Martingale Strategy in Trading: Why Doubling Down After Losses Breaks Accounts

The martingale strategy sounds almost too neat: every time a trade loses, you double the size of the next one, so the first winner wins back everything you lost plus your original profit. On paper it can’t fail. On a trading screen it fails in a very particular way, and usually after it has worked for weeks. Let’s walk through the maths with rupee numbers, see why the losing streak that kills it is far more common than it feels, and look at what traders do instead.

What the martingale strategy actually is

The idea comes from gambling. As the Wikipedia entry on the martingale betting system describes it, it is a family of betting strategies that became popular in 18th-century France. The simplest version is played on a coin toss: you win your stake on heads, lose it on tails, and you double the bet after every loss. The first win recovers all the earlier losses and leaves you up by the size of the very first bet.

Traders borrowed it almost unchanged. Investopedia’s explainer on the martingale strategy in forex describes the same doubling-after-a-loss rule, and notes that forex traders use it to bring down their average entry price, which needs a lot of capital. The same page is blunt that a trader without enough money behind the position can lose the whole account.

In the Indian market you will see it in a few disguises. A swing trader who keeps adding to a falling stock in bigger and bigger lots. An options buyer who loses ₹2,000 on Monday’s expiry and puts ₹4,000 into Thursday’s to “get it back”. An intraday trader who doubles quantity after every stop-loss. None of them call it martingale. The mechanics are identical.

A worked example: ₹1,00,000 and a ₹1,000 starting risk

Take a hypothetical trader with ₹1,00,000 of capital. Each trade risks ₹1,000 to make ₹1,000, so wins and losses are the same size. After a loss, the next trade risks double. After a win, the size resets to ₹1,000.

Look at the shape of that. Six losses in a row cost ₹63,000, and the seventh trade needs ₹64,000, which the account no longer has. And what was all that risk chasing? A net gain of ₹1,000, the size of the first bet. The trader is risking 63 times the reward on every cycle without quite realising it.

This is just the general formula at work. If the first bet is B and you lose n times in a row, the total loss is B × (2ⁿ − 1). With B = ₹1,000 and n = 6, that’s ₹1,000 × 63 = ₹63,000. The Wikipedia page uses the same arithmetic with a 63-unit bankroll: six straight losses wipe it out exactly.

Martingale doesn’t remove risk. It swaps a lot of small, visible losses for one large loss that arrives late and all at once.

Why six losses in a row is not rare

Most people’s gut says six losers on the trot is a freak event. For a single run of six trades, that’s roughly right. With a 50% chance of losing each trade, the chance of six losses in a row is 0.5⁶, about 1.56%. Push the loss rate to 55% and it becomes 0.55⁶, about 2.77%.

The trouble is that nobody trades only six times. The question that matters is: over a whole stretch of trading, what’s the chance that a run of six losses shows up somewhere? We worked that out for an idealised coin-flip trader (equal-sized wins and losses, each trade independent):

These are our own calculations from basic probability, not market data, and real trades are messier than coin tosses. But the direction is clear. An intraday trader taking four or five trades a day reaches 200 trades in a couple of months. Even with a fair 50-50 edge, our ₹1,00,000 account has roughly a four-in-five chance of hitting the streak that breaks it in that time. Wikipedia makes the same point about roulette: the odds of six losses in a row are much higher than people intuitively believe.

The expected value never turns positive

There’s a cleaner way to see the problem. Wikipedia’s analysis of a single martingale “round” (a run of losses that ends in either a win or bankruptcy) gives the expected profit per round as B × (1 − (2q)ⁿ), where q is the probability of losing a bet and n is the number of losses you can afford.

• If q is exactly 0.5, then 2q is 1, and the expected profit is zero. Doubling has added nothing except bigger swings.
• If q is above 0.5, (2q)ⁿ is greater than 1, so the expected profit is negative, and it gets worse as n grows.

Plug in our trader: B = ₹1,000, n = 6. At a 55% loss rate, the expected profit per round is ₹1,000 × (1 − 1.1⁶), which works out to roughly −₹772. Every cycle, on average, costs money, even though most cycles end with a small win.

Why would a trader’s loss rate sit above 50% on a trade that looks like a coin flip? Costs. Brokerage, STT, exchange charges, GST, stamp duty and slippage all come out of every trade, win or lose. A strategy that is a fair game before costs is a losing one after them. Wikipedia also notes that martingale fails even with unlimited time as long as there is a limit on bets or bankroll, and every real trader has both.

What makes trading worse than the casino version

In a casino, at least, a losing bet is a clean loss of the stake. Markets add a few twists.

Losses aren’t capped at the stake. A stop-loss on a stock can be jumped by a gap. If a company announces bad results after market hours and the stock opens 8% lower, your planned ₹32,000 loss on trade six can turn into something bigger. With a doubled position, the gap hurts twice as much.

Position sizes are lumpy. Futures and options in India trade in fixed lot sizes. You can’t always “double” neatly; you jump from one lot to two to four, and margin requirements jump with them. A broker’s margin call can close you out before your doubling sequence is even finished.

Losing streaks cluster. The coin-flip maths assumes each trade is independent. In markets, losses often bunch together: a strategy built for a range-bound Nifty can lose again and again once a strong trend sets in. That makes long streaks more likely than the table above, not less.

The psychology is brutal. At trade five, you are risking ₹16,000 to make back ₹15,000 and earn ₹1,000 on top. Most people start breaking the rules here, either skipping the doubled trade and locking in the loss, or going bigger than the plan. We cover the wider pattern in our piece on trading biases that wipe out accounts.

A martingale equity curve looks like a gently rising line, right up until the day it falls off a cliff. Weeks of good results tell you almost nothing about its risk.

How traders spot martingale behaviour in their own account

Very few people sit down and decide to run a martingale. It creeps in. A few warning signs:

• Your position size goes up right after a loss, not after a win or a change in volatility.
• You average down on a losing stock without having decided in advance how many times you will do it.
• Your trading journal shows lots of small wins and a handful of very large losses that wipe out months of gains.
• You catch yourself thinking “the next one has to work” because the last few didn’t.

If two or more of these sound familiar, pull up the last three months of trades and sort them by position size. If the biggest sizes sit right after losing trades, that’s martingale.

What to use instead of the martingale strategy

The fix is not a cleverer doubling rule. It’s sizing that shrinks when you’re losing and only grows from profits.

Fixed-fractional sizing. Risk a fixed percentage of current capital on each trade, say 1%. On ₹1,00,000 that’s ₹1,000. After a loss the account is smaller, so the next risk is slightly smaller too. Six losses in a row at 1% each take the account down to about ₹94,148 (₹1,00,000 × 0.99⁶), a setback of under 6%, compared with a 63% hole under martingale.

Anti-martingale, or pyramiding into winners. This flips the logic: you add to positions that are working and cut size on losers. Our guide to the pyramiding trading strategy walks through how to add in stages without giving back the open profit.

Edge-based sizing. If you have a measured win rate and a measured reward-to-risk ratio, you can size according to that edge rather than your last result. The Kelly criterion for position sizing is the formal version, and most traders use a fraction of it to stay safe. It pairs naturally with a clear risk-reward ratio for each setup.

For the bigger framework (daily loss limits, maximum open risk, when to stop trading for the day), our risk management in trading guide pulls it all together.

Is there any place where martingale makes sense?

Honestly, very few. Wikipedia’s analysis shows the strategy can only be argued to win with unbounded wealth, unbounded bets and unbounded time, none of which a retail trader has. Some systematic traders run small, capped “averaging” rules, for example adding once to a position in a long-term holding they would happily own at a lower price. But the moment the rule says “double after every loss, with no limit”, the result is a strategy that wins small, often, and eventually loses big.

If you remember one number from this piece, make it 63. That’s how many times the original bet you have riding on a six-loss martingale run, all to win back one unit. Ask yourself whether any trade is worth that before you double up.

Key takeaways

• The martingale strategy doubles the position after every loss so one win recovers everything plus the first bet’s profit.
• A run of n losses costs B × (2ⁿ − 1). Six losses on a ₹1,000 starting risk is ₹63,000.
• Long losing streaks are far more common over hundreds of trades than they feel. In our coin-flip example, a six-loss streak shows up in about 80% of 200-trade stretches even at a 50% loss rate.
• With a loss rate at or above 50%, which costs push most traders towards, the expected value per round is zero or negative.
• Fixed-fractional sizing, pyramiding into winners and edge-based sizing are the sensible alternatives.

Disclaimer: This article is for educational purposes only and is not investment advice. All examples use hypothetical numbers. Trading in equities and derivatives involves risk of loss; please consult a SEBI-registered investment adviser before making investment decisions.