Dennis Hoinkis / Profil

Worked example from my article "How a Gold Grid Works – and Why It Is Dangerous" (post below): five buys every 1.50 USD lower, one take profit for the whole basket 1.00 USD above the average price. A pullback to 3,998.00 closes all five trades together in profit. If price keeps falling instead, every level adds exposure – that is where the deep drawdown of a grid comes from.

Not investment advice. Trading grid systems can lead to a total loss.

How a Gold Grid Works – and Why It Is Dangerous

Grid Expert Advisors for gold are among the most popular bots on MetaTrader 5. Their equity curves often look impressive: up month after month, with a win rate well above 70%. That is exactly what makes them dangerous if you don't understand what is happening under the hood. In this article I explain step by step how a gold grid works, why it wins so often and why it can still wipe out an account at some point. As an example I use the logic of our own grid EA and its three settings (LOW, MEDIUM, HIGH), with real backtest numbers – including the bad ones.

WHAT IS A GRID EA?

A grid EA does not just open one position. When needed, it builds a whole basket of positions. If price moves against the first trade, the EA adds more trades at fixed distances. All positions belong together and are closed together. So the EA is not trying to get every single trade right. It is betting that price will eventually move back a little.

THE ENTRY: MOMENTUM AND TREND FILTER

Before a basket even starts, the market has to meet two conditions. The EA in this example trades XAUUSD (gold against the US dollar) only.

- Trend filter: a moving average (MA) 50 on M15 and an MA 20 on M5, plus a buffer of 0.05%. The EA only trades in the direction of the trend.
- Momentum: price must have moved at least 0.3% within no more than 71 M1 candles. Only then does the EA open the first trade.
- Pauses: after an entry there is a pause of 10 candles, after a closed basket it is 17 candles.

So the entry is not random. But even the best filter cannot predict whether price will keep going or turn around afterwards. And that is where the grid begins.

ADDING TRADES AT FIXED DISTANCES

If price moves against the basket, the EA adds another trade – every 150 points (1.50 USD in the gold price) on the LOW and MEDIUM settings, every 200 points on the HIGH setting. The number of levels is limited: at most 14 trades on LOW and at most 18 on MEDIUM and HIGH.

A WORKED EXAMPLE (see first image)

There is only one take profit for the entire basket: 100 points (1.00 USD) above the average price of all positions (below it for a sell basket). Let's say the EA buys gold at 4,000.00 USD with the same lot size on every level, and price falls:

Level - Buy price - Average price - Basket take profit
1 - 4,000.00 - 4,000.00 - 4,001.00
2 - 3,998.50 - 3,999.25 - 4,000.25
3 - 3,997.00 - 3,998.50 - 3,999.50
4 - 3,995.50 - 3,997.75 - 3,998.75
5 - 3,994.00 - 3,997.00 - 3,998.00

With every level the average price drops, and so does the target. After five levels price is at 3,994.00. For the whole basket to close in profit, gold only has to get back to 3,998.00 – a rise of 4 USD. The first trade at 4,000.00 even closes with a small loss, but the lower buys more than make up for it. The basket as a whole wins.

WHY THE WIN RATE IS SO HIGH

Gold moves constantly. A pullback of a few dollars almost always comes at some point. That is why most baskets close in profit. In our backtests (MT5 Strategy Tester, "every tick", XAUUSD, 01 Jan 2024 to 31 Oct 2026, starting with 3,000 USD, fixed 0.01 lot) it looks like this:

Setting - Final balance (backtest) - Max. equity drawdown - Profit factor - Trades / won
LOW - 17,767 USD - 52% - 1.54 - 13,488 / 77%
MEDIUM - 23,948 USD - 25% - 1.88 - 13,825 / 76%

76 to 77% winning trades sounds like an almost perfect system. But look at the column next to it: a maximum drawdown of 25% and 52% in the backtest. The high win rate is not a sign of safety. It is a consequence of how a grid is built.

WHY THE DRAWDOWN IS STRUCTURALLY HIGH

A grid does not take losses, it sits them out. Instead of closing a wrong trade with a stop loss, it adds more and waits. This has two consequences:

- Exposure grows with every level. At level 1 one position is open, at level 10 there are ten. Every further dollar against the basket now costs ten times as much.
- Profits are small, open losses can get large. A basket only earns 1 USD above its average price. A basket stuck deep in the red, on the other hand, puts heavy pressure on the account for as long as it stays open.

So the drawdown is not an accident, it is built in. Many small wins are paid for with rare but large hits.

WHEN ALL LEVELS ARE USED UP

It gets dangerous when price does not turn. On MEDIUM there are 17 gaps of 1.50 USD between the first and the 18th trade, which is 25.50 USD. On HIGH it is 17 x 2.00 = 34 USD. For gold, that is not an extreme move. Once the maximum number of trades is reached, the EA stops adding. If price keeps running, every open position loses with every dollar at the same time. The drawdown gets very deep. When the margin is no longer enough, the margin call comes, and finally the broker's stop-out. On aggressive settings that means losing the entire account.

EXAMPLE: THE HIGH SETTING AS A BACKTEST CHAIN (see second image)

How often this can happen is shown by our backtest of the HIGH setting. HIGH grows the lot size with the account (11% risk per grid level) and has no brakes. We ran the backtest as a chain: after every crash, a new backtest starts one day later with 3,000 USD again.

- 5 crashes in the backtest between 2024 and 2026: 18 Oct 2024, 13 May 2025, 23 Jul 2025, 04 Feb 2026 and 24 Mar 2026.
- Deposited: 18,000 USD (six starts of 3,000 USD each).
- Final balance: 73,147 USD, a net result of +55,147 USD.
- The entire gain comes from the last backtest run from 25 Mar 2026: it grew from 3,000 to 73,147 USD. Until then all five runs had been lost; on 24 Mar 2026 the chain stood at -15,000 USD.

The end result is a profit. But only because fresh money was added after every total loss and the last run happened to fall into the boom from March 2026. Anyone who had stopped after the second or third crash would simply have lost money. And nobody knows in advance which run will be the good one.

WHAT PROTECTIONS EXIST

A grid is not completely defenceless. On LOW and MEDIUM, the EA uses several brakes:

- Max trades: 14 on LOW, 18 on MEDIUM. The basket does not grow forever.
- Speed brake: if a basket reaches level 5 within 5 minutes, it is closed immediately. Moves that fast can be the start of a strong breakout.
- Time brake: if a basket has been open for more than 3 days, it is closed.
- Margin guard: no new order if the margin level would fall below 200% after the order. The basket must also be able to survive a 3% price gap. If the margin level drops below 100%, the EA closes the basket itself – before the broker does.

These brakes close baskets at a loss. That hurts, but at best it prevents a total loss. The HIGH setting deliberately has none of these brakes and no margin guard, and it can only be activated after explicit confirmation in the EA. It is built for maximum growth – with the full crash risk.

WHAT YOU SHOULD TAKE AWAY

1. Enough capital: as a rule of thumb, plan 1 to 1.5 times the gold price as capital per 0.01 lot. With gold at 4,000 USD, that is 4,000 to 6,000 USD. With less, the risk rises significantly. A cent account is an alternative for getting to know a grid EA.
2. Only money you don't need: never use money you need for rent, bills or savings.
3. Expect the crash: on aggressive settings a total loss is not the exception but part of the plan. Plan for it instead of hoping it won't happen.
4. Don't be dazzled by the win rate: always look at the maximum drawdown and at what happens when all levels are used up.

Not investment advice. Trading with leverage and grid systems can lead to a total loss. Backtests and past results are no guarantee of future performance.

MQL5.community'e kayıt oldu