Guide
No-Guess Minesweeper: What It Means and How It Works
Classic Minesweeper has a dirty secret: some boards can’t be finished by logic alone. At some point you have to flip a coin. No-guess Minesweeper removes that. This page explains what a forced 50/50 is, how a no-guess board is made, and what we do and don’t claim in FairMines.
What is a forced 50/50?
A forced 50/50 is a spot where two cells share exactly one mine, and nothing on the board can tell you which cell it is. The numbers around them are the same either way. You must pick one, and half the time you lose.
The simplest case: two unrevealed cells in a corner pocket, with one mine between them and every surrounding number satisfied by either choice.
Other guess situations exist too, such as a ring of cells with several consistent layouts where no cell is safe in all of them. Players often call all of them “50/50s”, but the pure two-cell case is the most common source of the name.
Classic Minesweeper places mines at random, so these traps arise by chance. On large, dense boards, a random layout often needs at least one guess.
What is a “no-guess” board?
A board where, starting from the cells already revealed, you can always find a cell that is certainly safe or certainly a mine by reasoning from the numbers, until the whole board is cleared. A loss on a no-guess board means a misclick or a missed deduction, not bad luck.
How no-guess generation works
There are two broad ways. One is to adjust mines after the fact (moving mines when you click). The other, which FairMines uses, is generate, then solve-check:
- Generate a random board with the right size and mine count, and a pre-opened safe starting area.
- Solve it with a program that only uses logic. It plays the board as a person could: it reveals only cells proven safe by the numbers, never guesses.
- Accept or reject. If the solver clears every safe cell, the board is kept. If it gets stuck, the board is thrown away and a new one is generated.
- The solver also records the hardest kind of deduction the board needed, which sets the difficulty (next section).
Generation happens in the background and is fast, so you don’t wait. As a safety net, the app also carries a bundled set of pre-checked boards, used if generation ever takes too long. The Daily board comes from a fixed, pre-checked list so everyone in the world gets the same one.
What difficulty means in FairMines
Most Minesweeper apps set difficulty by size and mine count alone. FairMines also sets it by what kind of reasoning a board needs:
| Difficulty | Board | Mines | Reasoning the board may require |
|---|---|---|---|
| Easy | 9 × 9 | 10 | Level 1: single-number rules (“this 1 touches one hidden cell, so it’s a mine”). |
| Medium | 10 × 14 | 22 | Up to level 2: comparing two numbers (if one number’s cells are a subset of another’s, the difference tells you something). Needs at least one such step. |
| Hard | 10 × 18 | 34 | Up to level 3: working out all possible mine layouts along the edge of the opened area. Needs at least one such step. |
| Expert | 12 × 22 | 54 | Level 3, plus using the total mine count. |
“Needs at least one such step” means a Medium board is not secretly an Easy one: some deduction beyond the basics is required.
Every board, even Expert, can be cleared by logic. Hard and Expert are harder because the logic is harder, not because the odds are worse.
How we checked it: 11,930 boards
The game’s own tests check the generator. Because a solver that grades its own work can share its own bugs, our QA wrote a separate solver from scratch, with no shared code, and used it to check the boards.
That solver is stricter than the in-app one: at each step it searches for every mine layout consistent with what is revealed and counts a cell as safe only if no consistent layout makes it a mine. A board passes only if this clears every safe cell with no guess. As a self-test, it correctly gets stuck on a real 50/50.
It checked 11,930 boards:
- 8,000 freshly generated boards, 2,000 each at Easy, Medium, Hard and Expert
- 3,200 boards from the bundled backup set (200 per difficulty, each in four orientations)
- 730 Daily boards, covering two years
Result: all 11,930 were clearable by logic alone, with zero failures. It also confirmed that every board’s opening was valid (a safe start with no mines around it) and that every board sat in its claimed difficulty tier. We re-ran a 7,930-board subset on the final build; again zero failures.
What we don’t claim
Honest limits matter more than a bold promise:
- Not a mathematical proof for all possible boards. It is a solver check on every board the app can serve from its generator, plus 11,930 independently verified samples. Boards the app serves are accepted by the in-app solver before you see them, but we have not exhaustively checked every possible board that could ever be generated.
- Not “easy.” No-guess does not mean simple. Expert boards can need long chains of reasoning and a count of the remaining mines.
- Not “no luck at all” in the opening. The app pre-opens a safe starting area so your first move isn’t a gamble, but where it opens is not your choice.
- Not a human difficulty rating. Levels describe the kind of logic needed. They are not a measured time-to-solve. Two Hard boards can feel very different.
- Not unique to us. Other apps offer a no-guess mode, often as an option. Ours is on for every board, including the Daily.
- Not a promise that you’ll never lose. You can still misclick, flag wrong, or miss a deduction. That’s the game.