No-Guess Minesweeper
Classic Minesweeper will eventually deal you a position where two cells are equally likely to be the mine and nothing on the board can tell them apart. Losing there is a coin flip, not a mistake. These boards are built so that never happens.
What “no guess” actually means
It does not mean the board is easy, and it does not mean you cannot lose. It means that at every point in the game there is at least one cell you can prove is safe using only what the board is showing you. If you lose, you missed something — the deduction was there.
That is a stronger guarantee than it sounds. Most Minesweeper implementations place mines at random and hand you whatever comes out, and at expert density a meaningful share of those layouts contain a position that no amount of reasoning can resolve.
How the generator works
Nothing is generated until your first click, which is what makes the opening move safe: the cell you clicked and the eight around it are excluded from mine placement, so the first click always opens a region rather than a single number.
Then it is generate-and-test. A candidate layout is drawn at random, and a solver plays it the way a careful person would — only ever taking moves it can prove, never guessing. If the solver clears the board, you get that layout. If it gets stuck, the candidate is thrown away and another is drawn. Up to 300 candidates are tried before the generator gives up.
The solver is the honest part of this. It is not checking that the board looks reasonable; it is checking that a complete chain of deductions exists from the opening cascade to the last safe cell.
Why the hardest setting stops where it does
Solvable layouts get rare fast as mine density climbs, and the rate was measured rather than guessed at. On a 20x16 board, the share of random candidates the solver can clear:
- 18.8% density — 17% of candidates are solvable
- 20.6% density — 7% of candidates are solvable
- 22% density — 3% of candidates are solvable
- 24% density — 0.3% of candidates are solvable
At 300 attempts, 20.6% never once fell through to an unchecked board across 200 games on every board size. 22% still did. So 20.6% is where the hardest difficulty sits — not because it felt right, but because it is the densest board the guarantee survives.
That figure is also exactly classic Minesweeper's Expert density, 99 mines on a 30x16 grid. The hardest setting here is as dense as Expert, and solvable all the way through.
The four densities
Difficulty is a mine density rather than a fixed count, applied to whichever board size you pick — so a small hard board and a large easy one are both available, and the mine count is always derived from the two.
- Easy — 12.3%
- Medium — 15.6%
- Hard — 18.8%
- Extreme — 20.6%
The honest caveat
If all 300 candidates fail, the generator returns the first one it drew rather than failing outright — an unchecked board, indistinguishable from a real one while you are playing it. That is why the density ceiling exists and why raising it is guarded by a test rather than left to judgement: the failure mode here is silent, so the only safe place to sit is well inside the measured margin.
Custom boards can be built up to 32 by 16 with any mine count under half the grid. Push the density past the presets and you are outside the measured range, where the guarantee is best-effort rather than reliable.
Where it matters most
The daily challenge — everyone plays the identical board and is ranked by time, so a forced guess would decide the leaderboard at random. Same for a 1v1 race: two players on the same layout should be separated by who reads it faster, not by who called the coin flip. The rules and chording are worth a read if you are new.