Dice guide
Sakura Dice: the doubling system, worked through
Sakura is the page label and refers to nothing in the game. This guide tests one betting system on number dice: the martingale, where the stake doubles after every loss and resets after a win.
How the martingale is supposed to work
Use a roll-under target of 49.50, which pays exactly 2.00x at a 1% edge. Bet 1 unit. If it loses, bet 2. If that loses, bet 4, and so on. The first win recovers all earlier losses plus 1 unit. The pitch is that a win must come eventually, so every sequence ends with +1.
The flaw is the word eventually. A bankroll is finite, and many services also cap the stake. After L straight losses the player has staked 1 + 2 + ... + 2 to the power L - 1 = 2 to the power L, minus 1 units, and cannot make the next doubling.
| Doublings allowed | Bankroll needed (units) | P(lose every step) | As odds |
|---|---|---|---|
| 4 | 15 | 6.5038% | 1 in 15.4 |
| 6 | 63 | 1.6586% | 1 in 60.3 |
| 8 | 255 | 0.4230% | 1 in 236.4 |
| 11 | 2,047 | 0.0545% | 1 in 1,835.7 |
What one sequence is worth
With 8 doublings allowed, the bankroll is 255 units. A sequence fails with chance 0.423%, losing all 255. Otherwise it wins 1 unit. The expected result per sequence is -0.0829 units. It is negative, as it must be, since every single bet inside the sequence has a negative EV.
With 11 doublings the bankroll grows to 2,047 units. Failure becomes rarer, 0.054%, but now it costs 2,047 units. The expected result per sequence is -0.1157. A larger bankroll only makes the failures rarer and bigger.
How often the failure comes
A run of 8 losses at 50.5% per roll shows up in 65.12% of 500-roll sessions, since a martingale player makes many short sequences in a row. The system turns a steady stream of small wins into a near certainty of an eventual large loss over a long enough session.
Other systems, same result
- Reverse martingale (doubling after wins): the same EV, with small frequent losses and rare large wins.
- Fixed-step systems that add a unit after a loss: slower growth in stake, same negative EV per unit.
- Changing targets between bets: each bet still has EV -1.00% per unit.
The total expected loss of any system equals the edge times the total amount wagered. Doubling raises the total wagered, so it raises the expected loss. No ordering of negative-EV bets adds up to a positive one.