Does martingale actually work?

The mathematics is correct under unlimited capital, which is why it produces long stretches of reliable gains. Under finite capital the expected outcome is ruin, because the required position size grows geometrically while the account does not. The strategy works every time until the sequence outlasts the balance, and then it takes all of it.

How do I tell if an EA uses martingale?

Look at the inputs for a multiplier, recovery factor or lot progression with a default above 1.0, usually beside a maximum levels cap and often with no stop-loss. In the Strategy Tester's trade list, volume rising through a losing run and resetting after a win is conclusive regardless of what the description says.

Is a soft martingale with a 1.3 multiplier safer?

It survives a longer run before failing, which is a real difference, but the shape is unchanged: size still grows geometrically and the loss still lands in one sequence. It buys length, not a different outcome, and it is best evaluated by the size of the capped sequence rather than by the multiplier.

Do any EAs on mt5depot use martingale?

No. Every listing here states its position handling explicitly — one position per leg, fixed lot, fixed stop and target, no grid, no martingale and no averaging down — and publishes its worst losing streak next to its drawdown so the two can be read together.

Martingale explained — why the sizing rule that always works eventually ends the account

In plain English

Lose, then trade twice as big; lose again, twice as big again. The next win pays for everything before it. It works every time until the sequence runs longer than the account, and then it takes the account.

Why it matters

Martingale is the single most common reason a beautiful equity curve turns into a total loss. It produces months of small, regular gains and one loss the size of everything, which is exactly the shape that looks best in a short record and worst in a long one.

It converts a low-probability event into a certain one. Each doubling is survivable; the sequence is what is not, and long sequences are ordinary rather than rare.
It flatters every statistic that a buyer looks at. Win rate above 95%, a smooth curve and a high profit factor are what the method produces right up to the failure.
The failure is not gradual. There is no drawdown warning of the right size — the account goes from a normal month to a stop-out inside one sequence.
It is easy to spot in an EA and easy to hide in marketing, so knowing what to look for in the inputs is the practical defence.

How the strategy works

1Open a position at a base lot. If it closes at a loss, open the next one at a multiple of that lot, conventionally two.
2Continue multiplying after each loss, so the position size grows geometrically with the length of the losing run.
3When a trade finally wins, it recovers every prior loss in the sequence plus the original target, and the size resets to the base lot.
4The mathematics is sound and complete on one condition — that capital is unlimited — which is the one condition no account meets.
5Grid variants place the added positions at fixed price intervals rather than after closed losses, and soft variants multiply by 1.3 or 1.5 instead of 2, which lengthens the survivable run without changing what ends it.

Market conditions that suit it

Ranging markets where price returns to a level repeatedly, which is where the sequence terminates quickly and the method appears to work.
Symbols with low overnight cost, because the losing legs are held while they are being added to.
Short evaluation windows — a few months of range-bound conditions produce the smoothest record the method will ever show.
In practice none of these are conditions that can be relied on ahead of time, which is the entire problem: the method requires the future to keep resembling a benign past.

Main advantages

A very high proportion of winning sequences, which feels like a high win rate and reads like one in a summary.
A smooth equity curve for as long as no sequence exceeds the account's capacity.
No forecasting skill is required. The rule needs only an exit and a multiplier, which is why so many EAs are built this way.
Recovery from a loss is immediate rather than gradual, which is genuinely useful psychologically and genuinely dangerous financially.

Main risks

Required capital grows geometrically: ten doublings is 1,024 times the base lot, and runs of ten to fifteen losses occur in ordinary market conditions.
The account is ended by margin, not by the strategy — a stop-out closes the sequence at the worst point, and the recovery trade that would have paid never happens.
Broker limits arrive before mathematical failure. Maximum volume caps and margin requirements refuse the doubled order while the sequence is still running.
Backtests understate the risk badly, because whether the record contains the fatal run is a matter of which years were sampled.
Combining it with the absence of a stop-loss, which is the usual pairing, removes the only mechanism that would bound a single leg.

Typical EA behaviour

Lot size rises after losing trades and resets to the base after a win — visible directly in the Strategy Tester's trade list.
The equity curve shows long smooth stretches interrupted by deep, near-vertical excursions rather than proportionate drawdowns.
Open drawdown far exceeds closed drawdown, because the losing legs are held rather than realised.
Trade duration on the losing side stretches as the sequence grows, since the EA is waiting for the reversal that closes it.
The worst losing streak in the report matters more than every other statistic, because the streak is what determines the capital required.

How to spot it in an EA

An input named multiplier, lot multiplier, martingale factor, recovery factor or similar, with a default above 1.0.
A maximum trades or maximum levels input, which exists precisely because the sequence would otherwise be unbounded.
No stop-loss input, or a stop-loss default of zero — the two features go together because a stop would end the sequence at a loss.
In the tester's trade list, volume increasing monotonically through a losing run and resetting to base after a win.
In marketing, a headline win rate above 95% combined with a smooth curve and no mention of the worst losing streak.

Example

A base lot of 0.01 doubled after each loss, on a $5,000 account. The sequence is entirely ordinary until the row where it is not.

Loss 1–5: 0.01 → 0.16 lots; Cumulative exposure still small. The method looks like it is working.
Loss 8: 1.28 lots; One position is now larger than most retail accounts would open deliberately.
Loss 11: 10.24 lots; Margin for this leg alone exceeds the account at most leverage settings.
Capital the full run needs: about 2,048× the base risk; A run of eleven losses is not rare — it is a normal feature of a long record.

Nothing unusual happened here. The strategy performed exactly as designed, and the design requires capital no retail account has.

Calculation 0.01 × 2^11 = 20.48 lots · cumulative 2^12 − 1 = 4,095 × base

Result The eleventh loss ends the account, and the twelfth trade would have won

How it is used

The practical use of this term is as a filter. Identify the method in an EA before buying it, and read the worst losing streak rather than the win rate.

Range	What it means
No lot multiplier, hard stop on every trade	Not martingale. A single loss is bounded by the stop.
Multiplier under 1.5 with a hard level cap and a stop	Bounded, but the cap is where the loss lands. Size for the capped sequence, not the base lot.
Multiplier of 2 with a level cap	The cap converts an unbounded loss into a very large one. Read the cap as the real risk per sequence.
Any multiplier with no stop-loss and no cap	Loss is bounded only by the account. The question is when, not whether.

Read the inputs before the results. A multiplier above 1.0 changes how every other number in the report should be read.
Treat the worst losing streak as the headline statistic, because the required capital is the base risk multiplied through that streak.
Check whether drawdown is reported on closed trades or on equity. Martingale hides its exposure in open positions, so closed-trade drawdown understates it.
Ask what happens at the level cap. A capped martingale still has a worst case; it is just written down somewhere.
Discount any record shorter than several years. Whether the fatal sequence appears is mostly a question of sample length.

None of the EAs listed in this catalogue use martingale, grid or averaging down — every listing states it explicitly and publishes its worst losing streak alongside its drawdown.

Common mistakes

Reading the win rate as the risk

A martingale EA can genuinely win 97% of its sequences and still end at zero, because the 3% is not a small loss but the whole account. Win rate describes how often the method resolves; it says nothing about what happens when it does not, which is the only number that decides the outcome.

Believing a level cap makes it safe

A cap converts an unbounded loss into a bounded but very large one. At a multiplier of two and a cap of ten levels, the capped sequence loses roughly a thousand times the base risk. That is a design decision, not a safety feature, and the account has to be sized for the capped loss rather than the base lot.

Trusting a short backtest

The method fails on a long losing run, and whether such a run appears is largely a function of how many years were sampled. A two-year backtest that missed it is not evidence of robustness; it is evidence that two years was not long enough to contain the event the design depends on avoiding.

Confusing it with grid trading

Grid places orders at price intervals; martingale multiplies size after losses. They are frequently combined, and a flat-lot grid is a genuinely different risk profile from a multiplying one. The question to ask about any adding strategy is whether the added positions are the same size as the first.

Frequently asked questions

Does martingale actually work?: The mathematics is correct under unlimited capital, which is why it produces long stretches of reliable gains. Under finite capital the expected outcome is ruin, because the required position size grows geometrically while the account does not. The strategy works every time until the sequence outlasts the balance, and then it takes all of it.
How do I tell if an EA uses martingale?: Look at the inputs for a multiplier, recovery factor or lot progression with a default above 1.0, usually beside a maximum levels cap and often with no stop-loss. In the Strategy Tester's trade list, volume rising through a losing run and resetting after a win is conclusive regardless of what the description says.
Is a soft martingale with a 1.3 multiplier safer?: It survives a longer run before failing, which is a real difference, but the shape is unchanged: size still grows geometrically and the loss still lands in one sequence. It buys length, not a different outcome, and it is best evaluated by the size of the capped sequence rather than by the multiplier.
Do any EAs on mt5depot use martingale?: No. Every listing here states its position handling explicitly — one position per leg, fixed lot, fixed stop and target, no grid, no martingale and no averaging down — and publishes its worst losing streak next to its drawdown so the two can be read together.

Related terms

Related resources

Tool Risk of Ruin How often a sizing rule ends an account, given its worst losing streak. Guide Martingale The strategy page, with the arithmetic and the variants set out in full. EAs Browse EAs by historical drawdown Sorted by drawdown, which is the figure a multiplying sizing rule is designed to keep small until it is not.