Martingale, Why Doubling Down Eventually Destroys Accounts
Martingale is a system where you double your position size after each loss, aiming to recover all previous losses with one win. On paper, it seems mathematically clever. In practice, it's one of the fastest ways to blow an account.
Example: you risk $10 and lose. Next trade, $20. Lose again. Then $40, $80, $160. After five losses, your next bet is $320, and total losses are $310. A single win would only bring you back to a small profit.
Why it fails:
- Losing streaks happen. Even a 70% win rate strategy can produce long streaks.
- Account size is finite. You eventually can't afford the next doubling.
- Brokers have limits. Maximum lot sizes and margin requirements stop you.
- Small gains, huge risks. Most of the time you win tiny amounts, but one bad streak erases everything.
Why it's tempting:
- Frequent small wins create a false sense of safety
- Backtests may look excellent until a long losing streak appears
Warning signs in products:
- Automated trading robots with smooth equity curves and high win rates but no stop losses
- Claims of "no losing months"
Safer ideas:
- Fixed or decreasing risk after losses, not increasing
- Strict stop losses
- Realistic expectations about streaks
If a system needs you to increase risk to recover from losses, it carries a hidden cliff.
Avoid the cliff. Survive first, then grow.
Martingale, Why Doubling Down Eventually Destroys Accounts
Re: Martingale, Why Doubling Down Eventually Destroys Accounts
Something worth trying if martingale still sounds tempting. Take a coin and pretend each flip is a trade, double the stake after every loss, start at 1 unit with a 100 unit account. Most runs, you'll be fine for a while. Then a string of six or seven losses shows up and the account's gone. It happens faster than people expect.
It’s Fairman 
-
LondonNewsTrader
- Posts: 421
- Joined: Mon Sep 21, 2026 9:30 am
Re: Martingale, Why Doubling Down Eventually Destroys Accounts
The arithmetic in your example checks out: 10, 20, 40, 80 and 160 lost is 310, the sixth bet is 320, and a win nets you 10 dollars. Risking 320 to make 10 is the whole problem in one line.
People underestimate how often the bad streak arrives. With a 70 percent win rate, the chance of at least one run of six straight losses is about 22 percent over 500 trades and about 40 percent over 1,000. Seven in a row over 1,000 trades is still around 14 percent. Those are not rare events for an EA that trades several times a day.
In lot terms it gets ugly faster than people expect. Start at 0.01 and ten doublings later you need 10.24 lots, which most small accounts cannot margin, and many brokers cap the maximum order size anyway. News days make it worse, because the losing streak tends to come in one directional move with no pullback for the grid to recover on.
People underestimate how often the bad streak arrives. With a 70 percent win rate, the chance of at least one run of six straight losses is about 22 percent over 500 trades and about 40 percent over 1,000. Seven in a row over 1,000 trades is still around 14 percent. Those are not rare events for an EA that trades several times a day.
In lot terms it gets ugly faster than people expect. Start at 0.01 and ten doublings later you need 10.24 lots, which most small accounts cannot margin, and many brokers cap the maximum order size anyway. News days make it worse, because the losing streak tends to come in one directional move with no pullback for the grid to recover on.
Re: Martingale, Why Doubling Down Eventually Destroys Accounts
Some people ask what about a softer version, like 1.5x after a loss? It's the same idea, just slower. The core problem is that it raises size exactly when your strategy is struggling. Any system that does that eventually gets caught by a long losing streak. With 1.5x, starting at 1 percent, eight losses in a row means the ninth trade risks about 25 percent of the starting balance, after already losing around 50 percent. Eight losses in a row happens to normal strategies with a 40 percent win rate over a few hundred trades. So it still breaks, just a few months later than full martingale does.
It’s Fairman 
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Martingale, Why Doubling Down Eventually Destroys Accounts
The coin flip test above also works on paper. Say the account is 1,000 dollars, the first stake is 10, and every loss doubles the next stake.
Six losses in a row cost 10, 20, 40, 80, 160 and 320, which is 630 dollars. The seventh stake would be 640 with only 370 left, so the system is finished after any run of six.
On a fair coin, six losses in a row from the start of a cycle happen once in 64 tries. The other 63 cycles make 10 dollars each, so 630 in total, and the one bad cycle loses 630, which leaves the expected result at exactly zero before costs.
The chart shows the stake and the running loss at each step against the 1,000 dollar balance. The seventh stake is the one that cannot be paid.
Spread and commission turn that zero into a loss, since every trade pays costs. So a long run of small wins in a backtest says nothing about whether the method has an edge.
The 1.5x version mentioned above only stretches the picture out. It needs a longer streak to break, but still raises risk exactly when the strategy is doing worst.
Six losses in a row cost 10, 20, 40, 80, 160 and 320, which is 630 dollars. The seventh stake would be 640 with only 370 left, so the system is finished after any run of six.
On a fair coin, six losses in a row from the start of a cycle happen once in 64 tries. The other 63 cycles make 10 dollars each, so 630 in total, and the one bad cycle loses 630, which leaves the expected result at exactly zero before costs.
The chart shows the stake and the running loss at each step against the 1,000 dollar balance. The seventh stake is the one that cannot be paid.
Spread and commission turn that zero into a loss, since every trade pays costs. So a long run of small wins in a backtest says nothing about whether the method has an edge.
The 1.5x version mentioned above only stretches the picture out. It needs a longer streak to break, but still raises risk exactly when the strategy is doing worst.
-
FXS Prop Desk
- Posts: 1691
- Joined: Wed Oct 07, 2026 9:52 am
Re: Martingale, Why Doubling Down Eventually Destroys Accounts
The problem with martingale is that it trades a high chance of small wins for a small chance of a total loss. The equity curve looks smooth for a long time, which makes it very attractive, until a losing streak long enough to wipe the account appears.
On prop accounts, martingale is usually forbidden or quickly breaches the daily loss limit. Even a modest losing streak of six or seven trades at doubled size can reach the limit in a single session.
On prop accounts, martingale is usually forbidden or quickly breaches the daily loss limit. Even a modest losing streak of six or seven trades at doubled size can reach the limit in a single session.
FXS Official Team | The Idea Lab & Strategy
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Martingale, Why Doubling Down Eventually Destroys Accounts
A common answer to the streak problem is a capped version, where the stake doubles at most twice and then resets. Say the steps are 10, 20 and 40 dollars on an even money trade.
Three losses in a row lose 70 dollars, which on a fair coin happens one cycle in eight. The other seven cycles end on a win and make 10 each, so the average is 0.875 times 10 minus 0.125 times 70, exactly zero.
The cap removes the blow up but not the problem. It swaps a rare huge loss for a more frequent 70 dollar one, and costs still push the total below zero.
The stake pattern only rearranges when the losses arrive. If a method has no edge at a flat 10 dollars a trade, no sequence of stakes will create one.
The reverse idea, adding size after wins instead of losses, at least raises risk while the method is working. It still needs a hard reset, or one losing trade at the top of a run takes back the whole streak.
When a vendor shows a smooth curve, asking for the largest stake ever placed and the balance at that moment tells you most of what the backtest hides. Has anyone seen an EA sales page that publishes that number?
Three losses in a row lose 70 dollars, which on a fair coin happens one cycle in eight. The other seven cycles end on a win and make 10 each, so the average is 0.875 times 10 minus 0.125 times 70, exactly zero.
The cap removes the blow up but not the problem. It swaps a rare huge loss for a more frequent 70 dollar one, and costs still push the total below zero.
The stake pattern only rearranges when the losses arrive. If a method has no edge at a flat 10 dollars a trade, no sequence of stakes will create one.
The reverse idea, adding size after wins instead of losses, at least raises risk while the method is working. It still needs a hard reset, or one losing trade at the top of a run takes back the whole streak.
When a vendor shows a smooth curve, asking for the largest stake ever placed and the balance at that moment tells you most of what the backtest hides. Has anyone seen an EA sales page that publishes that number?
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Martingale, Why Doubling Down Eventually Destroys Accounts
The idea is simple. After a loss, double the next position. When a win finally arrives, it recovers all previous losses plus a small profit.
The arithmetic works only with unlimited capital and no limits on position size. Real accounts have both.
A streak of seven losses in a row starting at one lot requires an eighth trade of 128 lots. The probability of seven consecutive losses at a 50 percent win rate is about 1 in 128 per sequence, so over a few hundred sequences it becomes almost certain.
Martingale produces beautiful equity curves for long periods, with steady small gains, then wipes out the account in one event. That shape explains why so many systems built on it look spectacular in demos and short track records.
Variations, such as grid systems and averaging down without a hard stop, share the same flaw. They exchange frequent small wins for rare catastrophic losses.
Has anyone run a martingale variant on a live account, and how did it end?
The arithmetic works only with unlimited capital and no limits on position size. Real accounts have both.
A streak of seven losses in a row starting at one lot requires an eighth trade of 128 lots. The probability of seven consecutive losses at a 50 percent win rate is about 1 in 128 per sequence, so over a few hundred sequences it becomes almost certain.
Martingale produces beautiful equity curves for long periods, with steady small gains, then wipes out the account in one event. That shape explains why so many systems built on it look spectacular in demos and short track records.
Variations, such as grid systems and averaging down without a hard stop, share the same flaw. They exchange frequent small wins for rare catastrophic losses.
Has anyone run a martingale variant on a live account, and how did it end?