Minesweeper Strategy

Getting better at Minesweeper is not about thinking harder. It's about recognising a handful of shapes so fast that you stop thinking about them at all. Strong players aren't out-reasoning you on every cell — they're pattern-matching most of the board instantly and saving their actual attention for the few spots that need it.

Every pattern below is stated as a board fragment with the deduction worked out. If you already know the rules and just want the shapes, skip to 1-2-1 — that one alone will change your win rate.

Notation

 1 2 3    revealed cells with adjacent mine counts
         .        revealed cell, zero adjacent mines
         ?        covered, unknown
         F        your flag
         ▓        wall (board edge)
         *        deduced mine
         ✓        deduced safe

The 1-1 pattern

Two adjacent 1s along an edge, with covered cells above them. Most guides say this pattern "finds a mine." It doesn't. It finds a safe cell, which is more useful anyway.

   a   b   c
           ?   ?   ?
           1   1   ?
           ▓   ▓   ▓

The left 1 touches only a and b — the wall closes off its left side. So exactly one mine sits in {a, b}.

The right 1 touches a, b, and c. It also needs exactly one mine.

Since {a, b} already accounts for the right 1's full quota, c cannot be a mine.

   a   b   c
           ?   ?   ✓
           1   1   ?
           ▓   ▓   ▓

a and b remain genuinely ambiguous — one of them is a mine and nothing here tells you which. Don't guess between them. Open c, which is free information, and let the new number resolve the rest.

The load-bearing condition is that the left 1 is closed off — by a wall, or by already-revealed cells. An open-ended 1-1 with covered cells on both sides tells you nothing. This is the single most common misapplication of the pattern.

The same logic runs vertically and in any rotation. Learn it as "the constrained 1 eats the shared cells, so the extra cell on the open end is safe."

The 1-2-1 pattern

The best pattern in the game. Three consecutive numbers along an edge, covered cells above:

   a   b   c
           ?   ?   ?
           1   2   1
           ▓   ▓   ▓

Work it from the 2. It touches a, b, c and needs exactly two mines among them.

The left 1 touches a and b, needs exactly one. The right 1 touches b and c, needs exactly one.

Suppose b were a mine. Then the left 1 is satisfied, so a is safe; the right 1 is satisfied, so c is safe. That gives the 2 only one adjacent mine. Contradiction.

So b is safe, and the 2's two mines must be a and c:

   a   b   c
           *   ✓   *
           1   2   1
           ▓   ▓   ▓

Both mines located, one safe cell, zero guessing. I verified this by enumerating every arrangement consistent with those three clues — there is exactly one.

Flag a and c, open b. Or better, flag both and then chord the 2, which opens b in the same motion.

The 1-2-2-1 pattern

Four in a row. Same setup:

   a   b   c   d
           ?   ?   ?   ?
           1   2   2   1
           ▓   ▓   ▓   ▓

Here the mines are the middle two, and the outer cells are safe — the mirror image of 1-2-1, which is exactly why people mix them up.

The left 1 touches a, b: one mine there. The right 1 touches c, d: one mine there. That's two mines total across the four cells, which the 2s confirm.

Now the left 2 touches a, b, c and needs two. If a were a mine, the left 1 is satisfied so b is safe, forcing the left 2 to find its second mine at c. But then the right 1 (touching c, d) is satisfied by c, so d is safe — and the right 2, touching b, c, d, sees only c. It needs two. Contradiction.

So a is safe, b is a mine, and by symmetry d is safe and c is a mine:

   a   b   c   d
           ✓   *   *   ✓
           1   2   2   1
           ▓   ▓   ▓   ▓

Also verified by exhaustive enumeration: exactly one consistent arrangement.

Quick recall for both: count the 1s at the ends. 1-2-1 → mines on the outside. 1-2-2-1 → mines on the inside.

Chording is where the speed comes from

Chording opens every unflagged neighbour of a number in one action, once that number's adjacent flag count matches its value. Click both mouse buttons together on the number (middle-click and double-click also work on this site's board).

Compare on this fragment, where the two mines are already flagged:

   F   1   ?
           2   3   ?
           F   2   ?

Without chording: open each covered neighbour individually, one click at a time. With chording: one action on the 3, and every unflagged neighbour opens at once.

Across a full expert board that's the difference between roughly 250 clicks and roughly 100. Nobody clears expert in under 100 seconds clicking cell by cell — the input alone won't fit in the time.

The tradeoff is that chording executes your flags without checking them. The engine verifies the flag *count* matches the number; it does not verify the flags are on actual mines. Chord with a misplaced flag and you open a live mine instantly. Chording is a commitment to your own reasoning, which is why it's fast and why it punishes sloppy flagging. The engine internals page has the exact code path.

Opening move: corner or centre?

Open in a corner or near an edge, and here's the reasoning, which cuts against most advice.

Cell neighbour counts differ by position: a corner cell has 3 neighbours, an edge cell 5, an interior cell 8. On this site, the first click is guaranteed to have zero adjacent mines and always triggers a cascade — so the question isn't safety, it's what kind of position you inherit.

A centre opening clears a larger blob, which feels productive. But it hands you a long, irregular frontier surrounded by covered cells on all sides, and edge-dependent patterns like 1-2-1 need a wall to resolve cleanly.

A corner opening clears less, but the resulting frontier is braced against two walls. Walls are constraints, and constraints are what make deduction possible. That's where 1-1, 1-2-1 and 1-2-2-1 actually fire.

Practical routine: open a corner, work the walls, and only push into open space when the edges are exhausted.

When you have to guess

Sometimes logic runs out. That's the game's design, not your failure — some boards are provably unsolvable, explained on the engine page. What you control is guessing *well*.

First, check whether the global mine count breaks the tie. This is the most underused information in Minesweeper. Consider an endgame with three covered cells: an isolated corner cell z, and a 50/50 pair x/y where one is definitely a mine.

   x   y   ...   z
           ?   ?         ?
           2   1         .

If the counter says 1 mine remaining: the pair already accounts for it, so z is safe. Open z for free.

If the counter says 2 mines remaining: one is in the pair, so the other must be z. Flag z. The 50/50 stays a 50/50, but you gained a certainty for nothing.

Always read the counter before guessing. It routinely converts an apparent coin flip into a free move.

Second, if it's still ambiguous, compare probabilities. Estimate per-cell mine probability from the constraint that binds it. A cell bounded by a 1 touching 4 covered cells sits at roughly 25%. A cell in a 50/50 pair is 50%. Prefer the lowest.

Also weigh cells with no adjacent numbers at all — the untouched interior. Their probability is the global rate: remaining mines divided by remaining covered cells. Late in an expert board with 8 mines left across 40 covered cells, that's 20%, better than any frontier 50/50. Counterintuitively, the safest click is often far away from everything you know.

Third, guess early rather than late. If you can see a 50/50 forming, resolve it while the board still has slack. A coin flip on move 30 costs you one board; the same flip on move 200 costs a nearly finished expert run.

Reasonable target times

Rough bands for players using flags and chording. Wide, because hardware and mouse habits move these a lot:

Board dimensions and mine counts are on the rules page.

Two things about times. Expert win rates are low even for strong players — unsolvable boards mean a meaningful share of games are lost to forced guesses regardless of skill. Judge yourself on decision quality, not on streaks. And speed follows recognition, not hurry. Trying to click faster produces misclicks. The clock drops when you stop *computing* 1-2-1 and start *seeing* it.

A practice order that works

1. Play beginner until 1-1 and 1-2-1 are automatic. Don't time yourself.
2. Move to intermediate. Add chording. Expect it to feel slower before it feels faster.
3. Force yourself to flag only what chording will use. Over-flagging is the most common habit that caps intermediate players — the reasoning behind that is in the beginner walkthrough.
4. Move to expert once intermediate wins are routine. Accept the losses to forced guesses.

The board at the top of the home page runs all three difficulties. Start on beginner, hunt specifically for a 1-2-1, and chord it when you find one.