The four-phase mental model
Every minesweeper game decomposes into four phases. Strong players learn to switch modes consciously; weaker players spend the whole board in phase one.
- Opening: the first click + the flood-fill it produces.
- Boundary scan: reading the numbers around the revealed region.
- Pattern deduction: applying named patterns (1-1, 1-2, subset) to find safe cells.
- End-game counting / forced guess: using the mine counter to pin down regions and, when forced, picking the lowest-probability guess.
Phase 1 — the opening click
On every modern minesweeper client the first click is guaranteed safe. The board is generated after your first tap so that the click and its immediate neighbours can't be mines. This means your opening click has a single job: maximize the size of the flood-fill region.
Where you click matters because the flood-fill expands through zero-cells. A click in the middle of the board has more candidate zeros to chain through; a click in a corner has only three neighbours and the chain dies sooner.
- Centre click: highest expected flood-fill area.
- Edge click: ~60% of centre's expected area.
- Corner click: the smallest expected flood-fill — typically only useful in puzzles where the mine density is so high the centre is too risky.
On a standard expert board (30×16, 99 mines, 20.6% density), a centre click produces a usable opening about 75% of the time. Speedrunners restart the remaining 25% — clicking restart and re-clicking centre is faster than fighting through a bad opening.
Phase 2 — reading the boundary
After the flood-fill, you have a region of revealed cells bordered by numbered cells. Those numbers are your information. The whole game lives at this boundary.
Two universal checks, run on every number:
- Mines fully flagged? If a number already has its mine count flagged among its neighbours, every other unrevealed neighbour is safe. Reveal them.
- Unrevealed count equals number? If a number's value equals the count of unrevealed neighbours, every one of those neighbours is a mine. Flag them.
These two checks alone clear the majority of a typical board. Sweep the entire boundary once before any clever deduction. You'll be surprised how many cells fall to the basics.
Phase 3 — pattern deduction
When the basic checks stop producing new information, you switch to named patterns. Three patterns cover ~80% of intermediate deductions.
The 1-1 pattern
Two adjacent 1s along a boundary, both touching the same pair of unrevealed cells, with one 1 touching an extra unrevealed cell. The extra cell is guaranteed safe — the proof and edge cases are in the dedicated 1-1 pattern guide.
The 1-2 pattern
A 1 next to a 2 along a boundary, both with shared neighbours plus one unique cell for the 2. The 2's unique cell is a mine; the 1's unique cell is safe. Walkthrough in the 1-2 pattern guide.
Subset deduction
When one number's unrevealed neighbours are a strict subset of another's, you can subtract the values. The difference is the number of mines in the extra cells of the larger set. If that difference equals the count of the extras, they're all mines. If it's zero, they're all safe.
This generalizes the 1-1 and 1-2 patterns and handles the majority of "non-obvious" board states.
Phase 4 — the chord move (skip flags entirely on small boards)
Once a number has the right count of flags around it, clicking the number itself reveals all its remaining unrevealed neighbours. This is the chord. On every board with even a handful of flag-able cells, the chord saves seconds.
Two non-obvious notes:
- Speedrunners on small boards often skip flagging entirely. Without flags, you reveal every safe cell directly. The flag step costs a click per mine, which on small boards is most of your time budget.
- On expert (30×16, 99 mines) the chord is mandatory. Manually revealing each cell costs minutes. The full chord drill is in the chord move guide.
Phase 5 — end-game counting and forced guesses
When the board is mostly revealed and a few patches of unrevealed cells remain, the mine counter at the top becomes your most valuable signal. The remaining mine total constrains the remaining regions globally.
If 3 cells are unrevealed in one region and the mine counter says 1 mine left in the whole board, exactly 1 of those 3 cells is a mine. The other 2 are safe — let the local numbers tell you which one.
When a true 50/50 (or 1-in-3, or 1-in-4) appears, you've reached a forced guess. Two rules:
- Pick the region with the lowest probability of a mine. A 1-in-3 region beats a 1-in-2 region.
- When two regions tie on probability, pick the one whose result will tell you the most about its neighbours. Information has value too. Full treatment in the late-game probability guide.
Practice plan
Time on the board beats theory. A useful weekly plan that produces measurable improvement in 6-8 weeks:
- Daily: one daily challenge run. Same board for everyone — compare your time to the leaderboard.
- 3× per week: 10 attempts on intermediate (16×16) focused on pattern speed, not perfect play.
- 1× per week: a single expert attempt taken slowly, narrating the deduction. Find the patterns you miss.
- Weekly: one 1v1 multiplayer match against a real opponent. Race conditions surface skills the solo board hides.
Track your median win-time, not your fastest. Improvement shows up in the median first.
What separates the players who improve from the ones who don't
Three habits, in order of impact:
- Restart fast on bad openings. If your first flood-fill is tiny, mash restart. Don't fight through a board you didn't earn.
- Learn one pattern per week, deeply. Don't read all the patterns at once. Pick the 1-1 one week, the 1-2 the next, subset the third. Spaced learning sticks.
- Play the daily even on bad days. Streaks lock in habit. A 30-second beginner run is enough to keep the muscle alive.
FAQ
How do you win at minesweeper consistently?
Master the boundary scan + the three named patterns (1-1, 1-2, subset). Most boards fall to those alone. Use the chord to clear neighbours fast. Reserve guessing for forced positions only.
What's the best opening click in minesweeper?
Centre — it maximizes the expected flood-fill area. Restart any opening with a small flood; cycling restarts is faster than playing through a bad opening.
Is minesweeper a game of skill or luck?
Both. Most boards are fully solvable by deduction, but ~15-30% contain a forced 50/50 or 1-in-3 guess. World-record runs require favourable openings and top-tier skill.