Crash gambling explained: one round, three numbers

A crash round has fewer moving parts than almost anything else in a casino lobby: a number starts near 1.00, rises, and stops. You choose when to take the number, and if the round stops first you take nothing. That is the whole game, which is why crash gambling is worth explaining with arithmetic rather than adjectives. This page works through the round, the multiplier, and the commitment scheme that lets a player check a result afterwards. It also counts which studios supply these games across the thirty-three lobby lists we hold, and states plainly the three things about them our files do not contain.

One round, described in order

A round runs in four stages. First a betting window, during which stakes are accepted and no multiplier is moving. Then the round starts and a displayed number begins at or just above 1.00 and climbs. At any instant during the climb a player may take the current number, and the stake is multiplied by it. At some point the round stops, and every stake not already taken is lost.

Everything interesting is in the third and fourth stages, and the relationship between them is the only thing a player controls. The stopping point is determined before the climb begins — it has to be, because it is the number the round was built around — and the climb is a display of time passing, not a process that decides anything. A player who takes 1.42 and a player who takes 8.10 in the same round were both acting against the same already-fixed stopping point.

Two facilities are standard and both are just ways of pre-committing the decision. An auto-cashout takes the number automatically when the display reaches a chosen value. An auto-bet repeats the stake for a set number of rounds. Neither changes the round; both change how consistently a player’s chosen number is applied, which matters because the entire distinction between one crash gambling session and another is the number chosen and how often it is stuck to.

The game is also unusually legible for a reader trying to compare sites, because the round produces a single public value — where it stopped — and many implementations keep a visible history of those values. That makes it one of the few lobby games whose outputs a player can write down.

Who supplies these games across thirty-three lobby lists

Thirty-three of our hundred operator records carry a list of the studios named on the site’s own lobby. Counting the studios associated with crash-style and instant games across those lists gives the following.

Two counting notes, because both bit us. Turbo Games appears as Turbo Games, TurboGames and Turbogames; SmartSoft as SmartSoft, Smartsoft, SmartSoft Gaming and Smartsoft Gaming. Tally the raw strings and you undercount both studios and invent several that do not exist. The figures above are after merging, and the merges are listed in our source notes so they can be disputed.

What the table does not say is which specific games sit behind those studio names. Our files record supplier strings only — no game titles anywhere, for any studio, on any site. So we can tell you a studio associated with crash gambling is named on nineteen of thirty-three lobby lists. We cannot tell you which of its games are switched on, and we will not guess from the studio name.

The two tags we can count

Our operator records carry a small set of tags taken from the operator’s own site description. Two of them bear on this subject. Twenty records carry original-games — in-house games rather than third-party studio content, which is where crash-style rounds usually live. Twenty-one carry provably-fair. Twelve carry both.

The overlap is the interesting part, because it is far from complete. Nine records carry provably-fair without original-games: Cloudbet, KatsuBet, LTC Casino, Mega Dice, Mirax, Stake, CryptoGames, bspin, Primedice. Eight carry original-games without provably-fair: BetFury, BitStarz, Bitcasino.io, Rocketpot, TrustDice, Fairspin, Sol Casino, Punt Casino.

That second group is the one to keep in mind. In our library, a site can advertise in-house instant games without describing them as verifiable, and eight records do exactly that. It follows that “has crash-style games” and “lets you check a crash-style result” are different claims resting on different site text, and anyone reading a crash gambling comparison should expect the two to be conflated. We keep them in separate fields for that reason.

A caution on the tags themselves: they come from how the operator describes its own site, not from an inspection of the game code. A tag is a claim we recorded, with a reading date attached. It is not a test result, and we have run no test of any crash round for this page.

Crash gambling arithmetic: the multiplier, with substituted numbers

Every number in this section is a substitution chosen to make the algebra visible. None of them is a reading of any game, from any studio, on any site in our files — our library contains no return-to-player figure and no multiplier distribution for anything.

Start from the one relationship that has to hold. If a round reaches a multiplier of m with probability p, then a stake of 1 taken at m returns m when it survives and 0 when it does not, so its expected return is m × p. For that expected return to equal the stake exactly — a game with no margin at all — you need m × p = 1, which is p = 1/m. A target of 2.00 would need to be reached half the time. A target of 10.00 would need to be reached one time in ten. A target of 100.00 one time in a hundred.

Now put a margin in. Suppose, as a substitution, a stated margin of 1%. Then the designer sets survival probability to 0.99/m instead of 1/m, and the expected return on any target becomes m × 0.99/m = 0.99 of the stake. Substitute 3%: every target returns 0.97. The key property is that m cancels. Whatever target you pick — 1.10, 2.00, 50.00 — the expected return is the same fraction of your stake, because the probability was set by dividing by the very number you chose.

This is the single most useful piece of arithmetic in crash gambling, and it is arithmetic, not a reading. It says the choice of cashout number is a choice about the shape of the results, not about their average. A target of 1.10 wins very often and adds very little each time. A target of 50.00 wins very rarely and adds a great deal. With the margin held constant, both grind down to the same expected fraction, and the difference between them is entirely how long you wait and how far the swings go.

What the swings look like when the average does not move

Continue the substitution to see why two players with the same expected return can have completely different afternoons. Take a no-margin target of 2.00 with survival probability 0.5, and a hundred rounds at one unit each.

• At a target of 2.00, expect about 50 wins of 2 units, so about 100 units back from 100 staked. The longest losing run in a hundred coin-flips is commonly six or seven; it costs six or seven units.
• At a target of 20.00 with survival probability 0.05, expect about 5 wins of 20 units — again about 100 back from 100 staked. But the gaps between wins average nineteen losing rounds, and a gap of sixty is entirely ordinary in a hundred rounds. The same expectation arrives in a completely different shape.

Both lines return the same substituted average and neither is a prediction. The point is the variance, and it is why the number a player types into an auto-cashout box is the only meaningful decision in the game: it does not move the average, and it moves everything else.

One more arithmetic consequence worth stating flatly, because it is where most writing about crash gambling goes wrong. Past stopping points carry no information about the next one if each round is generated independently, and independence is exactly what a commitment scheme is designed to make checkable. A run of low stopping points does not make a high one due. There is no arithmetic in this section, or any other, that produces a positive expectation from a negative-margin game by choosing when to bet. The claims that try are collected in the claims about crash rounds that the arithmetic does not support.

What the commitment scheme adds, in letters

The mechanism usually labelled provably fair is a commitment, and it can be described without a single number. Before the round, the operator picks a secret value — call it the server seed — and publishes a hash of it. A hash is a one-way summary: easy to compute from the value, impractical to reverse. The player contributes a value of their own, the client seed, which the operator cannot know in advance. The round’s outcome is derived from both seeds combined, by a stated rule. Afterwards the operator reveals the server seed, and anyone can hash it to confirm it matches the commitment published beforehand, then re-run the stated rule to confirm the outcome follows from the two seeds.

What that proves is narrow and worth stating exactly: it proves the operator did not choose the outcome after seeing the stake, because the commitment was fixed before. What it does not prove is that the margin is what the site says, that the distribution of outcomes across many rounds is what was advertised, or that the rule itself is favourable. A verified round can be a round from a badly weighted game.

Twenty-one records in our library carry the provably-fair tag, and the tag records the site’s own description rather than an audit. We have not verified a single round, for this page or any other, so nothing here is a test result. What we can do is say where the claim sits: it is a site-description claim, it appears on twenty-one of a hundred records, and on nine of those it appears without any accompanying in-house games tag at all. The mechanism itself is written out in the commitment scheme in full, hash by hash, and the same seed pair drives the same seed pair applied to a dice roll and how a grid game reuses the identical commitment.

The money clauses are not in the game

A crash round settles instantly on screen. Getting the balance off the site is governed by a different document, and in our library those two facts are recorded in different fields for a reason. Several operators whose lobby lists include crash-style suppliers also publish a payout cap: Bitcasino.io 1,000,000 USDT a week (clause 6.10), Wild.io $100,000 a week (9.6), CoinCasino €500,000 a month (9.6), CryptoLeo 50,000 USDT a month (7.9), Instant Casino €4,000 a day (8.1), BetPlays €5,000 a week (6.11), Bets.io $10,000 a transaction (clause 10).

Verification wording is a third field again. Empire.io, whose lobby list runs to 79 studio names, states additional checks above 2,000 USDT in clause 5.4. Bitcasino.io, with 120 names, states €2,500 in clause 6.6. Windice, with 44 names, reserves documents at any time in clause 15.1 with no figure at all. All three were read in the same week as the lobby lists.

So a page about crash gambling that stops at the round has described the fast half of the transaction and skipped the slow half. The round resolves in seconds; the cap, the clause and the document period are measured in days and weeks. Both halves are in our files, under separate keys, with separate clause numbers and separate reading dates, and the money half is kept in the caps and clauses that decide how the balance leaves.

Three things our files do not contain

No return-to-player figure, for any game, from any studio, on any site. Our provider field holds studio name strings and nothing else, so any percentage attached to a specific crash game on this site would have come from outside our sources, and we do not print figures we did not read.

No multiplier distributions or curve constants. We cannot tell you how often a named game reaches 10.00, because that is a property of the game’s implementation and no field in our library describes it. The arithmetic earlier on this page derives relationships from a substituted margin; it does not report a measurement of anything.

No game titles. Not one, anywhere in the library. This is the limit that most surprises people, because studio names look so much like game names in a lobby list. A studio associated with crash gambling being named on nineteen lists tells you the supplier is switched on somewhere in that lobby. It does not tell you which game, at what stake range, or with which settings.

What we do hold, for the record: 100 operator records checked on 26 August 2026, of which 38 were read live, 41 from archived copies of the same pages and 21 not at all; 33 lobby lists; 49 verification summaries; 38 country lists; 14 payout caps; and register lookups against two public sources keyed by domain, taken the same week. That set is what builds the top no KYC casinos whose lobby lists we counted. Every number on this page comes out of that set, and the substituted figures are labelled as substitutions where they appear.