Game Strategy · Provably Fair · Jun 2026

Dragon Tower strategy & odds

Dragon Tower on Stake is five ladders priced by one rule: 0.98 divided by the chance of getting that far. Below: the full multiplier table with the chance of each level, where rounding moves the return on Easy and Medium, how the eggs are placed from the seeds, and how Shuffle, Roobet and BC.Game price the same climb.

Not sure which to choose? See what we recommend for each type of player.

What Stake publishes about Dragon Tower

Dragon Tower is a nine-level climb. Each level is a row of tiles, some hiding an egg; pick an egg and you move up and the multiplier grows, pick an empty tile and the round is lost. You can cash out after any level. Stake's game page states: "This game has a theoretical return to player percentage of 98%, meaning that the Dragon Tower house edge is just 2%" (1 Nov 2026).

The fairness documentation defines the five difficulties by how many eggs a row holds and how many tiles it has: "A Dragon Tower game is generated with 9 separate game events, in the form of the levels up the tower", with Easy at 3 eggs in 4 tiles, Medium 2 in 3, Hard 1 in 2, Expert 1 in 3 and Master 1 in 4. Every tile in a row is equally likely to hold an egg, so the chance of passing a level is simply eggs divided by tiles.

The multiplier table, level by level

The fair multiplier for clearing L levels is one over the chance of doing so. The game's table on 1 Nov 2026 was 0.98 times that, rounded to the nearest cent.

LevelEasy (3 of 4)Medium (2 of 3)Hard (1 of 2)Expert (1 of 3)Master (1 of 4)
11.31x1.47x1.96x2.94x3.92x
21.74x2.21x3.92x8.82x15.68x
32.32x3.31x7.84x26.46x62.72x
43.1x4.96x15.68x79.38x250.88x
54.13x7.44x31.36x238.14x1,003.52x
65.51x11.16x62.72x714.42x4,014.08x
77.34x16.74x125.44x2,143.26x16,056.32x
89.79x25.11x250.88x6,429.78x64,225.28x
913.05x37.67x501.76x19,289.34x256,901.12x

Stake Dragon Tower multiplier table (1 Nov 2026).

Chance of clearingEasyMediumHardExpertMaster
175.00%66.67%50.00%33.33%25.00%
256.25%44.44%25.00%11.11%6.25%
342.19%29.63%12.50%3.70%1.56%
431.64%19.75%6.25%1.23%1 in 256
523.73%13.17%3.12%1 in 2431 in 1,024
617.80%8.78%1.56%1 in 7291 in 4,096
713.35%5.85%1 in 1281 in 2,1871 in 16,384
810.01%3.90%1 in 2561 in 6,5611 in 65,536
97.51%2.60%1 in 5121 in 19,6831 in 262,144

Arithmetic: (eggs / tiles) to the power of the level.

On Hard, Expert and Master the multipliers are exact to the cent, so the return is 98.00% at every level. On Easy and Medium the rounding leaves small differences:

Return if you cash out after level123456789
Easy98.25%97.88%97.88%98.09%98.01%98.07%97.98%98.01%97.99%
Medium98.00%98.22%98.07%97.97%97.97%97.97%97.97%97.97%97.99%

Arithmetic: table multiplier times the chance of clearing that many levels.

Cashing out after the first Easy level returns 98.25%, because 1.3067 was rounded up to 1.31. After the second or third it returns 97.88%. The gaps are fractions of a cent per dollar and too small to build a plan on, but they are the reason "98%" is not exact on the two gentlest settings.

The top of each ladder: 13.05x on Easy, reached in 7.5% of rounds; 37.67x on Medium (2.6%); 501.76x on Hard (1 in 512); 19,289.34x on Expert (1 in 19,683); and 256,901.12x on Master, 1 round in 262,144.

How the eggs are placed

  1. Stake shows the SHA-256 hash of the server seed before the round.
  2. Random numbers come from HMAC-SHA256 keyed with the server seed over clientSeed:nonce:cursor, four bytes per number.
  3. For each of the nine rows, egg positions are drawn without repeats. The documentation's example: "A level on difficulty easy would be represented like this: [0, 1, 3] - eggs would be present at tile 1 and 2 and 4."
  4. "The fisher-yates shuffle implementation is utilised to prevent duplicate eggs on a row."
  5. Rotating the seed pair reveals the server seed, which must hash to the value from step 1.

All nine rows are fixed by the seeds and the nonce before the first pick. Choosing the left tile or the right tile selects among positions that are already decided; it cannot move an egg. Stake's calculation page lists Dragon Tower with a difficulty selector and prints every row. Our provably fair verifier has no tower mode, so it covers step 5, the seed commitment, and the rows are checked on Stake's page.

Tower games at other casinos

The same idea runs under other names, and the price is not the same everywhere.

Stake Dragon TowerShuffle TowerRoobet TowersBC.Game Tower LegendRainbet Tower
Stated price98% RTPNo figure; multiplier chart published"approximately 96.5%"98 in the game data; 96% in the help article"Max RTP 96.00%"
Levels9989, or 6 on the two hardest settings8
Safe tiles per row3 of 4, 2 of 3, 1 of 2, 1 of 3, 1 of 4Same fiveSame fiveSame fiveSame five
First level, 1 safe tile of 21.96x1.96x1.93x1.96x at 98%Shown in the game
Top multiplier256,901.12x256,901.12x63,242.24x4014×Shown in the game
Stated limits$0.0001 minimum; profit cap per coin$25,000 bet, $700,000 winBet up to 100,000; payout up to 3,000,000Shown in the bet panelShown in the game

Shuffle help centre, Tower multiplier chart, game events and limits (13 Nov 2026); Roobet Towers game page (1 Nov 2026), configuration and verifier (13 Nov 2026); BC.Game help centre, game data and Tower verifier (13 Nov 2026); Rainbet game page (1 Nov 2026).

Shuffle publishes its chart in the help centre, and it is Stake's table with one cell different: Medium, row 8 reads 25.12 where Stake has 25.11. The unrounded value is 25.1163. Shuffle notes that the multipliers "have been visually truncated to 2 decimal points for readability".

Roobet states "Roobet Towers has an RTP of approximately 96.5%, regardless of the difficulty setting selected", and its public configuration carries an edge of 3.5. The displayed ladders bear that out: 1.29x, 1.72x, 2.29x on Easy against Stake's 1.31x, 1.74x, 2.32x, and 1.93x per level on Hard against 1.96x. There are eight levels, so the hardest setting tops out at 63,242.24x. Roobet's help centre describes Easy as "1 bomb in 5 tiles"; the game page and the verifier both give four tiles.

BC.Game runs nine levels on Easy, Medium and Hard and six on Extreme (1 safe of 3) and Nightmare (1 safe of 4). Its help article says "The game's Return to Player (RTP) rate is 96%"; its game data lists 98 and a maximum win of 4014×, which is 0.98 × 4 to the sixth power, the Nightmare top at a 98% price. The two documents disagree, and the maximum-win figure sides with 98.

Other casinos reviewed here list a tower game with their own figures (all as of 11 Nov 2026): Betpanda's default table gives Tower 99%, BetFury prints 98.2% on its Tower tile, BitStarz 97.5% with a $150,000 maximum win, Wild.io "97.5%+" with a 100,000x top multiplier, and 500 Casino states a 5% edge on Towers with a $200,000 win limit per round. On stated price the range across the sites on this page runs from 1% to 5%.

Gamdom has published verifier code for a Towers game with the same five layouts, nine rows and a 2% edge. Towers is not among the games listed on its fairness page (13 Nov 2026).

Difficulty and cash-out: what they change

A plan in Dragon Tower is a difficulty and a level to stop at. Each plan is a single bet with a known chance and payout:

PlanRound winsPaysStandard deviation per roundReturn
Easy, stop after level 175.00%1.31x0.5798.25%
Easy, stop after level 431.64%3.1x1.4498.09%
Medium, stop after level 329.63%3.31x1.5198.07%
Hard, stop after level 150.00%1.96x0.9898.00%
Hard, stop after level 312.50%7.84x2.5998.00%
Expert, stop after level 211.11%8.82x2.7798.00%
Master, stop after level 26.25%15.68x3.8098.00%

Arithmetic from Stake's table. Standard deviation is in units of the stake.

Hard to level 3 and Expert to level 2 pay nearly the same, 7.84x and 8.82x, at nearly the same chance. Different ladders reach the same kind of bet by different routes, and none of them is priced better than another. Stopping early wins often for little; climbing wins rarely for a lot.

Nothing about the order or size of bets shifts the 98%. Rounds are generated independently, so the tile that was safe last time carries no information, and doubling the stake after a failed climb changes the distribution of session results without improving their average. For the same climb at a lower price, Stake's Mines carries a 99.00% badge; the Martingale simulator shows what progressions do.

Limits on Stake

Stake's help centre gives a minimum of "$0.0001 USD" and a maximum profit per coin, 2,500,000 USDT for example. Against the upper rungs of the harder ladders that profit cap is the limit that matters. The table shows how rare each top prize is and the largest stake for which it fits inside the USDT cap.

DifficultyLevel 9 paysChanceRounds for an even chance of oneLargest USDT stake paid in full
Easy13.05x7.51%9207,468.88
Medium37.67x2.60%2768,175.62
Hard501.76x1 in 5123554,992.41
Expert19,289.34x1 in 19,68313,643129.61
Master256,901.12x1 in 262,144181,7059.73

Arithmetic from Stake's table and its 2,500,000 USDT maximum profit (help centre, as of 13 Nov 2026).

A Master climb to the top is a 1-in-262,144 event, and at any stake above about 9.73 USDT its prize is larger than the cap. How a result above the cap is settled is something to confirm in the bet panel before staking more.

Related: Pump · Mines at Stake · Mines at Shuffle · Mines at Roobet · Casino Originals compared · Stake review

Sources: Stake Dragon Tower game page and multiplier table (1 Nov 2026); Stake fairness documentation, game events and implementation (1 Nov 2026); Stake help centre, Crypto: Casino Betting Limits (article dated 24 August 2026, as of 13 Nov 2026); Shuffle Tower multiplier chart, game events and limits (13 Nov 2026); Roobet Towers game page (1 Nov 2026), game configuration, Towers verifier and Roobet House Games (13 Nov 2026); BC.Game help centre, How to play Tower Legend, game data and Tower verifier (13 Nov 2026); Gamdom Towers sample code and fairness page (13 Nov 2026); Rainbet Tower game page (1 Nov 2026). Chances and returns are our arithmetic.

Frequently asked questions

Stake states a theoretical return of 98%, a 2% house edge. The multiplier table is 0.98 divided by the chance of clearing each level, rounded to the cent: exactly 98.00% on Hard, Expert and Master, and between 97.88% and 98.25% on Easy and Medium depending on the level.
The chance of passing one level is 75% on Easy (3 eggs in 4 tiles), 66.67% on Medium (2 in 3), 50% on Hard, 33.33% on Expert and 25% on Master. Clearing all nine levels has a chance of 7.5%, 2.6%, 1 in 512, 1 in 19,683 and 1 in 262,144.
256,901.12x for nine levels on Master. The other tops are 13.05x on Easy, 37.67x on Medium, 501.76x on Hard and 19,289.34x on Expert.
No. Every tile in a row has the same chance of holding an egg, and the positions for all nine rows are fixed by the seeds before the first pick.
The layouts are the same and the prices are not. Shuffle's chart matches Stake's 98% table. Roobet states about 96.5% with eight levels. BC.Game's help article says 96% and its game data 98.
No. Difficulty and cash-out level decide how often you win and how much. The expected return stays at about 98%, and betting systems do not change it.
AM
Alex Mercer
Alex covers crypto casino game strategy and provably fair mechanics for the ProvenlyFair.com Editorial Team, with a focus on the mathematics behind Stake Originals.
Stake — Dragon Tower, provably fair
Play Dragon Tower