A no-guess Minesweeper puzzle is one where you can solve the entire board through pure logic. You never have to guess. There is no 50/50 waiting at the end. Every safe square can be proven safe from the information already on the board, given enough thought.
On a random Minesweeper layout this is not guaranteed. Most random Expert boards contain at least one position where a player is forced to flip a coin. You play perfectly for 80 seconds, you reach a corner with two squares and not enough constraints, and you lose because you picked the wrong one. That has always been the irritating part of the classic game.
No-guess generation is the fix. The puzzle generator builds a random arrangement of bombs, then runs a solver to verify that the arrangement can be solved from the starting position without guessing. If the solver gets stuck, the generator throws the board away and tries again. The only boards that ship to players are the ones a logical solver could solve cleanly.
Why no-guess matters for a leaderboard
Imagine two players race the same random board. Player A finishes in 38 seconds. Player B finishes in 42 seconds but had to flip a coin at the end and was lucky. Did Player B play better? Statistically the leaderboard rewards luck as much as skill.
On a no-guess board, that confusion goes away. If you finished, you deduced. If you blew up, you missed information that was on the board. The leaderboard tracks skill, not coin flips.
This is the entire reason MineRace ships no-guess puzzles. We are a competitive site. We want the daily race to reward the player who read the board fastest, not the player whose final corner happened to break the right way.
How the solver works
Our solver uses constraint propagation. Each revealed number on the board is a constraint that says: out of these specific unrevealed neighbors, exactly this many are bombs. The solver iterates over these constraints applying three rules until either every square is classified or no rule fires.
Rule 1: trivial deductions
For each numbered cell, check whether its remaining bomb count equals the number of unrevealed unflagged neighbors. If so, those neighbors are all bombs. Or check whether the remaining bomb count is zero. If so, those neighbors are all safe. This handles most of the easy work in the early and middle game.
Rule 2: subset reduction
For each pair of constraints, check whether one constraint's unrevealed cells are a subset of another's. If so, you can subtract. The difference of cells must contain the difference of bomb counts. This is the generalization of the named patterns like 1-2-1, 1-2-2-1, and similar shapes that show up on every Expert board.
Subset reduction is where most of the actual work gets done. The named patterns players memorize are just the most common shapes this rule produces. If you spot a 1-2-1 in your head, you are running subset reduction without naming it.
Rule 3: the global mine count
The total number of bombs left equals the total number of bombs on the board minus the number of correctly flagged ones. If the remaining bomb count equals the number of remaining hidden cells, they are all bombs. If the remaining bomb count is zero, they are all safe.
The global constraint resolves several classic endgame situations that otherwise look like 50/50s. A player with the discipline to count the remaining bomb counter and compare it against the local constraints can finish boards that look impossible.
Local perturbation
Generating no-guess boards by pure trial and error is slow. Most random Expert boards fail the solver test. So our generator uses a trick called local perturbation. When the solver gets stuck, we do not throw the whole board away. Instead we swap one bomb for one safe cell, somewhere near the region where the solver got stuck, and try again.
This is the same approach Simon Tatham uses in his classic Mines puzzle in the Portable Puzzle Collection. It converges much faster than full re-rolls. Our generator typically produces a valid no-guess Expert board in under a second.
Determinism and shareable puzzles
Every MineRace puzzle is generated from a deterministic seed based on the date and the difficulty size. So everyone on Earth playing today's medium puzzle is playing the exact same board. That is what makes the daily leaderboard fair.
The seed also produces a short puzzle id. You can share any past puzzle by url. When a friend opens the link, our backend recovers the same date and size from the id and regenerates the exact same board. No central database lookup, just deterministic seeding.
The math problem in the background
A neat fact about Minesweeper: deciding whether a partially revealed board is consistent is NP-complete. Richard Kaye proved this in 2000. So in the worst case, no fast algorithm can answer the question for arbitrarily large boards. Practical instances on Expert are small enough that our solver handles them in milliseconds, but the worst-case complexity is a fun piece of theory.
For very hard endgame positions, a stronger approach is to fall back to a SAT solver and enumerate consistent bomb arrangements. We have not needed that yet on Expert. Our constraint propagation plus global count is enough.
What no-guess does not mean
No-guess does not mean easy. You still have to find the deductions. A no-guess Expert board can absolutely take a beginner several minutes. The solver proves the information exists somewhere on the board. Your job is still to find it under the timer.
No-guess also does not mean unique solution. Often there are multiple correct orders to reveal cells. The constraint is only that every cell can be proven safe or proven bomb at some point during play. Two players can solve the same board with different click orders and both be playing correctly.
Why we share the math
We could just say "all our puzzles are solvable" and leave it at that. But Minesweeper attracts the kind of player who wants to know how the sausage gets made. So we documented it. If you want to dig deeper, the right keywords are constraint satisfaction problems, subset deduction, Kaye 2000, Studholme 2006 on interactive constraint Minesweeper, and Simon Tatham's Mines source code.
The short version is: every puzzle on MineRace can be solved through pure logic, and the leaderboard reflects who did it fastest. Today's puzzles are on the home page. Or read about strategies for the patterns that actually move your time.