Everyone at the table dives the same water against the same hull, on one clock. You are not steering — there is nothing to steer. There are three decisions, and the third is the only one that happens while the round is running.

• Your stake — set before the doors close. Once the round launches it is committed, and it is the whole of what you can lose.
• A standing order, optionally — a multiplier to get out at automatically. It fills the instant the round reaches that number, whether or not you are looking at the screen, and it is the only way to be paid at a figure you chose in advance rather than one you reacted to.
• When to surface — the one live decision. Every cash-out target has the same expected value, so this is a choice about variance, not about edge: getting out at 1.5× often and 20× rarely cost the same 3% to the house over enough rounds.

You do not have to wait for the doors. Half of all rounds are over inside ten seconds and the longest one possible is a hundred, against a six-second lobby, so the doors are shut most of the time — press DIVE anyway and the bet goes in when the next lobby opens. Nothing is taken until it does, so cancelling costs nothing, and the table places queued bets in the order they were asked for rather than in the order the network happens to deliver them. A queued bet can still be refused when the doors open, for the same reasons any bet can: the gates and the ledger run then, on the balance you have then.

Auto-bet places the same stake for a set number of rounds — ten, twenty-five or fifty — and shows what is left of the count. It stops when the count runs out, if the stake goes above your balance, the moment you press STOP — which is what the DIVE button says while it is running — and when a bet is refused for any reason but a shut door, a full table or the wager clock. Those three cost the round rather than the run: the count goes back, so a run always places the number of stakes you asked for and never one more. It is deliberately not a switch you leave on: a run has an end, and continuing past it is something you decide again. It changes nothing about the odds or the return.

Miss a round and you watch it finish without you. The table does not wait, and it keeps running whether anyone is betting or not — so the strip of recent dives shows rounds that really happened, not a demonstration. The dimmer ones are rounds you sat out; they show what the table did anyway.

The dents and the shed plating, the seams that start to leak, the creaking and the lamps beginning to stutter are all functions of your current depth alone. They are consequences of where you are, not predictors of where the hull gives out. Nothing on screen leaks the failure depth.

Draw u uniform on (0,1). If u > 1 − e the hull fails instantly at 1.00× — that branch is the entire house edge and it happens 3% of the time. Otherwise the failure multiplier is B = (1 − e) / u, clamped to the seafloor cap. So P(B ≥ m) = (1 − e)/m, and cashing out at any target m returns m · (1 − e)/m = 1 − e. Every cash-out target has identical expected value. No strategy beats another.

B is continuous, not rounded to 2dp. Rounding it would put a real probability atom on every displayed multiplier and force a tie-break at the failure depth; resolving those ties in the house's favour would quietly cost an auto-surfacing player up to 0.87 points of RTP. Displays round. The maths does not.

• House edge3%
• Return to player97%
• Instant failure at 1.00×3%
• Depth → multiplierM = G(d) · e^{d/1592.4131003119235}
• Salvage gain G(d)1× to 8×
• Max payout (seafloor)1000×

G(d) is the salvage gain: crates on the way down each multiply the curve by 1.25×, 1.5× or 2× and stay multiplied, so two players at the same depth in different rounds can be at very different multipliers. It never changes your odds of failing — the hull is drawn before you launch and salvage only moves what a given depth is worth, which is why the liability cap arrives shallower in a rich round rather than letting a payout past the maximum. Mines are the other half: they speed the descent for a few seconds, which reaches a depth sooner without changing what it pays.

Failure odds per second are shown live at the foot of the depth tape. Depth does not change them: the hull is no more likely to fail deep than shallow, which is what makes every cash-out target worth the same. The figure moves only while something in the water is driving the sub down faster, and it goes back when that passes.

A server seed is committed as a SHA-256 hash before each dive and revealed after it. The failure depth comes from HMAC-SHA256(serverSeed, entropy:nonce), where the entropy is the public commitment published for this whole table before the chain was used. This table runs shared rounds — one hull, one number, everybody on it — so the salt is that one public value rather than a seed each player picks, which is why the client seed field below is shown and disabled rather than removed: it is what a solo table would use.

Next dive — committed now, revealed after

Check it yourself —

This build has no limit register, so nothing you set here would be stored. The boot banner says the same thing and mainnet is refused while it is true.

Ends this session on this device. Your balance, your limits and any exclusion are untouched — this is not a break from playing, and it is not how you take one.

Keys: space or enter dive · space surface · X speak position · N speak session. Both are printed on the buttons.

Playing without sight: nothing is read out on a timer, because a number that changes hundreds of times a second cannot be narrated honestly and a stream of it in your ear would be an arousal device rather than an aid. Dive starts and results are announced; everything else is on request, with X and N or the “Read current position aloud” button at the start of the tab order. Auto-surface is the reliable way to play without watching — set a target before the dive and the game exits at exactly that multiplier, with no reaction time involved. The announcement gives the multiplier, what came back, and where the dive left you.