What a number tells you
A revealed number in Minesweeper is a precise statement: exactly that many of the eight surrounding squares contain mines. A 2 with exactly two unrevealed neighbours means both are mines. A 1 with exactly one unrevealed neighbour means that square is a mine. These cases need no probability at all, and a good share of any board falls to this kind of direct deduction.
Probability becomes useful when a number has more unrevealed neighbours than mines left to place. A 1 with three unrevealed neighbours says one of the three is a mine, but not which. Each clue is therefore a constraint on a group of squares, and the skill is in combining constraints.
The first estimate and why it is rough
The simplest estimate divides the mines still needed by the unrevealed squares around the number. A 1 with three unrevealed neighbours gives each of them a one-in-three chance, about 33 percent. That is a reasonable first guess when the number stands alone.
It stops being reliable when neighbouring numbers share those squares. A square touched by several clues has to satisfy all of them, so its true chance can be well above or below the simple figure. The estimate is a starting point, not an answer.
Overlapping clues: a small example
Take three unrevealed squares in a row, called x, y and z. One clue says x and y together hold exactly one mine. A second clue says y and z together hold exactly one mine. Which arrangements satisfy both?
- y is a mine, and x and z are safe. This uses one mine.
- y is safe, and both x and z are mines. This uses two mines.
Those are the only two possibilities. If they were equally likely, y would be a mine 50 percent of the time. But they are not equally likely, because the board has a fixed total number of mines. When the overall density of mines is low, an arrangement that uses fewer mines tends to be more likely than one that uses more, because the other mines then have to be placed elsewhere in the many remaining unconstrained squares.
A rough way to see the size of the effect: if the density of mines among the unconstrained squares is about 20 percent, each extra mine in the arrangement makes it roughly one quarter as likely, since 0.2 divided by 0.8 is 0.25. The one-mine arrangement then has a weight of about 1 and the two-mine arrangement about 0.25, so y is a mine in roughly 1 / 1.25 = 80 percent of cases. This is an approximation; an exact answer depends on the whole board. The direction is the useful lesson: with sparse mines, the arrangement that explains the numbers with fewer mines is usually favoured. This is the same reasoning as counting outcomes, with each arrangement being one outcome and the mine total deciding how it is weighted.
Board density sets the baseline
The standard difficulty levels have fixed mine counts, which gives a baseline chance for a square about which you know nothing.
- Beginner: 10 mines on 81 squares (9 by 9), about 12 percent.
- Intermediate: 40 mines on 256 squares (16 by 16), about 16 percent.
- Expert: 99 mines on 480 squares (30 by 16), about 21 percent.
Some versions or custom boards use other sizes, so check yours. As you reveal squares and mark mines, the density of what remains changes, and late in a game it can differ a lot from the starting figure.
Safe squares are worth more than they look
A guess has two possible costs: the chance of hitting a mine, and what you gain if you survive. Revealing a square with low mine probability that sits next to nothing useful gains little, but revealing one that is likely to open up a region or resolve a boundary can be worth a slightly higher risk. This is a small instance of comparing expected value across options instead of simply picking the lowest-risk square.
When forced to guess, there are two common choices:
- A boundary square with a lower chance than the baseline. If the best boundary square is, say, 15 percent against a baseline of 21 percent, it is the safer reveal and it also gives new information next to existing clues.
- A square away from any clue. This has the baseline chance, and the number it reveals may or may not connect to the rest. Corners are slightly useful because they have fewer neighbours, so a revealed zero is more likely to cascade.
The right choice depends on the board. Neither is guaranteed to be safe, and a guess at any positive probability can lose.
Common reading mistakes
- Treating squares as independent events. Two squares next to the same clue are linked: if one is a mine, the other is less likely to be. Estimating each separately from its own clue can miss that.
- Trusting a streak. Several safe guesses in a row do not make the next one safer, a trap that resembles the gamblers fallacy. Each guess faces the same underlying chance.
- Marking too early. A flag placed on a hunch can wrongly satisfy a number in your mind and push later deductions in the wrong direction. Flag only what you have proven.
- Ignoring the mine counter. The remaining mine count matters most near the end, when it can settle a position that the local numbers cannot.
Some positions have no safe move
Certain patterns cannot be resolved by logic alone. Two squares that are each a mine or not, with exactly one mine between them and no other clues touching them, form a true coin flip. Experienced players still hit these occasionally. The aim is not to avoid every guess but to guess rarely, choose the lowest-risk square each time, and accept that some losses come from chance rather than from error. A habit of honest risk assessment, estimating a number rather than going on a feeling, tends to help more than any single trick.