exit path is clear — removing a piece only ever empties squares. So any board reachable from a solvable board is still solvable. Every legal tap is safe, always, and there is no lose state to design. That is not a level-design choice; it follows from the rule.

The second question, for genres that pass:

What fraction of legal moves are safe?

If it is 99 per cent, you have a game where no single move is a trap but a long run of careless ones will still kill you. We call that death by accumulation, and it feels quite different from a puzzle with real traps in it.

The calm end — and calm is not an insult

Arrow Escape is the clearest case. Roughly half of all pieces are legal on an opening board, and it cannot be tuned lower — we tried five different levers and every one either failed or made the game easier. The full account is here.

Prism Sort — ball sort. With two spare tubes, 96 to 99 per cent of legal moves keep the board solvable. So you will lose boards, but never to a single mistake: a player choosing at random dies about 64 per cent of the time over thirty moves, purely by accumulation. It has a genuine efficiency chase — a competent player lands within 1.08 to 1.19× the optimal move count — but it is not a thinking game.

Gridlock, Gem Match, Tray Match, Jigsaw and Mahjong Solitaire all sit here too. Mahjong is structurally the same family as the arrow games, which surprised us when we worked it out.

This is not a complaint. These are some of the most-played games in the world and they are played for exactly what they are: something to do with your hands that will not punish you for thinking about something else. The failure is not making a calm game. The failure is labelling a calm game "Expert" and letting someone believe they are being tested.

The hard end

Unjam — Rush Hour. The one member of the sliding family that breaks the safety property, because pieces slide within the board and only your car leaves. It is the only one of its relatives that people write solvers for, and generating good levels for it needs a full breadth-first search rather than the free-lunch construction the calm games allow.

Star Grid (Star Battle), Bridgework (Hashiwokakero), Daily Sums (Kakuro) and Shikaku — all deduction puzzles where a wrong placement genuinely poisons the board.

For these we rate difficulty by necessity: a board is Hard only if it is still unsolvable when the harder technique is switched off. Our first attempt rated by which techniques fired, which turned out to measure nothing at all — the write-up is here, and the short version is that one technique fired on nearly every board while almost no board actually needed it.

Nonogram, Minesweeper and Sudoku are the familiar members of this group. Chess Puzzles is the hardest thing on the site by a distance, because the positions are real.

The middle: hard to execute, not hard to plan

Worth its own category, because these get misfiled in both directions.

2048, Ninefold, Block Blitz and Tile Order are not deduction puzzles. There is a good policy and you mostly know what it is — keep your largest tile in a corner, keep the board tidy. They are hard because executing that policy consistently over a long game is hard, not because any single move is a puzzle.

That is a real kind of difficulty. It is just a different one, and it wants a different mood.

Why we publish this

The uncomfortable version of this article is that we have measured our own games and about half of them cannot be made hard. We publish it anyway, because the alternative is putting "Expert" on a level that isn't, and a player who feels tricked by a difficulty label does not come back.

It also makes the recommendation honest. If you want to be tested, we can tell you precisely which four games to open. If you want an hour of quiet, we can tell you that too — and we will not pretend the second list is a lesser one.

Want to think: Unjam · Star Grid · Bridgework · Chess Puzzles

Want quiet: Arrow Escape · Prism Sort · Jigsaw · Mahjong Solitaire

All free, all in the browser, no account: browse everything