Grid and martingale EAs: the margin math that blows prop-firm accounts
요약 — A grid or martingale EA hides risk until one losing streak ends the account. We show the lot-doubling math, the real probability of a long losing streak, a martingale vs percent-risk table, and a real 42.91% drawdown from our own backtests.
Grid and martingale EAs: the margin math that blows accounts
TL;DR — A grid or martingale EA does not reduce risk; it hides risk until one losing streak ends the account.
- After each loss the lot size doubles, so margin and open exposure grow exponentially — not linearly — with the length of the streak.
- On a 10,000 USD account, a 10-trade losing run already needs more margin than the account holds, before the losses themselves are counted.
- A real grid-style build in our own backtests (goldgridm1) shows 42.91% relative drawdown — far past every prop-firm 10% limit.
- Fixed percent-risk sizing shrinks positions as equity falls, which is why our published EAs use it instead.
What is a grid or martingale EA, and why do traders reach for one?
It is an EA that adds to a losing position instead of closing it, betting that the market eventually reverses and the one winning trade recovers every earlier loss plus a profit. The martingale variant doubles the lot after every loss; the grid variant layers fixed-size orders at price intervals. Both are popular because they show long stretches of small steady gains — the account curve looks calm right up until it does not.
The appeal is real and that is exactly the danger: the strategy converts a visible, bounded risk into an invisible, unbounded one.
Why does a martingale blow up an account? (the lot-doubling math)
Because the lot size after each loss is not plus-one — it is times-two. Start at 0.01 lots and the positions you are holding grow like this (margin shown is illustrative, at 1:100 leverage, roughly 10 USD of margin per 0.01 lot on a EURUSD-style pair):
| Consecutive losses | Lot on next trade | Cumulative lots held | Margin required (illustrative) |
|---|---|---|---|
| 1 | 0.01 | 0.01 | 10 USD |
| 2 | 0.02 | 0.03 | 30 USD |
| 3 | 0.04 | 0.07 | 70 USD |
| 4 | 0.08 | 0.15 | 150 USD |
| 5 | 0.16 | 0.31 | 310 USD |
| 6 | 0.32 | 0.63 | 630 USD |
| 7 | 0.64 | 1.27 | 1,270 USD |
| 8 | 1.28 | 2.55 | 2,550 USD |
| 9 | 2.56 | 5.11 | 5,110 USD |
| 10 | 5.12 | 10.23 | 10,230 USD |
By the tenth straight loss the margin alone exceeds a 10,000 USD account, and that figure does not yet include the actual losses piling up on every open trade. The curve is not "about to recover" — it has already run out of room to double one more time, and the broker closes the positions at a total loss.
What is the real probability of a long losing streak?
Higher than intuition suggests, because traders confuse "per-trade probability" with "probability over a year." A single run of five losses is unlikely on any one trade, but over hundreds of trades it becomes a near certainty. Using the per-trade chance of a loss as (1 minus win rate):
| Win rate | Chance of 5 straight losses | Expected 5-loss streaks per 1,000 trades |
|---|---|---|
| 55% | 1.85% | about 18 |
| 50% | 3.13% | about 31 |
| 45% | 5.03% | about 50 |
A strategy that wins half its trades will, on average, walk into more than thirty separate five-loss streaks across a thousand trades. The martingale does not need a freak event — it needs an ordinary Tuesday. The longer you run it, the more certain the blow-up becomes.
How is this different from fixed percent-risk sizing?
This is the comparison that matters for account survival. Our published EAs size each trade as a fixed percentage of current equity, so a loss makes the next position smaller, not larger.
| Dimension | Martingale or grid | Percent-risk (our EAs) |
|---|---|---|
| Lot size after a loss | Doubles each loss | Shrinks with equity |
| Drawdown shape | Can exceed the account before one win | Bounded near the per-trade risk percent |
| Margin demand | Grows exponentially with the streak | Grows only with account size |
| A 10-loss streak | Account is blown | Loss is near 10 times risk per trade, recoverable |
| Fit for a prop-firm 10% limit | Effectively incompatible | Built to stay inside the limit |
Percent-risk does not make a strategy profitable, but it keeps a bad day from becoming a closed account. We explain the sizing formula in our position sizing guide.
What does a real grid-style backtest actually look like?
Our own library includes a grid-style build, goldgridm1, backtested on XAUUSD M1 from 2021 to 2026 on a 10,000 USD simulated account. The headline numbers are seductive and the drawdown is the warning:
- Net profit: 31,110 USD
- Total trades: 62,575
- Profit factor: 1.69
- Relative drawdown: 42.91% (recovery gain needed: 75.16%)
That 42.91% relative drawdown is more than four times a typical prop-firm limit, and it means the equity fell to roughly 57% of its peak at some point in the test. On a funded account with a 10% cap, this curve would have been shut down long before the profit showed up. High net profit and low risk are not the same EA — we covered that split in our drawdown recovery math piece. Full reports for every build are on the proof page.
How do you keep a grid EA inside a drawdown limit?
You mostly cannot — and that is the honest answer. A grid's drawdown is decided by the widest adverse move the market ever makes while positions are layered, not by any setting you tune. The only reliable controls are to not run it on a limited account, to cap total exposure at the account level with an equity guardian such as PropGuard, and to read relative drawdown before anything else. You can line builds up on the compare page.
FAQ
Is a martingale EA ever safe to trade? Only with unlimited capital and no margin limits — which no retail or prop account has. The moment a margin call stops the next doubling, the streak ends the account. The sentence worth remembering: a martingale EA does not reduce risk — it hides risk until the one losing streak that ends the account.
Can I make it safe by starting with a tiny lot size? No. A smaller starting lot only delays the blow-up; it does not change the exponent. Doubling 0.001 lots instead of 0.01 pushes the ten-streak margin down to about 1,023 USD, but a thirteen-streak still exceeds it, and over enough trades a thirteen-streak arrives. You trade a near-certain small loss later for a guaranteed large one.
Does a backtest showing profit prove a grid EA is safe? No. A grid EA can show years of net profit and still carry a single worst-drawdown number that breaches every prop limit. Always read the relative drawdown, not the net profit, and plan for at least 1.5 times the tested figure live. Our risk disclosure explains why backtests are not forecasts.
Risk caveat: the margin and probability figures above are worked examples on a 10,000 USD simulated account at 1:100 leverage; the goldgridm1 metrics are taken from its MetaTrader 5 strategy-tester report in our backtest library. Backtest results are historical simulations and do not predict future performance. Trading leveraged products carries substantial risk of loss. Nothing here is financial advice — see our risk disclosure.