D'Alembert roulette strategy
D'Alembert is the gentlest of the classic loss-chasing systems: the stake moves by exactly one unit. One unit up after a loss, one unit down after a win.
01How D'Alembert works
Named after the 18th-century French mathematician Jean le Rond d'Alembert, who believed a win becomes more likely after a loss. Start with one unit; add one after a loss, take one away after a win, never going below one.
| Spin | Result | Stake | Balance |
|---|---|---|---|
| 1 | loss | 1 | −1 |
| 2 | loss | 2 | −3 |
| 3 | win | 3 | 0 |
| 4 | loss | 2 | −2 |
| 5 | win | 3 | +1 |
| 6 | win | 2 | +3 |
Every win comes one unit higher than the loss before it, so a loss followed by a win leaves you one unit ahead. If wins and losses came in equal numbers, you would be up roughly one unit per win. That is the whole idea of the system.
02The numbers
Results do not balance out: the wheel has no memory, and every spin has the same 18 in 37 chance of winning. The zero makes losses slightly more likely than wins (51.4% against 48.6%), so losses tend to outnumber wins and the stake creeps up.
| Losses in a row | Next stake | Lost so far | Fibonacci | Martingale |
|---|---|---|---|---|
| 4 | 5 | 10 | 5 | 16 |
| 6 | 7 | 21 | 13 | 64 |
| 8 | 9 | 36 | 34 | 256 |
| 10 | 11 | 55 | 89 | 1,024 |
Stakes grow in a straight line, not exponentially, so one bad streak is far less dangerous than with Martingale. The risk is a long, uneven stretch where losses slightly outnumber wins: the stake and the deficit rise step by step, and a single win takes only one unit off the stake.
03Where it stops
D'Alembert almost never fails through one spectacular streak. It fails through attrition: a session drifts down, the stake climbs to ten or fifteen units, and the bank is gone in a series of small setbacks. The simulator's series stop-loss can cut a series earlier, but that only reshapes the loss distribution.
04What 200,000 sessions show
Classic rules, European wheel, bank 100 units, base stake 1 unit, target +20 units.
| System | Reached +20 | Average spins |
|---|---|---|
| D'Alembert (classic) | about 75% | about 51 |
| Fibonacci (classic) | about 74% | about 93 |
| Martingale | about 69% | about 34 |
| Flat betting | about 34% | about 2,200 |
Failed sessions lost 92 units on average, nearly the whole bank. Over many sessions the average result still matches the house edge: 2.70% of everything bet on a European wheel, 5.26% on an American wheel.
05Variants in the simulator
D'Alembert (classic) adds a unit after every loss, as above. D'Alembert (paced) only adds a unit when one win at the current stake can no longer cover the losses of the series, which keeps stakes much lower on bets with high payouts. Reverse D'Alembert raises the stake after wins instead, with a profit lock.
06What it means in practice
D'Alembert is the friendliest progression to play: no doubling, no drama, long sessions. That is also why it feels like it works. It doesn't: it spreads the same house edge over more spins.
07Try it yourself
In the simulator, click Even money (1:1), pick D'Alembert (classic) and spin. Monte Carlo runs up to 100,000 complete sessions, and a seed lets you compare it with Martingale and Fibonacci on exactly the same spins.