MidasA grid EA blows a gold account when price keeps going one way without a pullback: every new order is twice the previous one, so the floating loss grows much faster than the move, and with $1,000, lot 0.01 and 11 orders the stop-out comes only $65.58 below the first order.
Arithmetic of settings, not a forecast and not advice. Trading with leverage carries risk: if the deposit is too small, a stop-out is possible.
See also: 7 protections for a martingale grid bot on gold →
On this page (7)
How does a grid EA lose the whole deposit?
A grid EA of the martingale type opens a first order and, if price goes against it, adds a new order every fixed step, each one bigger than the last. When the account is lost, four things happen in this order:
- Price moves without a pullback. As long as gold turns back from time to time, the basket closes in profit near its last order. The danger is a move that does not turn back in time.
- The lot doubles. With a multiplier of ×2 the orders are 0.01, 0.02, 0.04 and so on. The 11th order is 10.24 lots, half of the whole basket.
- The floating loss outruns the deposit. On a cent account every $1 against the basket costs $0.01 for each 0.01 lot open, and the number of lots doubles with each order. The first $19.50 of the move costs $39.70; the next $46.08 costs about $943.
- Stop-out. When equity falls to the broker's stop-out level (40% of margin in our example), the broker closes the positions at market. The loss becomes real and the account is close to zero.
This is not a malfunction. It is how a doubling grid is built. The only question is how far the move has to go.
Worked example: $1,000, lot 0.01, ×2, 11 orders
The inputs are the defaults of the grid calculator: cent account, deposit $1,000, first lot 0.01, multiplier ×2, step $1.95, 11 orders, gold at $4,170, leverage 1:2000, stop-out 40%. A buy basket is shown; a sell basket is the mirror image. Spread, swap and commission are left out.
| Order | Lot | Lots in the basket | Price below the first order | Floating loss | Loss per further $1 |
|---|---|---|---|---|---|
| 1 | 0.01 | 0.01 | $0.00 | $0.00 | $0.01 |
| 2 | 0.02 | 0.03 | $1.95 | −$0.02 | $0.03 |
| 3 | 0.04 | 0.07 | $3.90 | −$0.08 | $0.07 |
| 4 | 0.08 | 0.15 | $5.85 | −$0.21 | $0.15 |
| 5 | 0.16 | 0.31 | $7.80 | −$0.51 | $0.31 |
| 6 | 0.32 | 0.63 | $9.75 | −$1.11 | $0.63 |
| 7 | 0.64 | 1.27 | $11.70 | −$2.34 | $1.27 |
| 8 | 1.28 | 2.55 | $13.65 | −$4.82 | $2.55 |
| 9 | 2.56 | 5.11 | $15.60 | −$9.79 | $5.11 |
| 10 | 5.12 | 10.23 | $17.55 | −$19.75 | $10.23 |
| 11 | 10.24 | 20.47 | $19.50 | −$39.70 | $20.47 |
| Stop-out | — | 20.47 | $65.58 | about −$983 | — |
Read it from the bottom. All 11 orders are open after a move of only $19.50, and the account has lost 4% of the deposit. From there the basket loses $20.47 for every further $1, and the stop-out comes $46.08 later, $65.58 below the first order. The last $46 of the move do 96% of the damage. The stop-out arithmetic article walks through the same formulas step by step.
Is a $65 move a lot for gold?
No. Our XAUUSD tick data (Dukascopy bid, UTC days, Monday to Friday) on the gold daily range page:
| Gold, one day from high to low | 2025 | 2026 (2 Jan – 22 Oct) |
|---|---|---|
| Average day | $61 | $122 |
| Median day | $52 | $101 |
A basket does not need the whole range, only a move in one direction. Measured from the day's opening price (gold move odds, 462 trading days from 2 Jan 2025 to 26 Oct 2026), gold went at least $50 one way on 55.4% of days and at least $75 on 31.4% of days. In the last three months the shares were 75.0% and 41.7%.
So $65.58 of room is about half of an average 2026 day. A basket also does not have to start at the open: it can start anywhere and carry into the next day. Past ranges do not predict future ones, but they show that this buffer sits inside an ordinary day, not a rare crash.
How do I calculate the reserve: 9 orders against 11?
The reserve is the deposit at which the stop-out comes only after a given move against the first order. Same settings as above (first lot 0.01, ×2, step $1.95, 1:2000, stop-out 40%); only the number of orders changes. For a different first lot, scale in proportion: lot 0.02 needs twice the deposit.
| Move against the first order | 9 orders (5.11 lots at the bottom) | 10 orders (10.23 lots) | 11 orders (20.47 lots) |
|---|---|---|---|
| $65.58 | $269 | $520 | $1,000 |
| $100.50 (median day, 2026) | $448 | $877 | $1,715 |
| $122.20 (average day, 2026) | $559 | $1,099 | $2,159 |
| $152.45 (one day in ten, 2025–2026) | $713 | $1,408 | $2,778 |
| $200 | $956 | $1,895 | $3,752 |
The rule behind the table: once every order is open, each extra $1 of room costs as many dollars as the basket has lots. That is $20.47 with 11 orders and $5.11 with 9 orders, for a first lot of 0.01 on a cent account.
Read the other way: on the same $1,000, a cap of 10 orders moves the stop-out from $65.58 to $112.54 below the first order, and a cap of 9 orders to $208.55.
The cap is not free. After the last allowed order the basket stops averaging, so in a long move it needs a bigger pullback to close and can stay in drawdown longer. A lower cap trades some exits for a much later stop-out.
What actually helps, and what only looks smart
What changes the outcome:
- Deposit per 0.01 lot. The one lever that buys room without changing the strategy. Count it per 0.01 of first lot, not as a total.
- A cap on the number of orders. Each order fewer halves the basket at the bottom. It costs slower recovery, not money up front.
- A hard equity stop or a lock, knowing they cost money. A stop closes the basket at a loss you chose instead of a stop-out at a loss the broker chose. A lock (an opposite order) freezes the floating loss but does not remove it. In our tests on gold ticks both cut the worst outcomes and also took a noticeable part of the income. They limit damage; they do not make a grid safe.
What does not change the arithmetic:
- Time, session and news filters, and "smart" entry filters. A basket is lost because of how far price goes, not because of the hour the first order opened. In our tests, filters by session, night hours and news days did not reduce how often a deep basket reached the stop-out. More in entry filters for a grid EA.
- A cent account by itself. It makes lots 100 times cheaper in dollars, so a test costs less, but the percentages and the stop-out distance are the same as on a standard account 100 times larger (what is a cent account).
Check your own settings
- The grid calculator shows the stop-out distance and the deposit for your lot, multiplier, step and number of orders.
- The martingale lot size calculator works the other way round: which first lot, number of orders or step fits a deposit and the move you want to survive.
- The cent account calculator converts cents to dollars and shows what a lot is worth.
- The gold daily range page shows how far gold has moved in a day, month by month.
- The Midas Grid emulator replays your settings on real gold ticks since January 2025, day by day. Free, in the browser.
FAQ
Why did my grid EA blow my account on gold?
Because price moved further in one direction than the deposit could carry the doubled lots. With lot 0.01, ×2 and 11 orders on $1,000 (cent account, 1:2000, stop-out 40%), all orders are open after $19.50 and the stop-out comes $65.58 below the first order, about half of an average gold day in 2026 ($122).
How much deposit does a gold grid EA need?
It depends on the first lot, the multiplier, the number of orders, the step and the move you want to survive. With 0.01, ×2 and a $1.95 step, carrying a $100.50 move (the median 2026 day) takes about $1,715 with 11 orders and $448 with 9 orders. No deposit rules out a stop-out.
How can I stop martingale drawdown on gold?
You cannot remove it, only limit it: more deposit per 0.01 lot, fewer orders, or a hard equity stop or lock that turns part of the floating loss into a realised one. Each of these has a price: money up front, slower recovery or lower income.
Do news or time filters protect a grid EA?
Not in our tests. They change when the first order opens, not how far price can run against the basket, and it is the depth of the move that decides the stop-out.
Is a grid EA safer on a cent account?
The risk in percent is the same. A cent account makes lots 100 times cheaper in dollars, so $1,000 behaves like $100,000 on a standard account with the same lots. It lowers the cost of a test, not the risk.