On an invented price path, the price swings for a while and then falls without a pause, with a purchase whose stake is doubled after every loss and returned to one after a win, until the stake can no longer be doubled: 10 trades are marked, 2 closed ahead and 8 closed behind.
A triangle marks where a trade opens and points the way it was taken. A green dot is a trade closed ahead, a red dot one closed behind. The path was invented to show how the approach behaves, including where it goes wrong. It is not market data and it proves nothing.
Also searched asmartingale strategy · martingale trading · doubling down strategy · martingale EA · averaging down
A warning, not an approach to copy
This is a description of an approach people use. It is not a recommendation. No approach works in every market, nothing here has been shown to be profitable, and the chart is invented.
The idea
The martingale is a betting system, far older than electronic markets, and it is a way of sizing, not a way of choosing a trade. After every loss the stake is doubled. When a win finally comes it covers all the losses before it and leaves a gain equal to the first stake. With unlimited money and no limit on size it could not lose. Nobody has unlimited money, and that is the whole of the problem. This page describes it as a caution.
Usually heldUntil a win, or until the money runs out · others held about as long
The rule, as people usually state it
As usually stated: stake 1 unit. After a loss, stake 2, then 4, then 8, doubling each time. After a win, return to 1. The arithmetic: after n losses in a row the total lost is 2ⁿ − 1 units, and the next stake is 2ⁿ. After 5 losses, 31 units are gone and the next stake is 32. After 10 losses, 1,023 units are gone and the next stake is 1,024, all to recover and gain 1 unit.
What it would need to survive
- Unlimited capital, or a run of losses that never reaches the limit of the capital there is.
- No limit on position size and no margin requirement.
- Neither condition is ever met.
What it costs
Every trade pays the spread, and the spread is paid on the doubled size: the eighth trade in a losing run pays 128 times the spread of the first. Margin required grows in the same way. The system does nothing to the odds of any single trade. It rearranges the results: the many small gains are real, and they are paid for by one loss that is certain to arrive if the system is used for long enough.
When it fails
- On a long run of losses. With an even chance on each trade, ten losses in a row has a probability of 1 in 1,024 for any given ten trades. Over thousands of trades that is not a remote event; it is an expected one.
- When the next doubled stake is larger than the account, the margin or the maximum order size allows. The sequence stops there, at its largest loss.
- In a trend, when the doubling is applied to adding to a losing position: the position is at its largest when the market is furthest against it.
The mistakes people make with it
- Taking a long record of small gains as evidence. That record is exactly what a martingale produces before it fails.
- Believing a loss is “due” to be followed by a win. Past outcomes do not change the next one.
- Softening the multiplier, to 1.5 times for example. This delays the end and does not prevent it.
- Not recognising it under other names: averaging down with growing size, “recovery” modes and many automated systems are the same idea.
In the rule bench
The rule bench cannot express a martingale, by design. It sizes every trade from a fixed share of the balance at risk, which is the opposite rule: the stake shrinks after a loss instead of doubling.
Questions people ask
- Does the martingale strategy work?
- No. It does not change the expected result of the trades it is applied to; it only changes how the results are distributed, turning them into many small gains and one loss large enough to take all of them and the starting capital. The arithmetic of doubling guarantees that the required stake outgrows any finite account.
- Why does a martingale look so good for so long?
- Because a run of losses long enough to break it is uncommon in any short period, and until it happens every sequence ends with a small gain. The risk is not visible in the record. It is in the size of the stake the next loss would require.
- What is the gambler’s fallacy?
- The belief that after several losses a win has become more likely. Where outcomes are independent, as with a fair coin, the chance on the next one is unchanged by what came before. The martingale is often defended with this belief.
The words on this page
An explanation for study. It is not advice, a recommendation or a forecast. This is a description of an approach people use. It is not a recommendation. No approach works in every market, nothing here has been shown to be profitable, and the chart is invented.