Mines myths, and how provably fair mines boards are generated

Most mines "tricks" fail because every hidden tile has the same chance of being a mine and the board is fixed before you click; in a provably fair game, you can check that by recomputing the mine positions from the revealed seeds after the round. This guide tests six popular myths, then walks through one common way a hash becomes a set of mine positions.

Six myths, checked

A quick test of "corners are safer"

With 3 mines on 25 tiles, any single tile holds a mine with probability 3/25 = 12%. There are 4 corners, 12 other edge tiles and 9 inner tiles, and each of them is at 12%. In our 300,000-round simulation for the EV guide, tiles were picked at random and survival matched the formula within sampling error — no location effect exists to find.

How a provably fair board is built

A provably fair game commits to its randomness before you play. A typical setup uses three inputs:

• Server seed: a secret chosen by the game. You see only its SHA-256 hash before play.
• Client seed: a value you can set or change.
• Nonce: a counter that goes up by one each round.

After you rotate seeds, the old server seed is revealed. You hash it and check it matches the hash you were shown. Then you recompute the board.

The method below is one common convention for illustration. Games differ in message format, byte handling and shuffling. Always compare with the documentation of the game you're checking.

Worked example: seeds to mine positions

All seeds here are made up for this example. Settings: 3 mines, server seed grid-desk-example-seed-7d41c0, client seed grid-reader-01, nonce 3.

Step 1 — check the commitment.

sha256("grid-desk-example-seed-7d41c0")
= f07f03b130201cdf401db74edeceeabf7f5f7b6a8fc00d9fc7ec4f97fce0a59a

If that matches the hash shown before play, the seed wasn't swapped.

Step 2 — make the random bytes. HMAC-SHA256 with the server seed as the key and client:nonce:round as the message:

HMAC("grid-reader-01:3:0")
= c95f69878de7a402c51065b8cd9988c7f38eaefa50aabe2e403ae2a0e30eb8ca

Step 3 — turn 4 bytes at a time into a number from 0 to 1. Each byte is divided by 256, 256², 256³ and 256⁴ and added up.

Step 4 — pick tiles without replacement. Number the tiles 0–24, left to right, top to bottom. Multiply each number by how many tiles are still available, round down, and take the tile at that position in the remaining list.

Note the third row: index 17 points to tile 18, because tile 13 was already removed from the list. Picking without replacement means a tile can never be chosen twice.

Because each step is fixed, anyone with the three inputs gets the same board. If you played this fictional round and your history shows a mine somewhere else, the result doesn't match the commitment and you should raise it with the site.

Why the client seed and nonce matter

The server seed alone would let the game choose boards in advance. The client seed is your contribution: because the game commits to its seed before knowing yours, it can't pick a server seed that produces boards it likes for you. The nonce makes each round different without changing either seed, so one pair of seeds can produce thousands of separate boards. If the nonce in your history skips or repeats, that's worth asking about.

Changing your client seed doesn't make mines easier to avoid. It changes which boards you get, not the chance of any tile holding a mine.

Reproduce it in Python

import hashlib, hmac
server, client, nonce, mines = "grid-desk-example-seed-7d41c0", "grid-reader-01", 3, 3
print(hashlib.sha256(server.encode()).hexdigest())

def numbers():
    rnd = 0
    while True:
        h = hmac.new(server.encode(), f"{client}:{nonce}:{rnd}".encode(), hashlib.sha256).digest()
        for i in range(0, 32, 4):
            yield sum(h[i + j] / 256 ** (j + 1) for j in range(4))
        rnd += 1

tiles, chosen, gen = list(range(25)), [], numbers()
for _ in range(mines):
    chosen.append(tiles.pop(int(next(gen) * len(tiles))))
print(chosen)   # [19, 13, 18]

What a valid proof does and doesn't show

It shows the board for that round came from seeds fixed before your first click, so mines weren't moved in response to your picks.

It doesn't show that the multipliers are fair (check them against 1 ÷ P using the multiplier guide), that the site is licensed, or that you'll win. The house edge is in the payouts, not the board.

Verification checklist

• Save the server seed hash before you play.
• Note your client seed and the nonce of the round you want to check.
• Rotate seeds to reveal the old server seed.
• Hash the revealed seed; it must match the saved hash exactly.
• Recompute the board with the game's documented method.
• Compare with the mine positions shown in your history.
• If the method isn't documented, treat the fairness claim as uncheckable.

New to the game? Start with how the mines game works.