What Stake publishes about Dice
Dice is the simplest of the Stake Originals: one number is rolled between 0.00 and 100.00 and you bet that it lands over or under a target of your choosing. The game page shows an RTP badge of 99.00%, and Stake's house-edge guide lists "Stake Originals Dice: 99% RTP (1% house edge)" (both 1 Nov 2026). The fairness documentation describes the roll precisely: "we cover a possible roll spread of 00.00 to 100.00, which has a range of 10,001 possible outcomes".
| Item | What Stake states | Where |
|---|---|---|
| Roll range | 00.00 to 100.00, 10,001 possible outcomes | Fairness documentation, "Game events" (1 Nov 2026) |
| RTP / house edge | 99.00% RTP, 1% house edge | Game-page RTP badge and house-edge guide (1 Nov 2026) |
| Win chance you can set | 0.01% to 98% | Bet panel on the game page (1 Nov 2026) |
| Multiplier range | 1.0102x to 9,900x | Follows from the win-chance range at a 1% edge |
| Numbers used per roll | One, from a single HMAC-SHA256 round (cursor 0) | Fairness documentation, "Implementation" (1 Nov 2026) |
| 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) |
Stake's own description of the game says you "directly control your Return-to-Player (RTP), risk level, and payout potential through the settings". The second and third are true. The first needs a correction that the next section makes with numbers: the slider changes how often you win and how much, and the return stays at 99% wherever it sits.
Every setting pays 99 divided by the win chance
The bet panel ties two numbers together: the win chance in percent and the payout multiplier. Their product is always 99. A 49.5% chance pays 2x, a 9.9% chance pays 10x, a 0.99% chance pays 100x. Multiply any row's chance by its multiplier and the expected return is 0.99 of the stake.
| Multiplier | Win chance | Roll-over target | Roll-under target | Wins about | Profit on a $100 win |
|---|---|---|---|---|---|
| 1.0102x | 98.00% | Over 2.00 | Under 98.00 | 98 in 100 | $1.02 |
| 1.1x | 90.00% | Over 10.00 | Under 90.00 | 90 in 100 | $10.00 |
| 1.5x | 66.00% | Over 34.00 | Under 66.00 | 66 in 100 | $50.00 |
| 2x | 49.50% | Over 50.50 | Under 49.50 | 1 in 2.0 | $100.00 |
| 3x | 33.00% | Over 67.00 | Under 33.00 | 1 in 3.0 | $200.00 |
| 5x | 19.80% | Over 80.20 | Under 19.80 | 1 in 5.1 | $400.00 |
| 10x | 9.90% | Over 90.10 | Under 9.90 | 1 in 10 | $900.00 |
| 50x | 1.98% | Over 98.02 | Under 1.98 | 1 in 51 | $4,900.00 |
| 100x | 0.99% | Over 99.01 | Under 0.99 | 1 in 101 | $9,900.00 |
| 1,000x | 0.0990% | Over 99.90 | Under 0.10 | 1 in 1,010 | $99,900.00 |
| 9,900x | 0.0100% | Over 99.99 | Under 0.01 | 1 in 10,000 | $989,900.00 |
Arithmetic from the stated 1% edge: multiplier = 99 / win chance in percent. Roll-over and roll-under targets are the two equivalent ways to get the same chance.
Rolling over 50.50 and rolling under 49.50 are the same bet. So are over 90.10 and under 9.90. Switching direction is cosmetic, and so is nudging the target between bets: no position on the slider is cheaper than another.
One small precision point. The roll has 10,001 possible values, from 0.00 to 100.00 inclusive, while the win chance on screen is quoted against a base of 10,000. The effect is that the exact return of a setting sits within about one hundredth of a percentage point of 99%, a difference of a cent per $100 wagered. It is the kind of gap that only shows up on paper.
What a 1% edge means over a session
Because the edge is the same on every bet, the expected loss of a session is simply 1% of the total amount wagered, whatever mix of targets was used. A thousand bets of $1 cost $10 on average. A hundred bets of $10 cost the same $10. What differs is the spread around that average, and it is wide at high multipliers.
| Losses in a row | At 2x (49.5% win chance) | That is | At 10x (9.9% win chance) |
|---|---|---|---|
| 5 | 3.2844% | 1 in 30 | 59.4% |
| 10 | 0.1079% | 1 in 927 | 35.3% |
| 15 | 0.0035% | 1 in 28,225 | 20.9% |
| 20 | 0.0001% | 1 in 859,355 | 12.4% |
Chance of losing that many bets in a row, from the win chances above.
Ten straight losses at 2x happen about once in 928 sequences of ten bets, and at Dice speed a player gets through many such sequences in an hour. This is the arithmetic that undoes doubling systems: a stake that doubles after each loss has multiplied by 1,024 after ten losses, and the run that requires it is not rare. A progression rearranges when losses arrive. It cannot change the 1% taken from every unit staked, so the expected cost of the whole sequence is still 1% of everything wagered in it.
The profit cap at high multipliers
Stake's help centre caps the profit of one bet on a Stake Original by coin. At ordinary Dice settings the cap is out of reach, but the top of the range is 9,900x, where a win is a profit of 9,899 times the stake:
| Coin | Max bet | Max profit per bet | Largest stake whose 9,900x fits the cap |
|---|---|---|---|
| BTC | 100 | 20 | 0.00202 BTC |
| ETH | 1,000 | 700 | 0.070714 ETH |
| LTC | 100,000 | 10,000 | 1.01 LTC |
| SOL | 100,000 | 20,000 | 2.02 SOL |
| USDT | 25,000,000 | 2,500,000 | 253 USDT |
| XRP | 1,000,000 | 250,000 | 25.26 XRP |
| DOGE | 2,000,000 | 500,000 | 50.51 DOGE |
Limits: Stake help centre, "Crypto: Casino Betting Limits", 12 Nov 2026. Last column is our arithmetic.
At 2x the same cap allows a stake equal to the cap itself, far above anything most players will bet, and the separate maximum-bet figure (for example 25,000,000 USDT) becomes the binding one. The help-centre article lists the limits without saying how a bet above them is handled, so check the bet panel when staking large at long odds.
How one roll is derived from your seeds
Dice uses the least machinery of any Stake Original. The documentation lists it among the "games with only 1 incremental number", meaning a single hash is enough:
- Stake computes HMAC-SHA256 with the server seed as the key and
clientSeed:nonce:0as the message. - The first four bytes of the result are combined into a number between 0 and 1: byte 1 over 256, plus byte 2 over 65,536, plus byte 3 over 16,777,216, plus byte 4 over 4,294,967,296.
- That number is multiplied by 10,001 and divided by 100, giving the roll with two decimals.
- The roll is compared with your target. Over wins if the roll is above it, under wins if it is below.
float = b1/256 + b2/256² + b3/256³ + b4/256⁴
roll = floor(float × 10001) / 100 → 00.00 to 100.00
The code sample on Stake's "Game events" page reads const roll = (float * 10001) / 100; with no rounding step, although a roll is always shown to two decimals; our checker rounds down, which is the reading that yields exactly 10,001 outcomes. Four bytes give 4,294,967,296 equally likely values spread across 10,001 rolls, so each roll's chance differs from the next by less than one part in 400,000. For practical purposes every hundredth is equally likely.
The server seed behind the hash is "a random 64-character hex string" that Stake shows only as a SHA-256 hash until you rotate the seed pair. The client seed is yours to change, and the documentation recommends doing so: "we highly recommend that you choose your own". The nonce counts your bets on the current pair, starting from zero after each rotation. To check a roll:
- Commitment: hash the revealed server seed with SHA-256 and compare it with the hash shown before the bets.
- Roll: our Stake seed checker and the Dice verifier take the server seed, client seed and nonce and print the roll with each intermediate step. Everything runs in your browser.
- Many rolls at once: the nonce is the only thing that changes between bets on one seed pair, so a whole session can be replayed by counting it up from the first bet.
Stake also lists a second dice game, Primedice, in the same documentation. It "uses the same dice roll mechanism as the regular Dice game" and adds bets on a roll between two numbers, outside a range, or inside one of two ranges. The roll formula and the 99% figure are the same; only the shapes of the winning zones differ. Background on the method: what provably fair gambling proves and the general verifier.
Stake Dice against Shuffle, Roobet and Gamdom
| Stake | Shuffle | Roobet | Gamdom | |
|---|---|---|---|---|
| Roll range | 0.00 to 100.00 (10,001 values) | 0.00 to 100.00 (10,001 values) | 0.00 to 99.99 (10,000 values) | Target slider 1.99 to 97.99 |
| Stated return | 99% RTP | 1.00% house edge | 99% RTP, 1% house edge | 99% once the first $5,000 of daily Originals wagering is used |
| Top multiplier | 9,900x | 9,900x at a 0.01% chance | Not stated on the game page | Set by the slider range |
| Limits | Per coin; profit cap e.g. 2,500,000 USDT | $20,000 bet, $250,000 win | Shown in the bet panel | $200,000 maximum bet per the help centre |
| Roll derivation | HMAC-SHA256, 4 bytes | Same formula as Stake | HMAC-SHA512 twice, 64 bits, cut to 4 decimals | HMAC-SHA512, own verifier |
Shuffle: game page and help centre, 12 Nov 2026. Roobet: game page 1 Nov 2026, verifier 12 Nov 2026. Gamdom: game page 1 Nov 2026, help centre 12 Nov 2026.
On price the four are level: each states a 1% edge. The differences are in the ceiling on a single win and in how the roll is computed. Shuffle documents the same formula as Stake, down to the 10,001 outcomes, so one checker covers both. Roobet's current verifier builds the roll from a 64-bit slice of an HMAC-SHA512 output and has no 100.00.
Sources: Stake fairness documentation, implementation, conversions and game events (1 Nov 2026); Stake Dice game page and house-edge guide (1 Nov 2026); Stake help centre, crypto betting limits (12 Nov 2026). Tables are our arithmetic from the stated 1% edge and the published caps.
The Dice calculator converts any target into win chance, payout and expected loss. Licence, KYC and withdrawal terms are in the Stake review.