The board, by the numbers
Expert is 30×16 with 99 mines. That's 480 cells, of which 381 are safe and 99 are bombs. Mine density: 20.6%. Compare to beginner's 12% and intermediate's 16% — expert is genuinely a different kind of board. You can't power through it; you have to read it.
Three numbers shape the strategy:
- 381 safe cells to reveal. At 6 clicks per second, that's a 64-second theoretical floor. Real players don't reveal each cell individually — they cascade through flood-fills and chords.
- 99 flags to (potentially) place. The "flag everything" approach is the slowest possible play. Top players place 40-60 flags per board, mostly for chord-prep.
- ~150 boundary cells at the half-cleared mark. This is the entire pattern-recognition surface area — every boundary cell is a 1-1, 1-2, or subset-deduction candidate.
Opening theory
The first click is always safe. Beyond that, where you click matters more than most beginners realise.
Corner-first (the conventional pick)
Click a corner cell. Corners have only three neighbours, so when the click reveals a 0, the flood-fill propagates only outward — never wrapping behind you. This produces a clean triangular region with a long, information-rich boundary.
Top speedrunners overwhelmingly favour top-left or top-right corners. The visual scan direction (left-to-right, top-down) aligns with the cleared region, so your eyes don't have to backtrack.
Center click (the high-variance pick)
Clicking the center maximises potential flood-fill area but produces a boundary that wraps in every direction. When it works, you might clear a third of the board on the opening click. When it doesn't, you get a tiny region surrounded by unclear boundaries.
Edge-of-corner (the hybrid)
Click one cell in from a corner. Slightly fewer revealed-area wins than a pure corner, but the boundary is denser — useful when you want pattern density over raw area.
Pick one opening, stick with it for a hundred boards. The muscle memory matters more than the specific choice.
The boundary, and why it's everything
Once you have a cleared region, every minesweeper decision happens on its boundary: the line of numbered cells touching the unrevealed area. Cells in the interior tell you nothing new; cells deep in the unrevealed area can't be deduced without boundary information.
Top players spend almost no time on cleared interior. Their eyes track the boundary like a coastline, scanning for the recognisable shapes:
- 1-1 → safe cell on the far side.
- 1-2 / 2-1 → mine in the 2's exclusive, safe in the 1's exclusive.
- 1-3 / 3-1 → subset deduction; the 3's extras hold two mines.
- 2-2-1 stretches → multi-step but solvable; usually pins one mine and two safes.
- Isolated 8 → all eight neighbours are mines; flag them and chord every adjacent number.
Pattern recognition is the single biggest gap between mid-tier and top-tier expert times. Click speed is a constant; what you click for is the variable.
Flag economy
Every flag costs about 0.2 seconds. Every chord saves 3-5 single clicks (also ~0.6-1.0 seconds). The math is favourable for flags only when you'll chord later. On expert, that condition is met for most 3+ numbers, common for 2s, and rare for 1s.
- Always flag mines around 3-8. The chord pays off massively.
- Flag mines around 2 only when adjacent numbers will also chord. Single-chord 2s are marginal.
- Skip flagging isolated 1-mine cells. Just avoid the cell. The flag is wasted motion.
- Don't flag in the late game. When the remaining mine count is small, you can usually count remaining mines directly without flagging.
Top runners often place 40-50 flags on an expert board, which sounds high until you realise it's roughly one flag per 8 cells revealed. Most of those flags enable chords; the rest are cognitive offload for late-game counting.
The 50% rule for late-game counting
Late expert games — say 30 cells unrevealed, 8 mines left — produce huge probability shifts. A boundary cell adjacent to a satisfied number sits at 0% mine probability. An unconstrained interior cell sits at 8/30 ≈ 27%. The math matters more than the patterns at this stage.
Rule of thumb: when remaining mines / remaining unrevealed cells drops below 30%, switch from pattern mode to counting mode. Track the mine counter at the top of the board, count what each constraint forces in its local region, and propagate outward.
For deeper coverage of probabilistic guessing, see the dedicated late-game probability article.
The mental clock
Expert speedruns aren't won by faster clicks — they're won by less hesitation. The biggest losses on a clean board come from 2-3 second pauses while you "think" about a deduction that should have been reflex.
Calibrate your mental clock by playing a five-board set with a deliberate one-second cap on every deduction. If you can't see the answer in a second, click the safest-looking cell and move on. You'll detonate more often in training — but you'll also start to notice which patterns you're slow on. That's the feedback you need.
After the calibration set, drop the cap and play normally. The muscle memory persists.
Click discipline
At sub-100 second pace, every click is in a flow state. The mouse hand or the thumb moves in straight lines, never backtracking, never hovering. Each click commits — no second thoughts.
- Don't hover. Hovering wastes 200-500ms per cell. If you can't decide, click the surer of two options.
- Don't go back. Moving the cursor left after moving right doubles your travel cost. Run rows or columns in single sweeps.
- Trust the pattern reads. If you read a 1-1, click the safe cell — don't re-verify.
- Accept some detonations. A 5% loss rate at fast pace usually beats a 1% loss rate at slow pace, depending on how you weight average time vs win rate.
The unsolvable 5-10%
Some expert boards force a guess. Around 5-10% of randomly generated layouts include at least one 50/50 that can't be resolved without information from elsewhere. There's no shame in losing those — top players average around 70% win rate on expert, and the missing 30% is mostly forced guesses plus the occasional rushed misclick.
Don't let those losses bleed into the next board. The next seed is independent; play it with the same confidence.
The progression to sub-100
- Sub-200. Hit it consistently before doing anything else. The path: clean pattern reads on 1-1 and 1-2, no flagging on isolated 1s, no hovering. Don't chase faster clicks; chase faster decisions.
- Sub-150. Now bring in chord cascades. Every satisfied 2+ should auto-chord. Pre-flag regions where you can see the cascade two moves ahead.
- Sub-120. Subset deduction unlocks the third pattern (1-3, 1-2-1 sandwiches). Late-game counting becomes the default mode below 25 unrevealed cells.
- Sub-100. No hesitation, no backtracking, deliberate opening, ride the cascade. At this level you're competitive with the upper rungs of the leaderboard.
- Sub-60. Mostly hardware and ergonomics. Mouse DPI, screen size, monitor refresh rate, keyboard placement. Beyond pure technique.
Track your progress on the leaderboard. The timestamps don't lie — twenty boards in a row at sub-150 means you've internalised the patterns. Now chase the next tier.