Martingale Strategy in Trading: The Maths of Doubling Down

Martingale doubles the position size after every loss so that a single win recovers everything. It works exactly as advertised, dozens of times in a row, and then the sequence runs one trade longer than the account can fund.

Martingale at a glance

What it is and why it works

Martingale comes from an eighteenth-century casino betting system. Bet one unit on a coin flip; if you lose, bet two; if you lose again, bet four. Whenever you finally win, you recover every previous loss and finish the sequence one unit ahead. The logic is airtight, and that is what makes it so persuasive: you really will win back everything, provided you can keep doubling.

In trading the same idea is dressed up as recovery, averaging, or a smart money-management overlay. You open 0.01 lots. It hits the stop. You open 0.02 lots. That loses too, so 0.04, then 0.08. Some versions never place a stop at all and simply add a doubled position at each adverse level, which is martingale and grid combined and is worse than either.

Here is what the doubling actually looks like. Take a 20-pip stop on an instrument where a standard lot is worth about ten units of account currency per pip, so 0.01 lots risks about two units. Losses run 2, 4, 8, 16, 32, 64, 128, 256, 512, 1,024. Ten consecutive losses have cost 2,046, and the eleventh trade, the one that is supposed to fix it, requires 10.24 lots and risks 2,048 on its own. You are staking two thousand to win two.

That is the entire strategy in one line. The reward for each completed sequence is fixed and tiny; the stake required to keep the sequence alive doubles without limit. A ten-loss streak is a one-in-1,024 event on a coin flip, which sounds comfortingly rare until you notice that a trader taking five trades a day will see roughly a thousand trades a year. The streak is not a freak occurrence you might avoid. It is a scheduled event whose date you do not know.

How to trade it, step by step

1. Write the doubling schedule out to fifteen levels before you risk anything. Start at your intended first lot size and double it fifteen times: 0.01, 0.02, 0.04, 0.08, 0.16, 0.32, 0.64, 1.28, 2.56, 5.12, 10.24, 20.48, 40.96, 81.92, 163.84. Seeing the numbers written down does more than any argument. Level 15 is over sixteen thousand times level one.
2. Convert each level into money using your actual stop distance. Multiply the lot size at each level by your stop in pips and by the pip value for your instrument; the pip value calculator gives you the exact figure. You now have a column showing what a loss costs at every step of the sequence.
3. Add up the cumulative loss down that column. Because each level is double the last, the running total is always roughly equal to the next single bet. That is the crucial property: at any point in the sequence, the amount you have already lost is about the same as the amount you are about to put at risk on one trade.
4. Find the level at which the required lot size exceeds your margin. Take your account balance, the leverage your broker gives you and the margin per lot on your instrument, and find the first row in the table you cannot fund. That is your true maximum streak length, and it is usually far shorter than traders expect: often eight to eleven steps rather than the twenty they imagine.
5. Work out how likely a streak of that length is over a year of trading. If your strategy wins roughly half its trades, the chance of any particular sequence of n losses is one in two to the power of n: one in 256 at eight, one in 1,024 at ten. Multiply by the number of trades you expect to take in a year. For most active traders the answer is that the streak is not merely possible but likely.
6. Compare the total profit of a completed sequence with the loss when one fails. Every successful sequence, however long, nets you one unit of profit: two units of account currency in the example above. One failed sequence costs the entire balance. Ask how many completed sequences it takes to fund a single failure, and note that you must complete them all before the failure occurs, not after.
7. Check whether your broker or funded programme permits it at all. Most prop firms explicitly prohibit martingale and other loss-doubling systems, and breaching the rule voids the account even in profit. Many brokers also impose maximum position sizes that cut the sequence short and leave you holding the largest loss with no way to continue.
8. If you want the position size to change with results, invert it. Anti-martingale, increasing size after wins and decreasing it after losses, has the opposite risk profile: drawdowns shrink as they deepen instead of accelerating. It is not a source of edge either, but it does not contain a built-in ruin point.
9. Replace the recovery instinct with a fixed fractional rule. Risk the same small percentage of the current balance on every trade regardless of what the last one did, and set the lot size from your stop distance using the position size calculator. Under that rule a losing streak reduces your risk automatically, which is the exact opposite of what martingale does.

Size every one of those entries with the position size calculator and check the trade is worth taking with the risk/reward calculator before you commit.

The conditions it needs

An unlimited bankroll and no position limits

Martingale is mathematically sound only under the assumption of infinite capital. With infinite money you would always eventually win, but you would also not need to trade. Every real account has a finite balance, every broker has a maximum lot size, and the strategy’s guarantee evaporates the moment either constraint binds.

A genuine edge underneath it, which martingale does not supply

A staking plan changes the sequence of your bets, not their expected value. If each individual trade has negative expectancy after spread and commission, doubling the size of a negative-expectancy trade makes the expected loss larger, not smaller. Martingale is often adopted precisely because a trader has no edge and hopes that money management can substitute for one. It cannot.

Independence between trades, which markets do not offer

The casino version assumes each flip is independent. Market losses are not independent: they cluster. A strategy that fails does so because conditions have changed (volatility has expanded, a range has broken, a trend has started) and those conditions persist for hours or days. Martingale demands its largest positions exactly when the market is behaving in the way that is causing the losses.

A hard cap on the sequence, which removes the point of it

You can limit the sequence to, say, four doublings and accept the accumulated loss at that point. That does bound the damage, and it turns the method into an ordinary strategy with a maximum loss of fifteen times the base risk. Whether that is acceptable is then just a normal risk-to-reward question, and the martingale framing adds nothing to answering it.

When it fails

• The failure is certain, only the date is unknown. This is the difference between martingale and an ordinary bad strategy. Given enough trades, a losing streak longer than your account can fund is not a risk, it is an eventuality. The strategy’s entire design converts a small, survivable loss into one enormous unsurvivable one.
• You risk everything to win almost nothing. A completed sequence returns a single base unit of profit no matter how many doublings it took. A failed sequence costs the balance. The reward-to-risk ratio at the deep end of the sequence is measured in thousandths, which no win rate can compensate for.
• Margin runs out before the streak does. Traders plan for ten or twelve levels and discover the broker will not let them open level nine, or that the required margin exceeds their equity two steps earlier than expected. The sequence stops where the money stops, and it stops holding the biggest position of the run.
• Losses cluster, which is the worst possible property for this method. Market losses arrive in runs because market conditions persist. Martingale is at maximum size in exactly the conditions that produced the losing run, and those conditions typically last longer than the sequence can.
• It feels like control right up to the end. Every recovered sequence reinforces the belief that the method works, because it did work, that is what makes it dangerous rather than merely wrong. Traders often increase the base lot size after a good run, which shortens the number of levels the account can fund and brings the failure forward.
• It disqualifies you from funded trading entirely. Prop firms detect doubling patterns in trade history and treat them as a breach of the risk rules. A challenge passed with a martingale sequence is normally voided on review, so the strategy is not just risky, it is unusable in that whole part of the industry.

For different levels of experience

If you are brand new

If someone has shown you an account statement with a hundred winning trades and no losses, you are almost certainly looking at a martingale system mid-sequence. Perfect records are not evidence of skill; they are the signature of a method that hides losses inside open positions and larger bets. The account with no losing trades is the one at greatest risk.

The instinct martingale exploits is completely natural. You lose money, you want it back, and doubling up promises to get it back in a single trade. Every beginner feels this. Learning to take the small loss and move on, without trying to win it back on the next trade, is the single skill that separates traders who last from traders who do not.

The practical replacement is simple and dull: decide before each trade what percentage of your account you are willing to lose, commonly 0.5% or 1%, and keep that percentage identical whether you have just won five in a row or lost five in a row. Under that rule losing streaks make your positions smaller, and no sequence of losses can end the account. Start with risk management and position sizing.

If your results are inconsistent

Most intermediate traders would never call themselves martingale traders and still do it occasionally. The pattern is the “get it back” trade: a normal loss, immediately followed by a larger position taken with less analysis, because the aim has quietly shifted from trading well to being flat again. That is martingale with one step.

It shows up clearly in a trading journal. Record position size alongside the result of the previous trade and look for correlation. If your size after a loss is systematically larger than your size after a win, you have a revenge-sizing habit, and it will eventually meet a bad run.

The fix is mechanical rather than psychological: fix the risk percentage in advance, calculate the lot size from the stop distance rather than choosing it, and impose a daily loss limit that stops you trading for the day. A rule that removes the decision at the moment you are least capable of making it well is worth more than any amount of resolve.

If you are experienced

Formally, martingale converts a symmetric return distribution into an extremely negatively skewed one without changing expectancy. The mean is unaltered by the staking plan; the distribution is reshaped so that almost all outcomes are small positives and a small number are total loss. Under the Kelly framework the position sizes involved are far beyond the growth-optimal fraction, usually beyond even the zero-growth threshold, so the median outcome is ruin even when the underlying edge is positive.

The deeper problem is that martingale destroys the informational content of the equity curve. Win rate, profit factor and Sharpe are all computed over a sample that systematically excludes the tail, so every standard performance statistic is biased upwards and the bias is invisible in-sample. Any evaluation must therefore be done on the underlying signal at constant size, with the staking plan modelled separately as a path-dependent overlay and assessed on risk of ruin rather than on return.

If size is to vary with results at all, the defensible direction is anti-martingale: fractional-Kelly scaling on realised equity, which decreases exposure into drawdown and produces a bounded, path-dependent but non-ruinous outcome. That is what professional risk systems implement, and it is the exact inverse of what is being sold to retail traders as a recovery method.

Risk management for this strategy

Martingale is not a strategy with bad risk management; it is the absence of risk management expressed as an algorithm. Risk management means deciding your maximum loss before you enter. Martingale explicitly refuses to accept any loss as final, which means its maximum loss is undefined and therefore equal to the balance.

The replacement is fixed fractional sizing. Pick a percentage of the account to risk per trade, place the stop where your idea is proven wrong, and let the distance between entry and stop determine the lot size: not the other way round. Use the position size calculator for the arithmetic. The important property is that risk scales down with the balance, so a losing run shrinks your positions automatically and no streak, however long, can take you to zero.

Add a circuit breaker on top: a maximum number of consecutive losses or a daily loss limit after which you stop for the day. Martingale is fundamentally an emotional response to loss, so the countermeasure has to be a rule that operates when your judgement is compromised rather than a resolution to behave better.

Where Market Structure Pro fits

Martingale is what traders reach for when they cannot tell a bad trade from a bad market. If every loss feels like an accident that the next trade should correct, doubling down seems reasonable. The alternative is a reliable read on whether conditions currently support taking a trade at all, because the honest answer to a string of losses is usually not “trade bigger” but “stop trading for now”.

Market Structure Pro is built around that verdict. It combines 27 tools into one output (TRADE, TRANSITION or NO TRADE) with a confidence percentage, an A/B/C grade and a plain-English explanation of the reasoning. Its ranging and chop filter exists specifically to return NO TRADE in the dead and directionless conditions that generate the clustered losing runs martingale cannot survive. It is session-aware and spread-aware, so a low-quality setup in thin hours is graded as what it is rather than taken as a recovery opportunity.

Because the state locks on the closed bar and does not repaint, the record of what it said is fixed and reviewable, which makes it useful in a journal: you can see whether your losing streak came from trading good setups in a difficult market or from taking trades the read had already flagged as poor. MSP is decision support only (it places no trades, is not a signal service and guarantees nothing) but replacing “win it back” with a graded read on current conditions removes the impulse martingale feeds on.

Frequently asked questions

What is the martingale strategy in trading?

Martingale means doubling your position size after every losing trade so that one eventual winner recovers all the previous losses plus the original intended profit. It originated as a casino betting system for even-money bets. In trading it is applied either by doubling the size of the next trade or by adding doubled positions at successively worse prices without a stop.

Does martingale work in forex?

It works in the narrow sense that completed sequences do recover prior losses, which is why short-term results look consistent. It fails because it requires an unlimited bankroll: with a finite account and broker position limits there is always a streak length you cannot fund, and reaching it costs the whole balance. Over a large number of trades that streak becomes likely rather than remote.

How quickly does martingale lot size escalate?

It doubles at every step, so the growth is exponential. Starting at 0.01 lots the sequence runs 0.01, 0.02, 0.04, 0.08, 0.16, 0.32, 0.64, 1.28, 2.56, 5.12, 10.24; the eleventh trade is over a thousand times the first. Cumulative losses grow at the same rate, so at any point in a sequence the amount already lost is roughly equal to what the next single trade risks.

How many losses in a row can a martingale account survive?

Fewer than most traders assume, because margin runs out before the doubling schedule does. Depending on account size, leverage and base lot, the practical limit is often around eight to eleven consecutive losses. At a roughly even win rate a ten-loss streak occurs about once in every 1,024 sequences, which an active trader taking a few trades a day can expect to encounter within a year.

Do prop firms allow martingale?

Almost none do. Funded-account programmes routinely prohibit martingale, grid systems and other methods that increase exposure into losses, and reviewers look for doubling patterns in trade history. A challenge passed using martingale is typically voided even if the account is in profit, so the method is unusable for anyone wanting to trade firm capital.

What is anti-martingale?

Anti-martingale increases position size after winning trades and decreases it after losses, which is the reverse schedule. It reduces exposure during drawdowns instead of amplifying it, so it has no built-in ruin point. Like martingale it creates no edge of its own, but it does not convert a normal losing streak into a total loss.

Is martingale better than grid trading?

They are close relatives with the same flaw: neither closes a losing position. A grid adds equal-sized positions at fixed intervals, so exposure grows linearly with the adverse move, while martingale doubles, so exposure grows exponentially. Martingale therefore reaches the account-ending point considerably faster.

Can martingale be made safe with a small starting lot?

A smaller base lot buys you one or two extra levels in the sequence and nothing more, because each additional level requires double the capital of the last. Halving the starting size adds roughly one step before ruin. Sizing changes how long the strategy takes to fail, not whether it fails.

What should I do after a losing streak instead of doubling down?

Keep the risk per trade at the same fixed percentage of your current balance, which automatically reduces the money at risk as the balance falls. Stop trading for the day once a preset daily loss limit is hit, and review the losing trades to establish whether the setups were poor or the market conditions were unsuitable. Recovering losses is a by-product of trading well, never a target for the next trade.

Related reading

• Grid Trading: The same refusal to close a loser, escalating linearly rather than exponentially.
• Position Sizing: The fixed fractional rule that makes ruin arithmetically impossible.
• Trading Psychology: Why the urge to win it back on the next trade is so hard to resist.
• Risk-Reward and Expectancy: Staking plans do not change expectancy. This explains what does.
• Risk Management: Defining the maximum loss before entry; the thing martingale removes.