Minesweeper Patterns: Which Ones You Actually Need
Minesweeper patterns are mostly two numbers read together, and two of them do most of the work: the 1-1 and the 1-2. I played 50,000 boards with a solver that always uses the simplest reasoning available and logged what each move needed. Reading numbers in pairs like that, after subtracting flags, took the share of Intermediate boards it could clear without a guess from 25% to 59%. The famous 1-2-1 is just two 1-2s, and big multi-number chains almost never mattered. The mine counter mattered more.
I make a browser Minesweeper and had already measured how much of the game is luck. Pattern guides say the basics cover most of the game, but none I found says how much, or which pattern you’ll need most often. So I counted.
The patterns, quickly
In the diagrams, ? is a covered square, · is an open square whose number doesn’t matter here, | is the edge of the board, S is a square the pattern proves safe and M one it proves is a mine. The row below the numbers is already open.
1-1. Two 1s side by side along a straight edge, the first one against a wall. The first 1 touches only two covered squares, and the second 1 touches those same two. Whichever one holds the mine also satisfies the second 1, so its third square is safe.
| ? ? S
| 1 1 ·
1-2. A 1 next to a 2 along a straight edge. The 1 allows at most one mine in the two squares they share, so the 2 must have its other mine in the square only it touches.
? ? ? M
· 1 2 ·
1-2-1. Read the 1-2 from each end. The squares beside both 1s are mines, the one beside the 2 is safe, and so are the squares just past the ends.
S M S M S
· 1 2 1 ·
1-2-2-1. The same trick on four numbers. The mines sit beside the two 2s.
S S M M S S
· 1 2 2 1 ·
Reduction. Subtract the flags already touching a number before you read it. A 3 with two flags next to it is a 1. Authoritative Minesweeper’s patterns page explains all of these well, and names the 1-2 as the rule the 1-2-1 and 1-2-2-1 are built from.
What I measured
The boards come from my game’s own code: Beginner 9 × 9 with 10 mines, Intermediate 16 × 16 with 40, Expert 30 × 16 with 99, first click in the middle, seeds 1 to 20,000 (1 to 10,000 on Expert). That’s the same 50,000 boards as the luck post.
The player never guesses. Each turn it uses the cheapest reasoning that finds anything, and records which kind it was:
- One number. A number that already has all its mines flagged, so the rest of its squares are safe, or one with exactly as many covered squares as missing mines, so they’re all mines. Flags are subtracted first.
- Two numbers. A square that two overlapping numbers prove only when read together. Every pattern above is one of these.
- Three or more numbers. A square that only a whole group of connected numbers proves.
- The mine counter. A square that is only certain once you count how many mines are left in the whole board.
Then I played every board four times, letting the player go one level further each time, and counted the boards it finished without a guess.
Results
| Reasoning allowed | Beginner | Intermediate | Expert |
|---|---|---|---|
| One number at a time | 64.6% | 24.5% | 0.1% |
| Plus two numbers (the patterns) | 82.6% | 59.0% | 9.3% |
| Plus bigger groups of numbers | 82.9% | 59.7% | 10.1% |
| Plus the mine counter | 90.7% | 70.7% | 15.3% |
| Boards played | 20,000 | 20,000 | 10,000 |
The last row is everything logic can do. Beginner and Intermediate match the luck post exactly. Expert comes out at 15.3% rather than 15.1%, because this player clears the easy parts first and leaves the solver smaller tangles: only 9 boards were too big for it to finish this time, against 135 before, and they count as not cleared.
Three things stand out.
Single numbers are nearly every move. Of all the squares the player proved safe or a mine, 95% of Beginner’s needed only one number, 96% of Intermediate’s and 91% of Expert’s. Two-number patterns were 4% to 8% of moves.
But the patterns decide whether you finish. That small share of moves is the difference between clearing a quarter of Intermediate boards and clearing three fifths of them. On Expert, one number at a time cleared 13 boards in 10,000. Without the 1-1 and the 1-2 you are guessing on almost every Expert game.
Big chains almost never matter, the counter does. Reasoning across three or more numbers at once added under one point on every board. Watching the mine counter added 8 points on Beginner, 11 on Intermediate and 5 on Expert. On Expert, a third of the boards that could be cleared without a guess needed the counter at least once.
Which patterns actually came up
Every time the player had run out of one-number moves and two numbers together proved something, I recorded the shape. A 1-1 or 1-2 counts as the textbook shape when the two numbers sit side by side and every covered square they touch lies in one straight row (or column) on the same side.
| Shape | Beginner | Intermediate | Expert |
|---|---|---|---|
| 1-1 along a straight edge | 38% | 37% | 30% |
| 1-2 along a straight edge | 35% | 39% | 42% |
| … of which part of a 1-2-1 | 22% | 22% | 19% |
| … of which part of a 1-2-2-1 | 2% | 2% | 2% |
| Other 1-1 (round a corner, or one inside the other) | 21% | 18% | 18% |
| Other 1-2 | 5% | 6% | 7% |
| Anything with a 2-2, or a 3 or more | 0.5% | 1.2% | 2.7% |
| Two-number moves recorded | 16,725 | 65,300 | 130,656 |
The numbers here are what’s left after subtracting flags. So the 1-1 and the 1-2 between them are about three quarters of all pattern moves, and some kind of 1-1 or 1-2 is over 97%. The 1-2-1 is common: it was needed on 64% of Expert boards. The 1-2-2-1 is the one guides always list second, and it turned up about a tenth as often as the 1-2-1, on 14% of Expert boards and 3% of Intermediate ones.
Most patterns are hiding behind flags. In 47% of Beginner’s pattern moves, 68% of Intermediate’s and 75% of Expert’s, at least one of the two numbers already had a flag next to it, so what you see on screen isn’t a 1-1 or a 1-2 until you subtract. On Expert, 84% of the 1-1s looked like something else at first. Reduction isn’t an advanced extra. On the bigger boards it is how you find most of the patterns.
How I did it, and what it doesn’t show
The code is a test file kept next to the game’s own code, and the run took about 90 seconds on a laptop for all 200,000 plays. The player checks its own work: if it ever opened a mine, the run would have stopped with an error, and it never did. The two-number step works out every arrangement of mines the pair allows and keeps the squares that are the same in all of them, so it finds every pair pattern there is, including ones no guide has named. The shape labels are mine, a simple check of where the numbers and their covered squares sit, so a few odd cases will be filed under “other”.
It counts what a careful player needs, not what you’ll notice. A shape only counts when single numbers had nothing left, because that’s when you have to look for it. And it says nothing about speed. Good players see a 1-2-1 at a glance instead of reasoning it out, which is the real reason guides tell you to memorise them.
What to learn, in order
- Subtract flags before you read a number. On Expert most patterns only show up after it.
- Learn the 1-2. The square beside the 2 that the 1 doesn’t touch is a mine. The 1-2-1 and 1-2-2-1 come free with it.
- Learn the 1-1 against a wall, and the same thing round a corner. The square past the second 1 is safe.
- Watch the mine counter near the end. When the numbers account for every mine left, every covered square they don’t touch is safe. It mattered more than all the long chains together.
- Don’t drill rare shapes. Patterns with a 3 or higher in them were under 3% of pattern moves even on Expert.
Play it
My Minesweeper is free in the browser on a phone or a computer, with no download. It deals boards the way they were measured here: random mines, and a first click that’s always safe and always opens an area. It has the three standard boards and a mine counter that drops as you flag. Flag by right-click, long-press or the Flag mode switch, and tap a number whose flags are all placed to clear around it. Your best time for each board is kept on your device.
The honest limits, as of November 2026:
- There’s no hint button and no pattern trainer, so it won’t point out a 1-2-1 for you.
- Boards are random, so some still end in a forced guess. There’s no no-guess mode yet.
- Best times stay on your device. There’s no account and no online leaderboard.
Questions people ask
Is there a pattern to Minesweeper?
The mines are placed at random, so there is no pattern to where they are. The patterns people talk about are arrangements of numbers that always have the same answer, like the 1-2-1 along a wall. Almost all of them come down to two numbers next to each other, a 1-1 or a 1-2, read together.
What is the 1-2-1 pattern in Minesweeper?
Three numbers in a row reading 1, 2, 1 along a straight edge of covered squares. The squares beside the two 1s are mines and the square beside the 2 is safe. It is two 1-2 patterns, one read from each end.
What is the 1-2-2-1 pattern?
Four numbers in a row reading 1, 2, 2, 1 along a straight edge. The mines are beside the two 2s, and the squares beside the 1s are safe. Like the 1-2-1, it is two 1-2 patterns read from each end.
Which Minesweeper patterns should I memorise?
The 1-1 and the 1-2, and the habit of subtracting flags before you read a number. In 50,000 boards I played with a solver, those two shapes made up about three quarters of every move that needed two numbers at once, and on the bigger boards most of them only appeared after flags were subtracted.
What do you do when you're stuck in Minesweeper?
Subtract the flags around each number and look again for a 1-1 or a 1-2. Then check the mine counter, especially near the end. If none of that finds a safe square, the board may need a guess, which on Expert happens in most games.