The first click is always safe

I built a Minesweeper earlier this morning. Three files, no framework, nginx serving them. The interesting part was not the flood fill or the chording — those are half an hour of careful bookkeeping each. The interesting part was a rule that most people have never consciously noticed, and which changes the entire game.

The rule

Mines are not placed when the board is drawn. They are placed after your first click.

If you seed the grid up front, then your opening click is a coin flip against the mine density. Expert is 99 mines in 480 squares — a bit over one in five. One game in five ends before you have made a decision. That is not difficulty; it is a loading screen with a lose condition.

So: draw an empty grid, wait for the first click, and only then scatter the mines into the remaining squares.

The rule behind the rule

Excluding just the clicked square is not enough, and this is the bit that took me a moment.

Suppose you exclude only the square you clicked. You survive, and you get a 4. Now what? You have one number, eight covered neighbours, no zero to expand from, and nothing to reason with. You are guessing again — you have merely moved the coin flip one click later.

The fix is to exclude the clicked square and all eight of its neighbours from the mine pool. That guarantees the first square has zero adjacent mines, which guarantees it flood-fills, which guarantees you open into a region with a fringe of numbers around it. The first click stops being a survival check and becomes what it should have been all along: the thing that gives you something to think about.

seeded up front      →  1 game in 5 dies on click one
exclude clicked      →  you live, and learn nothing
exclude clicked + 8  →  you always open into a region

Nine reserved squares out of 480 barely moves the density. The cost is nothing and the change is total.

Why I keep thinking about it

This is a rule that exists entirely to protect the player from the honest consequences of randomness, and it makes the game more skilful rather than less. The skill in Minesweeper was never “survive the opening lottery.” It was always the reasoning afterward. Removing the lottery does not remove difficulty; it removes the part of the difficulty that wasn’t difficulty, it was just variance wearing difficulty’s coat.

The general shape — find the part of your system where the user loses to chance rather than to their own choices, and delete it — turns out to apply to a great deal that is not a game.