What the D'Alembert system is
The D'Alembert system is a stake-sizing rule for even-money bets, published by the French mathematician Jean le Rond d'Alembert in the 1700s. You choose a base unit first. Then exactly three things happen: after a loss you stake one unit more, after a win you stake one unit less and never below the base, and when the stake has walked back down to the base the cycle is finished and the next round starts again from one unit.
That is the whole rule. It is frequently misquoted as βdouble after a loss, halve after a winβ, which is the Martingale system — a different strategy with a different risk profile, compared honestly in the table below. D'Alembert does not double anything. Its stake moves in single units, one step per outcome.
The clearest way to see the mechanics is one complete cycle. Take a base unit of 1 and start at a stake of 1:
| Round | Stake | Result | Change | Running total | Next stake |
|---|---|---|---|---|---|
| 1 | 1 unit | Loss | −1 | −1 | 2 units |
| 2 | 2 units | Loss | −2 | −3 | 3 units |
| 3 | 3 units | Win | +3 | 0 | 2 units |
| 4 | 2 units | Win | +2 | +2 | 1 unit — cycle closed |
Eight units went in and two came out, because a closed cycle nets exactly one base unit for every win in it. The deepest point in the middle was −3. That bounded, tiny gain is the whole promise of the system — and it is what you are managing against an open-ended loss.
The other half of the rule is what a losing run costs. Because the stake rises by one unit each time, a run of L consecutive losses from the base stake asks for
1 + 2 + 3 … + L = L(L + 1) / 2
Ten losses in a row costs 55 units. Twenty costs 210. Thirty costs 465. The gain per completed cycle never grows; the requirement to survive a run grows without limit. That asymmetry is the entire reason this strategy does not work, and it is exactly what the equity curve on this page draws.
How to use this simulator
The calculator is a ledger, not a game. There is no money here and nothing is ever wagered — every button below records an outcome you tell it about and works out what the D'Alembert rule would have done next.
- Set the base unit. Use the Base Unit stepper in the Round Actions card. The next-stake display then reads in units as well as currency, so
4 unit(s) @ 1/unitmeans the stake has climbed four steps above where the cycle started. - Set the bankroll. Click the pot figure, or use the −10, −1, +1, +10 and +Unit buttons. This is the number that decides when a run has cost more than you have, so set it to whatever you want to watch degrade.
- Record rounds. Press WIN or LOSE by hand (or the ↑ and ↓ keys), and Reset Cycle to Base Unit with →. To watch a long run instead, open Auto-Simulate and choose a round count, a win percentage and a speed; it runs up to 50,000 rounds and stops the moment the pot can no longer cover the next bet.
- Read the ledger. Every round writes one row — round number, outcome, stake, and the pot at that moment. The table keeps the latest 200 entries and the full session exports to CSV, so you can inspect the exact arithmetic rather than trusting the curve.
- Read the curve and the stats. The equity chart plots the pot after every round; hover a point for the round number and balance, or expand it to zoom. Above it, Total Rounds, Win Ratio, Max Bet Reached and Net Profit / Loss summarise the run so far.
Everything stays in your browser. The pot, the settings and the ledger are held in this device's local storage, and Reset All erases them and starts a new session with a fresh bankroll.
D'Alembert vs Martingale
Both are negative progressions: both raise the stake after a loss. They differ in how far, and that single difference decides how each one fails. Neither is a way to beat a negative expectation game, and this table is not a ranking.
| D'Alembert | Martingale | |
|---|---|---|
| Stake after a loss | Rises by one base unit: 1, 2, 3, 4… | Doubles: 1, 2, 4, 8… |
| Stake after a win | Falls by one base unit, never below the base stake | Returns to the base stake |
| How the stake grows | Linearly — one extra unit per loss | Exponentially — it doubles per loss |
| Bankroll for 10 losses in a row | 55 units | 1,023 units |
| Bankroll for 20 losses in a row | 210 units | 1,048,575 units |
| With a 100-unit bankroll, the run that ends it | The 14th consecutive loss — 105 units would be needed | The 7th consecutive loss — 127 units would be needed |
| How the damage arrives | A steady bleed across every round of the run | Almost all of it at once, on the final bet |
| How a recovery works | One win does not repair the level above it; the stake has to walk back down, one step per win | A single win at the doubled stake repairs the entire run |
| Effect on the house edge | None | None |
| Effect on expected return | Negative, by the size of the edge | Negative, by the size of the edge |
So Martingale fails sooner and harder, and D'Alembert lasts roughly twice as long before a 100-unit bankroll gives out. The honest reading of that is not that D'Alembert is the safer system. It is that D'Alembert loses in instalments — which is considerably easier to mistake for progress, because the ledger keeps ticking over round after round instead of hitting a wall in one dramatic bet. Both finish in the same place, at the same rate of loss measured against turnover. If you want to watch the difference for yourself, set both base units to 1 and a pot of 100, then run the auto-simulation at 48.65 percent and watch how long the two curves last.
The reverse D'Alembert is also in the calculator. Under Settings › Progression Model you can flip the rule to run the other way — up one unit on a win, down one unit on a loss. It is a positive progression, which behaves quite differently under a losing run, and switching between the two while keeping the bankroll and win rate fixed is a useful way to see that the stake rule changes the shape of the equity curve without changing its drift.
What the equity curve actually shows
The chart on this page plots the pot after every round, starting from the balance the session opened with. It is worth reading carefully, because it is the one part of a simulation that says something a summary statistic cannot.
The win rate is an average, and the curve is a path. Two sessions can produce the same 49 percent win ratio and end at wildly different balances, because in one of them the losses arrived early and the wins arrived late. Order is invisible in an average. It is the only thing that matters once the average has already told you the losing side wins over time.
This is why a curve that looks positive can still end at zero. A session can sit above its starting line for two hundred rounds, bank a real-looking gain, and then give all of it back and more. Nothing in the curve up to that point is protection: a long stretch of gains does not make the next losing run any less likely, and one run of fourteen losses needs 105 units whether you have had a good week or a bad one. The curve above water is a description of the past, not a buffer against the future.
Watch where the curve returns to, not just where it ends. The useful signal is how long the line spends away from its starting level. A curve that keeps crossing back over the base line is one where the losing runs stay short. A curve that leaves and does not come back is one where the runs have grown past what the bankroll can carry — and no amount of winning in between will change that.
Two things about this particular chart are worth knowing before you draw conclusions from it. The curve is anchored on the pot the session opened with, so it is an honest record of the run rather than a chart rescaled to each cycle. And in an auto-simulation the last plotted point is simply where funding stopped — the auto-run halts the instant the pot can no longer cover the next bet — not a final verdict on anything.
One setting that is easy to misread. By default Compound Starting Baseline is on, so every completed cycle re-anchors the target to the new balance. The cycle figure beside the pot then returns to +0 after each cycle even as the session figure keeps falling. That is the setting working as designed, not a sign that the system is holding steady — the session total is the number that carries over.
What this simulator cannot tell you
A simulator is only honest if it is clear about its own reach. Here is where this one stops.
- It cannot predict anything. The auto-simulator draws each round from the win percentage you type, using your browser's own random number generator. It is a model of arithmetic, not a model of a wheel. Whatever curve it produces, the next real outcome is completely independent of it.
- It does not include the house edge unless you type it in. The ledger credits a win with exactly the stake and debits a loss with exactly the stake. At a 50 percent win rate that is a fair game, so the expected drift is precisely zero and the curve is nothing but variance. Real even-money propositions are not 50 percent: red or black on a single-zero wheel is 18/37, about 48.65 percent, so the expected loss is 2.7 percent of everything staked. Set the win rate to a realistic number and a slow downward drift appears — caused entirely by the edge, and entirely independent of the stake rule.
- A progression cannot change the edge. The edge is a percentage of turnover. Raising the stake raises how much you lose per round, not how much you are charged. It changes the order in which money leaves, never the amount.
- Simulation is for understanding a sequence, not predicting one. Run it to watch what a fourteen-loss run does to a 100-unit pot, or how far a stake climbs before it is capped. It cannot tell you when the next such run arrives, because nothing can.
- It models a ledger, not a person. There is no real bankroll here, no table limit, no session length, no fatigue and no tilt. In practice the stake discipline a system requires is the hardest part to hold, and no spreadsheet measures that.
This calculator has no advertising, no account and no server — it runs entirely in your browser, and nothing you enter is transmitted or stored anywhere. Hosting a free simulation like this still costs money every month, and if it made the arithmetic of a betting system clearer, you are welcome to cover the running costs on the support page.
Frequently asked questions
What is the D'Alembert system?
It is a stake-sizing rule for even-money bets, described by the mathematician Jean le Rond d'Alembert in the 1700s. Choose a base unit, add one unit after each loss, subtract one unit after each win without going below the base, and start again from one unit once the stake has walked back down.
It is often confused with the Martingale, which doubles after a loss. D'Alembert adds a single unit instead — see the comparison above for what that difference does to the bankroll required.
Does the D'Alembert system beat the house edge?
No. The house edge is a fixed percentage of every amount staked, so it applies just as much to a 40-unit bet as to a 1-unit bet. Changing the size of a stake changes how much you lose when you lose and how quickly — it never changes the probability of winning or the size of the edge.
On a single-zero roulette wheel red or black pays even money on 18 of 37 pockets, about 48.65 percent, so the expected loss is 2.7 percent of turnover. A negative progression rearranges your wins and losses in time. Rearranging them in time is not removing them.
How is D'Alembert different from the Martingale system?
Martingale doubles after every loss, so the bankroll needed to survive a run grows exponentially — 1,023 units to absorb ten consecutive losses, and more than a million for twenty. D'Alembert adds one base unit per loss, so the same runs cost 55 and 210 units.
Martingale therefore fails sooner and more sharply. D'Alembert lasts longer and bleeds steadily instead. Neither recovers the edge, and neither turns a game with negative expectation into a positive one.
How much bankroll does D'Alembert need to survive a losing run?
A run of L consecutive losses starting from the base stake costs 1 + 2 + … + L, which is L(L+1)/2 units: 55 units for ten losses, 210 for twenty, 465 for thirty. That formula is why the simulator asks for a bankroll — set a pot, set a base unit, and watch exactly how far the stake climbs before the pot can no longer cover the next bet.
What it cannot do is make the number reachable, because runs of that length keep arriving at the same rate whatever the stake rule does.
What does the equity curve show that the win rate does not?
Order. The win rate is one averaged number and cannot show how long a session spent below its starting balance, which is usually the part that matters. Two sessions with identical win rates can end at completely different balances depending on whether the losses arrived first.
It is also why a curve that looks positive can still end at zero: a long stretch of gains is not insurance against the next losing run. On this calculator the final plotted point is where funding stopped, not a verdict on the strategy.
Is this simulator free, and does anything leave my browser?
Yes to both. There is no account, no sign-up and no advertising, and the calculator runs entirely in your browser with no server involved, so a session built here is never transmitted or stored anywhere else.
Your bankroll, settings and ledger live in this browser's local storage on this device, and the Reset All control erases them at any time.