What Stake publishes about Crash
Crash is a shared round: one multiplier climbs from 1.00x for everyone at the table and stops at a point fixed in advance. A bet wins if you cash out before the stop, at the multiplier showing when you do. The game page carries an RTP badge of 99.00% (1 Nov 2026). For the mechanics, Stake's fairness documentation does not restate anything; it says "See the BitcoinTalk seeding thread", and that thread is where the formula lives.
| Item | What Stake states | Where |
|---|---|---|
| RTP | 99.00% | RTP badge on the Crash game page (1 Nov 2026) |
| House edge in the formula | "0.01 will result in 1% house edge with a lowest crashpoint of 1" | BitcoinTalk seeding post, 8 Jul 2019 |
| Hash chain | "a chain of 10,000,000 SHA256 hashes", each the hash of the previous one | Same post |
| Last hash of the chain | 78a9757d3be42b74a3f70239078ad9317125fe9ee630d5bdada46de963e56752 | Same post |
| Public salt | Hash of Bitcoin block 584,500, "which has not been mined at time of posting" | Same post; the block hash was added in a follow-up |
| Minimum bet | $0.0001 or the equivalent | Help centre, "Crypto: Casino Betting Limits" (12 Nov 2026) |
| Maximum bet and profit | Per coin for all Stake Originals, e.g. 100 BTC bet and 20 BTC profit | Same help-centre article (12 Nov 2026) |
Crash differs from Dice, Mines or Plinko in one structural way. Those games draw from your personal seed pair. Crash has no client seed of yours in it, because every player in a round must get the same result. The role your seed plays elsewhere is taken by a public value nobody could choose: the hash of a Bitcoin block that had not been mined when Stake committed to its chain.
The formula, line by line
The seeding post gives the whole computation in six lines of code:
gameHash = hashChain.pop()
hmac = HMAC-SHA256(key = gameHash, message = blockHash)
int = first 8 hex characters of hmac, read as a number (0 to 4,294,967,295)
crashpoint = max(1, 2³² / (int + 1) × (1 − 0.01))
Read in plain terms: take the round's hash, mix it with the block hash, turn the first 32 bits into a number, and divide. A small number gives a large multiplier; a number near the top of the range gives a multiplier near 0.99, which the max(1, …) lifts to 1.00x. The factor 0.99 is the entire house edge.
The block hash is public. Stake's follow-up post gives it as 0000000000000000001b34dc6a1e86083f95500b096231436e9b25cbdd0075c4, and block 584,500 on the Bitcoin chain carries exactly that hash, with a block time of 15:00 UTC on 8 July 2019, the day of the post. The order matters: Stake published the end of its chain first and let a later block supply the salt, so it could not have picked a chain that suits a known salt.
How often each multiplier is reached
From the formula, a round reaches a target t when the 32-bit number is small enough, which happens with probability 0.99 divided by t. Paying t on a 0.99/t chance returns 99% at every target.
| Cash-out target | Round reaches it | That is about | Round ends before it | Expected return |
|---|---|---|---|---|
| 1.01x | 98.02% | 98 in 100 | 1.98% | 99.0% |
| 1.1x | 90.00% | 90 in 100 | 10.00% | 99.0% |
| 1.5x | 66.00% | 66 in 100 | 34.00% | 99.0% |
| 2x | 49.50% | 1 in 2.0 | 50.50% | 99.0% |
| 3x | 33.00% | 1 in 3.0 | 67.00% | 99.0% |
| 5x | 19.80% | 1 in 5.1 | 80.20% | 99.0% |
| 10x | 9.90% | 1 in 10 | 90.10% | 99.0% |
| 20x | 4.95% | 1 in 20 | 95.05% | 99.0% |
| 50x | 1.98% | 1 in 51 | 98.02% | 99.0% |
| 100x | 0.990% | 1 in 101 | 99.010% | 99.0% |
| 1,000x | 0.099% | 1 in 1,010 | 99.901% | 99.0% |
Arithmetic from the published formula. Probabilities are exact to within one part in four billion, the resolution of the 32-bit number.
Three things stand out. Half of all rounds end below 1.98x, so the median crash point is just under 2x. One round in a hundred has a number so large that the formula gives less than 1 and the round is set to 1.00x; a bet on any target loses in those rounds. And the expected-return column does not move: a 1.1x target and a 100x target cost the same 1% of the amount staked, paid in very different patterns. An auto cash-out at a low target wins most rounds and loses the whole stake in the rest; a high target loses most rounds and occasionally pays a multiple. Neither is cheaper.
The same arithmetic covers runs. The chance that ten consecutive rounds all fail to reach 2x is 0.505 to the tenth power, about 0.11%, or once in roughly 930 ten-round windows. With a round every few seconds that occurs routinely, and it carries no information about the eleventh round, whose hash was fixed years ago.
The profit cap decides the real ceiling
The formula itself has no maximum that matters: the smallest possible number would give a multiplier in the billions. What limits a win is Stake's per-coin cap on profit from one bet on an Original. The table shows the largest stake that a cash-out at each target still fits under the cap.
| Cash-out target | USDT (cap 2,500,000) | BTC (cap 20) | ETH (cap 700) |
|---|---|---|---|
| 2x | 2,500,000 USDT | 20.00 BTC | 700 ETH |
| 10x | 277,778 USDT | 2.22 BTC | 77.78 ETH |
| 100x | 25,253 USDT | 0.20202 BTC | 7.07 ETH |
| 1,000x | 2,503 USDT | 0.02002 BTC | 0.700701 ETH |
Caps: Stake help centre, "Crypto: Casino Betting Limits", 12 Nov 2026. Stakes are our arithmetic: cap divided by (target minus 1).
The help-centre article states the caps and not the mechanism, so it does not say whether a running bet is cashed out automatically when its profit reaches the cap. Anyone letting a large stake ride to high multipliers should read the limit shown in the bet panel first.
Verifying a round: the chain and the point
There are two separate things to check, and neither needs your account's seeds.
- The round belongs to the committed chain. Each round shows its game hash once it is over. The SHA-256 of that hash must equal the game hash of the round played immediately before it. Repeating the step walks back toward the hash Stake posted in 2019. A round whose hash did not come from the chain would break the link.
- The crash point follows from the hash. Put the game hash and the block-584,500 salt through the formula above. The result, shown to two decimals, must be the multiplier at which the round ended.
Our Crash verifier and the Stake seed checker do both from a pasted game hash: they print the previous round's hash for the chain check and compute the crash point with the salt pre-filled. Stake's own calculation page has a Crash entry that takes the hash and the salt.
The limits of the proof are worth stating. It shows the sequence of crash points was fixed before any of today's bets and that Stake cannot steer a round in response to who is in it. It does not make any cash-out target profitable, and it does not cover what happens between your click and the server: a cash-out that arrives after the stop is a loss, which is why an auto cash-out set before the round is the only instruction that cannot be late. More background: what provably fair gambling proves.
Stake's Slide game uses the same salt-and-chain model with its own chain and a stated 98% RTP. Aviator-style titles in the lobby come from outside studios and are not part of this chain.
Stake Crash against Shuffle, Roobet and Gamdom
All four casinos run Crash on a hash chain. The formulas at the end of the chain are not the same, and they price the game differently.
| Stake | Shuffle | Roobet | Gamdom | |
|---|---|---|---|---|
| Chain length | 10,000,000 hashes | 10,000,000 hashes | 2,000,000 hashes | 10,000,000 hashes |
| Salt | Bitcoin block 584,500, unmined when committed (Jul 2019) | Bitcoin block 779,588, announced before it was mined (Mar 2023) | Ethereum block 26,008,955, named before it was mined (Sep 2026) | Hash of Bitcoin block 339,300 (mined Jan 2015) |
| Rounds forced to 1.00x | 1 in 100, from the max(1, …) step | Not documented | 1 in 25 | 1 in 100 |
| Stated return | 99% RTP | 99% RTP at every target | None stated | 99% after the first $5,000 of daily Originals wagering |
| Return at a 2x cash-out, from the published formula | 99.0% | Formula not published | 95.5% | 98.0% |
| Where the formula is published | BitcoinTalk seeding post | Chain and salt in the help centre; multiplier step not given | Verifier linked from its fairness page | Sample code linked from its fairness page |
Shuffle: help centre and game page, 12 Nov 2026. Roobet and Gamdom: verifier code linked from each fairness page, 12 Nov 2026. Returns are our arithmetic from each published formula.
Stake's version is the cheapest of the three whose formula can be read, and the only one of the four where a stated return and a published formula can be set side by side and agree. The individual pages go through each casino's code.
Sources: Stake.com Crash seeding event, BitcoinTalk, 8 Jul 2019 (read 12 Nov 2026); Bitcoin block 584,500; Stake Crash game page and fairness documentation, game events (1 Nov 2026); Stake help centre, crypto betting limits (12 Nov 2026). Probabilities and stake ceilings are our arithmetic from the formula and the caps.
The auto cash-out calculator shows hit rate and expected loss for any target. Stake's licence, KYC and withdrawal terms are in the Stake review.