Stake dice strategy: how the game pays and why no system beats the edge

Stake dice is about as simple as a casino game gets: pick a number, choose whether the roll must land over or under it, and the game draws a roll from 0.00 to 100.00. Because it is so simple, it attracts more "strategy" than almost any other game, from Martingale doubling to auto-bet scripts. This guide explains exactly how Stake Dice pays, shows with numbers why no betting system changes the 1% edge, and walks you through checking a real roll in our verifier.

How does Stake Dice work?

You set three things: your bet, a target number, and a direction (roll over or roll under). The game then draws a roll. The roll can be any of 10,001 values from 0.00 to 100.00 in steps of 0.01, all equally likely.

Your win chance is the share of those rolls that beat your target. Roll over 75.00 wins on about 25% of rolls; roll under 75.00 wins on about 75%. Moving the slider or typing a chance changes the target and the payout together.

The payout follows one rule:

Multiplier = 99 / win chance (in percent)

The last column is the expected return: whatever chance you choose, a $1 bet returns $0.99 on average. A fair game would use 100 instead of 99. That missing 1 is the house edge, and it sits in the payout, not in the roll. Stake's help centre does not publish an edge table, so the 1% here is what the multiplier formula implies.

Roll over or roll under: does it matter?

No. Roll over 50.50 and roll under 49.50 both win 49.5% of the time and both pay 2x. The roll is uniform, so which side of the range you bet on changes nothing except which rolls you cheer for.

Is there a Stake dice strategy that works?

Not one that beats the edge. Each roll comes from a fresh hash of your seeds and a new nonce, so the previous roll tells you nothing about the next. And every bet, whatever its size or chance, returns 99% of the stake on average. Add up any sequence of bets and the expected result is the same:

Expected loss = 1% x total amount wagered

A system can change the shape of your results: many small wins and a rare big loss, or many small losses and a rare big win. It cannot change that line. Wagering $5,000 in a session costs about $50 on average whether you placed it as 5,000 flat $1 bets or as a Martingale ladder.

The Martingale, worked through

The Martingale is the most common dice system. You bet at 49.5% for 2x, start at $1, double after each loss and go back to $1 after a win. Every win recovers all the losses in the run plus $1. It feels safe because most runs end in a win.

The catch is the losing streak. The chance of losing one bet is 50.5%, so the chance of losing k bets in a row is 0.505 to the power of k. Covering k losses in a row from a $1 start needs a balance of 2^k minus 1 dollars. Here is what happens if you try to win $100, one dollar at a time (100 completed runs):

We computed the table in Python. The expected result per run is 1 minus 1.01 to the power of k dollars, which is always negative.

Read across the rows and the trade is clear. A bigger balance makes a bust rarer, but each bust costs more, and the expected loss grows rather than shrinks, because a deeper ladder means more money wagered. With $32,767 at risk to win $100, you will usually win the $100. About 1 time in 250 you lose $32,767, and that rare loss is bigger than all the small wins put together.

Two more limits make it worse in practice: any maximum bet stops the ladder early, and a bad streak can arrive on the first run as easily as the hundredth.

Other systems and myths

• Reverse Martingale (doubling after wins). Same edge, opposite shape: frequent small losses and rare big wins.
• D'Alembert and Labouchere. Slower ladders than Martingale, so the busts are smaller and the wins smaller too. The 1% stays.
• High chance "grinding". Betting 98% for 1.0102x wins most rolls, but the edge on total wagered is still 1%, and you wager far more to make the same profit.
• Changing your client seed after losses. Your client seed makes the results unpredictable to Stake; it does not let you pick better rolls. A new seed gives a new, equally random sequence.
• Auto-bet scripts. Stake's auto-bet can increase your bet on a loss or reset on a win. It runs a system faster; it does not change its maths.

What you can control is how fast you lose the edge and how wild the swings are. Low win chances swing more; fast auto-bet wagers more per hour. Neither creates profit.

How to verify a Stake dice roll

Stake Dice is a provably fair Stake Original. Before you bet, Stake shows the SHA-256 hash of a secret server seed. Each roll is built from HMAC-SHA256 with the server seed as the key and client seed:nonce:round as the message. The first 4 bytes become a random float between 0 and 1, and the roll is that float times 10,001, rounded down and divided by 100.

Here is a worked example with made-up seeds:

1. SHA-256 of the server seed gives b780b898...dd209a2, which matches the hash, so the seed was fixed before the bet.
2. HMAC-SHA256 over lantern-harbour-9:0:0 starts b9be8a43. Those bytes are 185, 190, 138 and 67.
3. 185/256 + 190/256^2 + 138/256^3 + 67/256^4 = 0.725564.
4. 0.725564 x 10,001 = 7,256.36. Round down to 7,256, divide by 100: the roll is 72.56.

A bet on roll over 50.50 wins. Nonces 1, 2 and 3 on the same seeds give 20.11, 20.09 and 62.99. You can open this exact check in our verifier.

For your own bets:

1. In Stake's fairness settings, copy the hashed server seed and set your own client seed if you want.
2. Play. Each bet in your history shows its nonce.
3. Rotate your seed pair to reveal the old server seed.
4. Open our Stake verifier, paste the revealed seed, the hash, your client seed and the nonce, and choose Dice.
5. Compare the roll with your bet history. If they match, the roll was fixed before you bet.

A mismatch usually means a seed from a different pair or a nonce out by one. Our step-by-step provably fair dice walkthrough covers the byte maths in more detail, and Stake Originals explained shows how the same engine drives Stake's other games.

Play it for what it is

Verification proves the roll was fair. It does not make the game profitable: at a 1% edge, every $1,000 you wager costs about $10 on average, and dice lets you wager $1,000 in minutes. Set a loss limit and a time limit in Stake's responsible gambling settings before you open the game, and see our responsible gambling page if betting stops being fun.

Sources

• Stake help centre: provable fairness for Stake Original games
• Dice scheme as recorded in our Stake audit and implemented in our verifier
• RFC 2104: HMAC
• NIST FIPS 180-4: Secure Hash Standard
• Martingale figures computed by us in Python from a 49.5% win chance at 2x