MidasThere is no acceptable drawdown percentage for a grid EA, because its drawdown is not linear: with $1,000, lot 0.01, ×2 and 11 orders the floating loss is only 2% of the deposit when the 10th order opens, yet the stop-out comes $48 later — so size the move the deposit must survive (dollars of gold per 0.01 lot), not a drawdown %.
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 (6)
Drawdown terms in one place
Drawdown is how far the account falls from its last high. For a grid EA four different numbers hide behind that word:
| Term | What it measures | Where you see it |
|---|---|---|
| Floating drawdown | Loss of positions still open (equity below balance) | MT5 "Equity", trade tab |
| Realized drawdown | Losses already closed and booked to the balance | Account history, balance curve |
| Equity drawdown | Fall of equity from its peak, open and closed together | Backtest report "Equity drawdown maximal" |
| Distance to stop-out | How far gold can still go before the broker closes everything | Grid calculator |
A grid closes most baskets in profit, so its realized drawdown is usually small and its balance curve looks smooth. The risk lives in the floating part: a basket that is open and deep. The broker's stop-out does not look at the balance; it looks at equity against margin.
Why a grid's drawdown is not linear
Here is one basket of the default grid, $1,000, first lot 0.01, ×2, step $1.95, 11 orders, as gold moves against the first order:
| Gold against the first order | Orders open | Total lots | Floating loss | % of $1,000 |
|---|---|---|---|---|
| $5.85 | 4 | 0.15 | −$0.21 | 0.02% |
| $9.75 | 6 | 0.63 | −$1.11 | 0.11% |
| $13.65 | 8 | 2.55 | −$4.82 | 0.48% |
| $17.55 | 10 | 10.23 | −$19.75 | 1.98% |
| $19.50 | 11 | 20.47 | −$39.70 | 3.97% |
| $30.00 | 11 | 20.47 | −$254.64 | 25.46% |
| $40.00 | 11 | 20.47 | −$459.34 | 45.93% |
| $50.00 | 11 | 20.47 | −$664.04 | 66.40% |
| $60.00 | 11 | 20.47 | −$868.74 | 86.87% |
| $65.58 | 11 | 20.47 | −$982.93 | stop-out |
For the first $19.50 the drawdown stays under 4%. After the last order, every further $1 costs $20.47, and the remaining 96% goes in $46. A median gold day in our data ranged $70.90 from low to high (one day in ten more than $152.45), so the gap between "2% drawdown" and "account gone" is less than one ordinary day's range — if that range runs one way without a pullback.
Why "max drawdown %" in a backtest misleads for grids
A backtest report shows the deepest drawdown that happened in the test. For a grid that is the deepest basket the test met, not the risk of the settings.
- It measures the past path, not the ladder. If the deepest basket in the test reached 10 orders, the report shows about 2% for the default grid. The same settings are $48 of gold away from stop-out.
- It is near zero right up to the end. The table above goes 0.5% → 2% → 4% → 100% within $52. A test that missed one bad hour reports a small number; a test that met it reports the end of the account. There is little in between.
- It depends on data and costs. Missing hours or a too-small spread hide exactly the deep baskets (why a profitable backtest blows a real account).
Read the backtest's drawdown as "the deepest basket we happened to see", and calculate the stop-out distance separately.
What to size instead: the move per 0.01 lot
The useful question is: how many dollars can gold go against the first order before stop-out? It depends on the deposit behind each 0.01 lot and on the number of orders (the total lots of the full basket).
Stop-out distance below the first order, lot 0.01, ×2, step $1.95:
| Deposit | 9 orders (5.11 lots) | 10 orders (10.23 lots) | 11 orders (20.47 lots) |
|---|---|---|---|
| $500 | $110.70 | $63.66 | $41.15 |
| $1,000 | $208.55 | $112.54 | $65.58 |
| $2,000 | $404.24 | $210.29 | $114.43 |
| $3,000 | $599.93 | $308.04 | $163.28 |
| $5,000 | $991.32 | $503.54 | $260.99 |
The other way round — the deposit that survives a given move:
| Move to survive | 9 orders | 10 orders | 11 orders |
|---|---|---|---|
| $30 | $87.63 | $155.65 | $271.71 |
| $50 | $189.83 | $360.25 | $681.11 |
| $100 | $445.33 | $871.75 | $1,704.61 |
| $150 | $700.83 | $1,383.25 | $2,728.11 |
| $200 | $956.33 | $1,894.75 | $3,751.61 |
The formula behind both: after the last order, the deposit must cover the floating loss of the full basket at the bottom plus the stop-out share of its margin. Each extra $1 of move costs the basket's total lots in dollars on a cent account (×100 on a standard one), which is why one order more roughly doubles the deposit needed (basket tonnage).
Scale is linear: a first lot of 0.02 needs twice these deposits, 0.05 five times.
So what drawdown should I accept?
Turn the question into three numbers you choose:
- The move to survive, in dollars of gold against the first order. Look at real moves: the biggest 1-hour gold moves and daily ranges.
- The deposit per 0.01 lot that survives it with your number of orders, from the tables above or the martingale lot size calculator.
- The realized loss you accept if you add a basket stop: for example 10% of the balance per basket. In our model tight basket stops fired 6–15 times a week and about 95–98% of the stopped baskets would have recovered, so a stop turns rare stop-outs into frequent small losses. That is a choice, not a free protection.
In our model on 22 months of gold ticks, at $1,050 per 0.01 and step $1.95, the number of account blow-ups was 6 with 9 orders, 9 with 10 and 10 with 11. The deposit per 0.01 lot remained the strongest single lever. A model is not a guarantee: live baskets can go deeper than modelled ones.
FAQ
What is a good maximum drawdown for a grid EA?
There is no good percentage, because a grid's drawdown jumps from a few percent to the whole deposit within a short move. Decide instead how far gold may go against the first order before stop-out and size the deposit per 0.01 lot for it.
What is the difference between floating and realized drawdown?
Floating drawdown is the loss of positions that are still open; realized drawdown is losses already closed. A grid usually shows small realized drawdown and large floating swings, and the stop-out is triggered by the floating part.
Why did my backtest show 5% drawdown and the live account still blew up?
Because the backtest's maximum is the deepest basket it happened to meet. One basket a few orders deeper, or a move a few dollars longer, takes the same settings from single-digit drawdown to stop-out: in the default example from 2% to the end of the account in $48.
Is drawdown in percent or in dollars more useful for a grid?
In dollars of gold price. "$65.58 to stop-out" can be compared with real gold moves; "4% drawdown" cannot, because it says nothing about the next $10.
How much deposit do I need per 0.01 lot?
It depends on the move you want to survive and the number of orders: to survive $100 against the first order with ×2 and step $1.95, about $445 with 9 orders, $872 with 10 and $1,705 with 11, before spread and swap.