Is Minesweeper Luck Based? I Solved 150,000 Boards
Is Minesweeper luck based? Less than people think on small boards, more than they think on big ones. I ran 150,000 random boards through a solver that never guesses: it cleared 91% of Beginner boards, 71% of Intermediate and only 15% of Expert without a single guess. Every other board reached a point where the numbers left two or more squares that logic couldn’t tell apart.
I make a browser version of the game, Minesweeper, and wanted a real number instead of the ones floating around online. Figures quoted for Expert run from about 30% to 85% solvable, usually without saying how anyone got them. So I measured it with the game’s own code.
The numbers
Each board is the standard size: Beginner 9 × 9 with 10 mines, Intermediate 16 × 16 with 40, Expert 30 × 16 with 99. The first click is in the middle. “Won” means the solver cleared the whole board without ever guessing.
| Board | Won without guessing | Boards played |
|---|---|---|
| Beginner | 90.7% | 20,000 |
| Intermediate | 70.7% | 20,000 |
| Expert | 15.1% | 10,000 |
On Expert, 135 boards (1.4%) had tangles too big for the solver to finish in its time limit, and they count as not won here. If every one of them turned out solvable, Expert would still only reach 16.5%.
The other number worth knowing is how far the solver got on the boards it couldn’t finish. On Expert it had cleared 74% of the safe squares, on average, before it was stuck. The guess usually comes late, after most of the work is done, which is part of why losing on Expert hurts.
How I measured it
The solver plays like a careful human with perfect patience. After every move it looks at all the open numbers and works out every arrangement of mines that fits them, including the total mine count. A square that is safe in every possible arrangement gets opened. A square that is a mine in every arrangement gets flagged. When no square is certain, the board is “stuck” and counts as needing a guess.
That is the strongest a no-guess player can be: if the solver can’t prove a square is safe, no amount of reasoning can. It is also why some positions take it so long. Deciding whether a Minesweeper position is even possible is NP-complete in general, which Richard Kaye proved in 2000, so a big enough tangle of numbers can take a computer a very long time to untangle.
The boards are the game’s own, generated from seeds 1 to 20,000 (10,000 on Expert) with the same mine-laying code the game uses. I also checked the solver against the real mine positions on hundreds of random boards: it never opened a mine and never flagged a safe square.
What this doesn’t measure is how often a good player wins. A forced guess is often a coin flip or better, so someone who guesses well wins more often than these numbers. They measure how many boards can be finished on logic alone.
Why the first click matters so much
Not every Minesweeper deals the first click the same way. Some versions only promise that the first square you click isn’t a mine. Others, mine included, also keep its eight neighbours clear, so the first click always opens an area to work from.
That one rule changes everything:
| Board | First square safe, area opens | Only the first square safe |
|---|---|---|
| Beginner | 90.7% | 29.8% |
| Intermediate | 70.7% | 17.3% |
| Expert | 15.1% | 2.2% |
When only the first square is safe, it usually shows a number and nothing else. With nothing to reason from, the second click is a guess. On boards that got stuck under that rule, the solver had cleared only 5% to 9% of the safe squares, so the luck arrives on the very next move.
Where you click first matters too. Starting in a corner instead of the middle, with the area rule, drops the no-guess rate to 81.7% on Beginner, 57.7% on Intermediate and 10.4% on Expert. A corner opening touches fewer squares, so it gives the numbers less to work with.
Where the luck comes from
The classic forced guess sits against a wall. Two covered squares, one mine between them, and every number that touches them touches both. Here F is a mine already flagged and _ an open, empty square:
? ? F
1 3 F
_ 1 1
The 1 says there is exactly one mine in the two squares above it, and the 3 agrees whichever one it is. No number touches one of them without the other, so nothing on the board can say which. Players call it a 50/50. Bigger versions exist, where three or four squares share mines in a way no number separates, and they get more common as the board gets denser. Expert packs 99 mines into 480 squares, about one in five, against one in eight on Beginner, which is why it falls off so sharply.
So is Minesweeper luck or skill?
Mostly skill, with a ceiling set by luck:
- On Beginner, skill decides nine games in ten. A player who never makes a mistake almost always finishes.
- On Intermediate, about seven in ten. The rest need at least one guess somewhere.
- On Expert, luck decides most games eventually. Only about one board in seven can be finished on logic alone, and the forced guess tends to come late. Skill decides how far you get and whether you survive the parts that can be reasoned out. Luck often decides the last few squares.
Over many games the skill shows: better players get further, make fewer mistakes and win far more often. On any single Expert game, though, losing to a 50/50 says nothing about how well you played.
How to need less luck
- Start in the middle. From the numbers above, a centre start beats a corner start on every board size.
- Play a version where the first click opens an area. It is the biggest single difference in the whole experiment, roughly tripling the no-guess rate on Beginner.
- Try a no-guess mode if you hate coin flips. Some versions only deal boards that a solver can finish, so a careful player never has to guess.
Play it
My Minesweeper is free in the browser on a phone or a computer, with no download. It deals boards the way this post measured them: random mines, and a first click that is always safe and always opens an area. It has the three standard boards, flags by right-click, long-press or a flag-mode switch, clears around a number when its flags are placed, works from the keyboard, and keeps your best time for each board on your device. Expert stands upright on a phone so it fits.
The honest limits, as of October 2026:
- Boards are random, so on Expert you will sometimes reach a 50/50. There’s no no-guess mode yet.
- Best times are kept on your device only. There’s no account and no online leaderboard.