Mines

Nothing about the board is hidden, and the mines are no longer merely asserted — they are published. The chain produces verifiable randomness (Chainlink VRF on Arbitrum, Switchboard on Solana — committed before your bet and verified at reveal), your bet carries the map of tiles you have already opened, and the program draws the mines among the tiles that map says are STILL FACE DOWN. The strike is settled first — with r tiles face down and M mined, you bust when randomness mod r lands in the top M ranks, so survival is exactly (r − M)/r — and only then are the other mines arranged. That order is the point: the arrangement decides which mines you are shown, never whether you won. When you bust, the chain publishes the whole mined board for that press, and you can redraw it from the same published value. A surviving press publishes nothing, on purpose: every press re-draws over the tiles still face down, so posting a board after each one would show mines that appear to move — and a tile you have already opened is excluded from every later draw, so it stays safe forever. Every step pays ⌊98·r/s⌋ percent (s = r − M), from a table generated from a single source and compiled into the program. That floor is the honest part: a two-outcome bet has one payout knob and the label grid is 0.01×, so the return lands at 98% or just under it, never above — 97.14% to 98.00% across the table, a house edge of 2.00% to 2.86%. That figure is PER STEP. A climb is a chain of separate bets that each paid their own edge, so an eight-rung climb returns about 0.98⁸ of what was staked through it, not 98% — the board shows your real running total instead of a headline the maths cannot keep, and each rung prints its own return beside its own multiplier.