The shape
A 1 sits next to a 2 along a revealed boundary. Both touch three unrevealed cells: two are shared between them, plus one extra cell that only the 2 sees. Like this (? unrevealed, # the wall):
? ? ? ?
1 2 # #
# # # #Two things are simultaneously true: the cell only the 2 sees is a mine, and the cell only the 1 sees is safe. One read, one click for a flag, one click to reveal — three pieces of information at once.
Why both deductions hold
The 1 has three unrevealed neighbours and exactly one mine among them. The 2 has four unrevealed neighbours and exactly two mines among them.
The two 1-mine and 2-mine constraints overlap on the shared cells. If the 1's mine were in the leftmost cell (its exclusive one), the two shared cells would have to hold both of the 2's mines — but that's two mines in cells the 1 also touches, breaking the 1's "exactly one mine" rule.
So the 1's mine has to sit in one of the shared cells. That means the shared cells contain exactly one mine. The 2 needs two mines total, but only one of them is in the shared region, so the second mine has to be in the 2's exclusive cell. Done: 2's exclusive is a mine, 1's exclusive is safe.
The mirrored shape
The shape works in either direction. A 2 sitting next to a 1 on the right (instead of left) gives the same result mirrored:
? ? ? ?
# # 2 1
# # # #Now the leftmost cell (only the 2 sees it) is a mine, and the rightmost cell (only the 1 sees it) is safe. Identical logic.
The 1-2-1 sandwich
A common extension. When a 2 sits between two 1s along the boundary:
? ? ? ? ?
1 2 1 # #
# # # # #Apply 1-2 from the left: rightmost cell of the 2's row is a mine, leftmost is safe. Apply 1-2 from the right: same result mirrored. The whole region collapses: both outer cells of the 2's row are mines, and the middle cell is safe. The two outer 1s have their mines pinned, and you can chord directly after flagging.
Variations along a wall
1-2 on a flat wall (canonical)
Two 1 and 2 along the bottom edge with three unrevealed cells above them and one to the side. This is the textbook case described above.
1-2 along a vertical edge
Same logic rotated. Two 1 and 2 stacked along a side wall, with three unrevealed cells to one side. Just read the shape as if the board were rotated 90°.
1-2 with one number in a corner
When the 1 is in a corner, its neighbour count drops. If the corner 1 sees only two unrevealed cells (both shared with the 2), the deduction shrinks but still works: the 2's exclusive cell is a mine, the 1 has no exclusive cell to declare safe.
1-2 in the open
Sometimes the pattern appears mid-board rather than against a wall. The proof still holds, but you need to scan more carefully — the "shared neighbours" can be in any direction, not just above. Trace each number's three or four unrevealed neighbours before applying the rule.
A worked example
Mid-game on intermediate. The boundary along your revealed region reads:
· · · · · ·
? ? ? ? ? ?
1 2 1 1 2 1
# # # # # # Left side: 1 2 — the cell above the 2's exclusive (the third cell from the left in the unrevealed row) is a mine, and the cell above the 1's exclusive (the first cell) is safe. Right side: 2 1 — the cell above the 2's exclusive (the fifth cell) is a mine, and the cell above the 1's exclusive (the sixth cell) is safe. Middle: 1 1 is the 1-1 pattern, but no exclusive cells exist between them because both share the same neighbours — that one's neutral.
Net result from a five-second scan: two mines flagged, two safe cells clicked. Four cells of progress with zero risk. That's the kind of deduction that turns a 3-minute board into a 90-second board.
Stacking 1-1 and 1-2 in the same scan
Once both patterns are memorised, run them together. Scan the boundary left to right: every pair of adjacent numbers gets a one-second check.
- 1-1 → one safe cell on the outside.
- 1-2 → one mine in the 2's exclusive, one safe in the 1's exclusive.
- 2-1 → mirrored 1-2.
- 2-2 → less common but solvable: typically both shared cells are mines if the exclusives don't help (see the no-guessing strategy section).
- 1-3 / 3-1 → reduces to subset deduction — the 3 has all the 1's neighbours plus extras, so the extras must contain 3 − 1 = 2 mines.
With four patterns in muscle memory, 80% of mid-game decisions become reflex.
Common misreadings
Confusing 1-2 with 1-1-2
Three numbers in a row read differently. A 1-1-2 stretch has two adjacencies to evaluate. Don't apply the pattern to the whole run at once — slice it into pairs and check each pair independently.
Ignoring already-flagged mines
If one of the 2's shared neighbours is already flagged, the 2 only needs one more mine — and that effectively turns it into a 1. The pattern then becomes 1-1, not 1-2. Always re-count flags before applying the rule.
Applying it mid-region instead of on the boundary
The pattern relies on three or four total unrevealed neighbours. If the 1 or 2 has more unrevealed cells (because the region is wider than expected), the deduction doesn't fire. The shape only works on a clean boundary.
Drill it on the daily
The daily challenge is built for pattern drilling. Same seed for the whole world, so once you've played, you can re-load tomorrow's puzzle slowly the next day to see which 1-2 spots you missed. Players in the top decile on the leaderboard usually have both 1-1 and 1-2 reads running on autopilot — that's the unlock.
If you want a structured progression: start with beginner 9×9 and play five boards where you deliberately verbalise every 1-2 you see. Move to intermediate once you can run the loop without slowing down. Expert demands the pattern be invisible — you click, flag, click in a rhythm without conscious thought.