# D'Alembert Betting System Explained: Mathematics, Risks and Why It Fails
The D'Alembert betting system is a negative progression in which the stake rises by one unit after a loss and falls by one unit after a win. It is less aggressive than the Martingale and easier to understand than many sequence-based systems, which helps explain its enduring appeal. The central claim, however, does not survive mathematical scrutiny.
D'Alembert changes the size of the next bet. It does not improve the selection, remove the bookmaker's margin or make a win more likely after a loss. Its steady progression can produce many apparently tidy sessions, while a longer run of adverse results creates a deficit that one later win cannot repair.
This guide explains the method in full, works through realistic sequences and compares its exposure with Martingale staking, the Fibonacci betting system and flat staking. All figures are worked illustrations, not a recommendation to chase losses.
The essentials
Choose a base unit, such as £5. Begin with one unit. After every loss, add one unit to the next stake. After every win, subtract one unit, but do not go below the original base stake.
| Previous result | Next stake when one unit is £5 | Rule |
|---|---|---|
| Starting bet | £5 | Begin at one unit |
| Loss at £5 | £10 | Add one unit |
| Loss at £10 | £15 | Add one unit |
| Win at £15 | £10 | Subtract one unit |
| Win at £10 | £5 | Subtract one unit, stopping at the base |
The system is commonly illustrated at decimal odds of 2.00. At that price, a winning bet produces profit equal to the stake. Prices below 2.00 make recovery slower because the profit is smaller than the stake; prices above 2.00 usually imply a lower chance of winning.
Key point: D'Alembert is a staking rule, not a way to find value. If the underlying bets have negative expected value, increasing stakes after losses increases the amount exposed to that disadvantage.
Where the idea comes from
The name is associated with Jean le Rond d'Alembert, the eighteenth-century French mathematician. Modern gambling descriptions often connect the system with a belief that results will balance over time. That connection is misleading when it is used to imply that a win becomes due after losses.
Over a very large number of independent fair trials, the proportion of wins may move towards the true probability. That does not mean short runs must alternate or that the next event compensates for the last one. A fair coin remains 50% likely to land heads after five tails. A sporting event is more complex than a coin toss, but earlier losing bets do not automatically improve the probability of the next unrelated selection.
The distinction is the difference between long-run frequency and short-run correction. D'Alembert quietly relies on the second idea, yet probability supplies only the first.
A complete worked sequence
Assume a one-unit starting stake and every selection is offered at decimal 2.00. The sequence below contains six losses and five wins.
| Bet | Stake | Result | Profit or loss | Running profit or loss | 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 | Loss | -2 | -2 | 3 units |
| 5 | 3 units | Loss | -3 | -5 | 4 units |
| 6 | 4 units | Loss | -4 | -9 | 5 units |
| 7 | 5 units | Win | +5 | -4 | 4 units |
| 8 | 4 units | Win | +4 | 0 | 3 units |
| 9 | 3 units | Loss | -3 | -3 | 4 units |
| 10 | 4 units | Win | +4 | +1 | 3 units |
| 11 | 3 units | Win | +3 | +4 | 2 units |
The session finishes four units ahead even though there were more losses than wins. That is the attractive result used to promote the system: wins happened at several of the larger stakes.
Change only the order of the same results and the outcome changes. If the five wins arrive first, the stake remains at one unit because it cannot fall below the base. The following six losses cost 1, 2, 3, 4, 5 and 6 units. The session finishes 16 units down. Nothing about the overall win-loss count changed; only the path changed.
D'Alembert therefore makes timing unusually important. The bettor cannot control whether wins arrive when stakes are large or small, so the apparent recovery feature is not a dependable edge.
How quickly the liability grows
D'Alembert is described as gentle because the next stake rises linearly rather than doubling. The next bet after eight consecutive losses is nine units, not the 256 units demanded by a classic Martingale. The cumulative loss still grows faster than many users expect.
| Consecutive losses | Stakes placed | Total loss in units | Total loss with a £5 unit | Next stake |
|---|---|---|---|---|
| 1 | 1 | 1 | £5 | £10 |
| 3 | 1 + 2 + 3 | 6 | £30 | £20 |
| 5 | 1 to 5 | 15 | £75 | £30 |
| 8 | 1 to 8 | 36 | £180 | £45 |
| 10 | 1 to 10 | 55 | £275 | £55 |
| 15 | 1 to 15 | 120 | £600 | £80 |
| 20 | 1 to 20 | 210 | £1,050 | £105 |
After ten losses, the next £55 bet looks modest beside a Martingale stake, but the bettor has already lost £275. A single winner at £55 would recover only £55 at even money, leaving the account £220 behind. The system needs a favourable sequence of later results, not one rescue bet.
This slower form of damage can be behaviourally dangerous. It may allow a bettor to remain in a losing progression for longer, normalising ever larger stakes before the financial risk feels urgent.
The expected-value test
Stake progression cannot be judged separately from price. Suppose a bet has a true 50% chance, but the available decimal odds are 1.91. A £10 winner earns £9.10 profit, while a loser costs £10.
| Component | Calculation | Result |
|---|---|---|
| Expected winning contribution | 0.50 × £9.10 | £4.55 |
| Expected losing contribution | 0.50 × -£10 | -£5.00 |
| Expected value per £10 staked | £4.55 - £5.00 | -£0.45 |
| Expected return on stake | -£0.45 ÷ £10 | -4.5% |
Every £10 exposed has an expected loss of 45p under those assumptions. Raising the next stake to £20 does not repair the price; it doubles the expected loss to 90p. The same principle applies at every stage.
If the bettor has a genuine edge, sensible staking can determine how much risk to take. If there is no edge, no sequence supplies one. The starting point is always probability and price, covered in our guides to expected value and implied probability.
Why equal numbers of wins and losses do not guarantee profit
At even money, supporters sometimes argue that paired wins and losses will favour the system because a loss increases the stake and a following win occurs for one unit more. A loss of one unit followed by a two-unit win does indeed make one unit.
Reverse the pair and the effect disappears. A one-unit win is followed by another one-unit stake because stakes cannot drop below the base; the loss then returns the pair to zero. Longer sequences create many possible arrangements, including runs in which losses accumulate at increasing stakes and wins later arrive too slowly to repair them.
Prices also move the arithmetic. At decimal 1.90, a two-unit winner earns only 1.8 units, so the simple loss-then-win pair earns 0.8 units rather than one. At 1.80 it earns 0.6 units. Repeated exposure to an overround remains repeated exposure to an overround.
D'Alembert compared with other staking approaches
| Method | After a loss | After a win | Growth during losses | Creates an edge? | Main risk |
|---|---|---|---|---|---|
| D'Alembert | Add one unit | Subtract one unit | Linear next stake, quadratic cumulative loss | No | Slow accumulation can hide the drawdown |
| Martingale | Double stake | Reset | Exponential | No | Bankroll or limit reached quickly |
| Fibonacci | Move one step forward | Move two steps back | Accelerating sequence | No | Large later stakes and path dependence |
| Flat staking | Keep stake unchanged | Keep stake unchanged | None | No | Poor selections remain unprofitable |
Flat staking does not make a bad bet good, but it keeps the financial weight of each decision consistent. That makes results easier to analyse and prevents one late wager from dominating the entire record. Readers comparing systems should also see how to evaluate a betting system, which sets out sample-size, drawdown and price-availability checks.
What happens away from even money
The standard D'Alembert explanation assumes each winning unit earns one unit. Many real markets do not offer precisely 2.00, and the odds can change between selections.
| Decimal odds | Profit on a 5-unit winner | Break-even probability | Effect on the standard recovery story |
|---|---|---|---|
| 1.50 | 2.5 units | 66.67% | One win recovers much less than one equal stake lost |
| 1.80 | 4 units | 55.56% | Recovery is slower than the simple illustration |
| 2.00 | 5 units | 50.00% | Matches the textbook arithmetic |
| 2.50 | 7.5 units | 40.00% | Larger profit, normally paired with lower win probability |
A bettor cannot make the comparison fair by selecting only a price band. Two bets at 2.00 can have very different true probabilities. The useful question is whether each price is greater than the bettor's defensible estimate of fair odds.
Bankroll, limits and the illusion of control
Because the progression is slow, users may believe a large bankroll makes D'Alembert safe. It only moves the stopping point. A finite bankroll cannot guarantee completion of an open-ended sequence, and scaling the base unit upwards recreates the same proportional exposure.
Bookmaker limits introduce another constraint. A sequence may reach a stake the account cannot place, while a market can be suspended, repriced or settled under rules the user did not anticipate. The recovery narrative assumes identical opportunities remain available indefinitely. Sport does not provide them.
Personal limits should not be redesigned to accommodate a progression. A deposit cap and affordable-loss boundary exist to stop exposure, not to be treated as obstacles between the bettor and a theoretical recovery.
The psychology that keeps the sequence running
D'Alembert gives loss chasing a procedure. Instead of deciding emotionally to stake more, the bettor can say the system required it. That can make the escalation feel disciplined even when its purpose is to recover money.
Three biases are particularly relevant:
- The gambler's fallacy: believing a win is increasingly due after losses.
- Loss aversion: feeling the pain of a deficit more strongly than the value of stopping safely.
- Selective memory: remembering the many modest recoveries while discounting the rare sequence that caused most of the damage.
A rule is not the same as control. Genuine control includes the freedom to stop without completing the cycle. If stopping feels impossible because the ledger is negative, the progression has become a behavioural problem regardless of its recent results.
How to test D'Alembert without betting money
Use a spreadsheet or simulation and compare exactly the same selections under D'Alembert and a fixed stake. Do not restart or delete an unfinished losing sequence.
Record:
- the market and selection;
- odds actually available;
- estimated fair probability;
- base unit and current stake;
- profit or loss after every result;
- cumulative amount staked;
- maximum drawdown;
- highest stake required;
- the result using one flat unit on every bet.
| Test question | Why it matters | Warning sign |
|---|---|---|
| Were all prices realistically available? | Paper results can overstate returns | Using best prices that could not be placed |
| How large was maximum drawdown? | Final profit can conceal severe risk | Drawdown far beyond the intended session loss |
| Did one large stake dominate the result? | A fragile record depends on timing | Most profit came from one late winner |
| Was flat staking compared? | It isolates the effect of progression | No benchmark using identical selections |
| Were abandoned cycles counted? | Restarts can erase genuine losses | Losing sequences quietly removed |
Our bet calculator can check returns, but it cannot determine whether a selection is value. That assessment must be made before any stake is chosen.
Common claims examined
"It is safer than Martingale"
It is less aggressive. That is not the same as safe. The next stake grows slowly, while the cumulative deficit keeps rising and may require several well-timed wins to recover.
"Wins and losses eventually balance"
Long-run proportions do not control short-run order. Even if the counts become similar, wins can occur at smaller stakes and losses at larger ones.
"It works with a high strike rate"
A high strike rate says nothing about value without odds. Markets with frequent winners commonly offer shorter prices, so one loss can outweigh several wins.
"Stopping after a small profit removes the risk"
A profit target changes when the session ends, not the expectation of each bet. Many small winning sessions can still be outweighed by one larger losing run.
"The progression is too slow to cause serious harm"
Slow escalation can delay recognition of the problem. At a £5 unit, twenty straight losses total £1,050 before the next bet is considered.
A more defensible alternative
The most transparent approach is to separate selection from staking:
1. estimate the true probability using relevant evidence; 2. convert that probability into fair odds; 3. compare the estimate with a price that is genuinely available; 4. pass when the margin of safety is insufficient; 5. use a small, pre-set stake that remains affordable if it loses; 6. review results across a meaningful sample, including maximum drawdown.
This process cannot guarantee profit, but it keeps the decision connected to price and risk rather than the previous result. Bankroll management in practice explains how fixed units and loss limits can reduce volatility without pretending to manufacture an edge.
Verdict
D'Alembert is easier on a bankroll than Martingale during the early part of a losing run, but it does not solve the problem shared by every negative progression. It assumes that raising stakes after losses can improve the outcome without improving the bet.
The mathematics says otherwise. Expected value follows probability, price and stake. A larger stake magnifies whatever expectation already exists. The system's smoother shape can make it feel measured, yet its cumulative losses grow, its results depend heavily on order and its recovery story weakens outside even-money prices.
No bettor should use a staking sequence to recover money. Set affordable limits before betting, never borrow to continue and stop when the limit is reached. If gambling is becoming difficult to control, use the practical tools and support listed in BetOwl's responsible gambling guide.
Frequently asked questions
What is the D'Alembert betting system?
It is a negative progression that begins with one unit, adds one unit after a loss and subtracts one unit after a win, without dropping below the starting stake.
Does D'Alembert guarantee recovery?
No. Results depend on the order of wins and losses, the prices taken and how long the bettor can continue. A long losing run can create a deficit that several wins are needed to repair.
Is D'Alembert safer than Martingale?
Its stakes rise more slowly, so it is less aggressive. It still encourages larger stakes after losses and cannot remove the risk of a damaging drawdown.
Which odds are best for D'Alembert?
The textbook example normally uses decimal odds of 2.00. No price makes the method profitable by itself; the key issue is whether the available odds represent genuine value.
Can D'Alembert beat the bookmaker's margin?
No. A staking progression cannot change the true probability or the price. If the bets have negative expected value, changing stake sizes does not create a positive edge.