How hard is a Minesweeper board?
A GregSweeper daily takes between twenty seconds and four minutes. The game predicts which, for that specific board, before anyone has played it, and calls the prediction par.
Two boards the same size are not the same puzzle
Put the same number of mines in the same number of squares twice and you get two puzzles that play nothing alike. On one, the first click opens half the grid and the rest falls out in seconds. On the other you stall three times and work for every square.
The familiar difficulty labels do not describe that. Beginner, Intermediate, and Expert are set by the grid and the mine count, so every Expert board wears the same label, and anyone who has played a few hundred of them knows they are not equally hard.
What 3BV measures, and what it leaves out
The measure the Minesweeper community uses is 3BV, short for Bechtel's Board Benchmark Value. It counts the fewest left clicks that can clear a board if you never chord, chording being the click on an already-satisfied number that opens all of its remaining neighbors at once.
As a count of clicking, 3BV is exact, and speedrunners compare times with it for good reason. What it leaves out is how much thinking a click costs. Take a square because one neighboring 1 is already satisfied, or take it only after comparing four overlapping clues, and 3BV counts one click either way.
What par counts instead
A solver checks every daily, weekly, and ladder board before anyone sees it, because GregSweeper guarantees there is always a next move you can prove. That check records how each square was settled, and par is built from the record.
Moves are filed by the deduction that opened them:
- Pattern moves, where one clue and its neighbors settle the square. These are the shapes an experienced player reads without slowing down.
- Search moves, where no single clue is enough and several constraints have to be weighed together before anything is certain.
Both are counted per square, so a big gentle board and a small vicious one are described on one scale. Board size, mine density, the shape of the open areas and the number of each special modifier go in as well. Move type is the signal the rest of the model is arranged around.
So how hard is hard?
Daily par runs from twenty seconds at the easy end to four minutes at the hard one. Most days are drawn from the gentler part of that range, and roughly one in seven from the harder part, so in a given week you can expect one board that makes you work.
Weekly boards are bigger and the range shifts with them, from one minute to six.
Those are predictions for a typical player. What you personally should expect is par adjusted by your handicap.
Can Minesweeper be solved without guessing?
On these boards, yes, and that is the point of checking them first. Every daily, weekly, and ladder board has a proven path from the marked opening square to the end, so a 50/50 coin flip is never the reason you lose.
Removing the guesswork removes the luck rather than the difficulty. Once nothing can be settled by a coin flip, every remaining square has to be earned, and the twenty seconds to four minutes above is the spread of how much earning that takes.
What a Minesweeper handicap is
Par describes the board and says nothing about who is playing it, which is no use for comparing two people.
So each player also gets a handicap, in the golf sense: a multiplier fitted from your own past games. Finish consistently in three quarters of par and your handicap is 0.75, and your personal par on any board becomes three quarters of that board's par. Below 1 means faster than the model's typical player, above 1 means slower.
That is what makes the daily leaderboard a fair comparison: a quick player and a slow one play the same board, and each is scored against their own expectation. New players get a provisional handicap from their first few games, which settles toward the fitted one as games accumulate.
Where the number comes from
Par is fitted, not hand-tuned. A Bayesian mixed-effects regression runs every night over completed daily and weekly games, with each player estimated as their own effect so that the board coefficients describe boards rather than the handful of people who play the most.
One consequence is worth knowing: par moves. A board priced today may be priced a little differently next week, once more games have accumulated for that combination of features.
Everything the model looks at
This list is read from the same file the game itself uses, so it is whatever the fit shipped last night rather than a description written once and left behind.
Play today's Minesweeper board →One puzzle a day, the same board for everyone, and always solvable by logic.