Most strategy guides for Minesweeper are written by people who skipped the boring middle. They tell you to chord and to look for 1-2-1 and then they show off some endgame trick. That is fine, but it skips the actual journey from beginner to solid player. I want to write the guide I would have wanted when my Expert times were stuck in the three-digit range and I could not figure out why.
A note before we start. Every puzzle on MineRace is solvable through pure logic. There are no forced guesses. If you find yourself about to flip a coin on two squares, the answer is somewhere else on the board. Train yourself to look around before you commit. Guessing gets you killed and it is almost never necessary.
The mental model that fixes most beginners
A revealed number is a constraint. It says: out of these specific neighbors, exactly this many are bombs. That is it. The whole game is solving a system of these little constraints until every square is resolved. When you read a board, you are not looking at squares, you are looking at constraints overlapping each other.
Once you start thinking that way, the patterns below stop feeling like tricks and start feeling like consequences. A 1-2-1 works because the 2 in the middle has to be satisfied by exactly two of its three new neighbors, and the 1s on either side each cover one of the outer ones. The middle is forced safe. You can derive that yourself any time you need to.
Patterns to learn cold
These are the four you should be able to spot without thinking. Not read, not solve, just spot. They are the bread and butter of every Expert game.
1-1 along an edge
You have two 1s next to each other, both touching the same edge of unrevealed squares. There are three unrevealed squares in front of them: one touches only the left 1, one touches both, one touches only the right 1. The far ones are bombs. The middle one is safe.
Spot this constantly. Especially when one of the 1s is on a corner, this resolves a whole side of the board in two flag placements.
1-2-1 along an edge
Same setup but with a 2 in the middle. Three unrevealed squares in front. The outer two are bombs, the middle one is safe. The 2 forces it. This is the most common pattern on Expert and the one that pays the most.
1-2-2-1
Four numbers in a row, four unrevealed squares in front. The two outer hidden squares are safe, the two inner ones are bombs. Same family of reasoning as the 1-2-1, just one wider.
The completed number
A number that already has its full quota of flags around it. Every other neighbor is safe. Chord click it. This is the engine of speed play. Half of all your clicks on Expert should be chord clicks on completed numbers.
Subset reduction, the meta-pattern
All the named patterns above are special cases of a more general idea called subset reduction. It sounds fancy. It is not.
Take two numbers on the board, call them A and B. Look at the unrevealed squares that A is counting. Now look at the unrevealed squares that B is counting. If A's set is a subset of B's set, you can subtract: the squares that are in B but not in A must contain (B's remaining bomb count minus A's remaining bomb count) bombs.
If the answer is zero, those squares are all safe. If the answer equals the number of squares, they are all bombs. Either way you just deduced something new.
Real example. A 2 on the edge has three unrevealed neighbors, and a 1 just below it has two of those same neighbors plus one extra square. The 1's neighbors are a subset of the 2's neighbors. Subtracting: the one square that only the 2 sees must contain 2 - 1 = 1 bomb. So that square is a bomb. The 1 is now fully constrained and its other unique neighbor is safe.
Doing this in your head, on the fly, while a timer is running, is basically what high-level play looks like. The named patterns are the most common shapes of subset reduction. Once you internalize the underlying logic, you start spotting new ones on your own.
Endgame and the global mine count
Late in the game, you have a few hidden squares left and a number telling you how many bombs are still unplaced. This number is more information than people realize.
Say you have ten hidden squares left, three of them are bombs by the mine counter, and your local constraints say one specific patch of four squares has two bombs in it. That means the other six squares contain only one bomb, which often resolves things like classic 50/50 corners by counting.
On MineRace the mine counter is the red LED on the HUD. Watch it. When you have under twenty hidden squares left, the global count becomes a real solver tool, not just a curiosity.
Reading the opening
Every MineRace puzzle starts with an opening cascade. The first thing to do is not click. The first thing to do is read.
Scan the border of the opening for the pattern of numbers. You are looking for the easy reads first: any 1 with only one unrevealed neighbor is a free flag. Any 2 with two unrevealed neighbors that touch a 1 next door usually gives up a 1-2-1. Any 0 you can chord into. Knock those out before you start working harder.
On Expert this initial sweep can sometimes resolve a quarter of the board in ten seconds. You go from a tight opening to a wide open canvas. The slower path is to start in one corner and grind outward. The faster path is to harvest the easy reads first, no matter where they are.
The flag-or-not question
Every flag placement costs you a click. If the flag does not enable a chord, it might be wasted time. Strong players flag only when:
- The flag unlocks an immediate chord click on a neighbor.
- The flag prevents a misclick later when the board gets dense.
- They need to mark a bomb to keep track of their thinking.
Speedrunners often play with very few flags down. They reveal safe squares as fast as they can identify them and they keep the bomb positions in their head. This is called no-flag style and it is faster at the top of the leaderboard. For most players, judicious flagging is enough.
On the other hand, if you are still learning, flag liberally. Flags are a thinking tool. They tell you what you already know so you do not re-derive the same constraint over and over. There is no shame in playing slow and correct.
Avoiding self-induced 50/50s
A 50/50 is when two squares are equally likely to be a bomb and your local information cannot resolve it. On a true no-guess board you should never have one. But you can accidentally create them by revealing things in the wrong order.
Imagine a corner with three unrevealed squares. The information you need to resolve it lives two squares away. If you blindly open the corner first, you might trigger a cascade that loses the extra constraint you needed. So the rule is: when you have a choice, open the square that gives you the most information, not the square that feels most urgent.
Concretely, this means working from the edges of the cleared region inward. The cells along the edge of the cleared region are the ones with constraints touching them. The cells deep in the unrevealed dark are floating in the dark. Open the constrained ones first.
When you actually do have to guess
On a no-guess board, the answer is to look harder. But if you ever play on a random board that is not guaranteed solvable, here is what to do.
Count. For each remaining hidden square that has any constraint touching it, work out its probability of being a bomb. The square with the lowest probability is the safest bet. Squares with no constraint at all carry the global density, which on Expert is about 20 percent. Constrained squares are usually higher or lower than that.
If your safest square is at 15 percent, take it. If your safest is at 50 percent, accept that you might lose and move on. Players who freeze on a guess waste seconds without changing the odds.
The flow state of fast play
When you watch a world record video, the player is not thinking per click. They are pattern matching at speed. They see a 1-2-1, their hand executes the flag-flag-chord without conscious intervention, and they are already scanning the next region while their fingers finish.
That fluency comes from playing the same patterns hundreds of times until your hands know them better than your head. You do not need to be a speedrunner. But the same principle applies at every level: the more you play, the less you think per click, and the faster you finish.
Daily play is the best teacher
Variety beats grinding the same puzzle. MineRace ships a new puzzle every day at every size. Play the small one to warm up, play the medium to practice patterns, and play the large when you want a real session. Over a couple weeks of daily play your pattern recognition gets sharp without you trying.
And if you really want to push, study your losses. When you blow up, go back and look at what information was on the board that would have saved you. Eight times out of ten it was there and you missed it. The ninth time, you missed a global mine count clue. The tenth time, you really did guess wrong. The first eight are the ones to learn from.
Next steps
If you have not read it yet, the how to play page covers the basics like chording and the cascade. The rules page is the canonical reference. And if you are curious about how we generate puzzles that are guaranteed solvable, the no-guess page has the math.
Otherwise, go race. Today's puzzles are sitting on the front page.