0
Before you start
Two checks. Tick them as you go.
Words you'll see. Backtest = running the EA over past prices in the Strategy Tester. Drawdown = the fall from a peak before the account recovers, if it does.
Guide · Learn · MetaTrader 5 · about 20 minutes
How to check a gold EA before you run it.Seven checks, before any real money.
Written by Dan Freeman · last checked
Gold (XAUUSD) Expert Advisors are some of the most advertised on MetaTrader 5. This guide gives you seven checks to run on any of them before it touches a real account. It names no product and recommends none. A good backtest is not a promise of future results.
~20 minstart to finish
7 stepstick each one off
Any gold EAno product named
Learningnothing to install
✓
Nothing here places a trade.
Every check is something you read or test on a demo account. Keep the EA off any real account until it passes them all.
1
Know what it actually does
The strategy, in plain words.
- Find a plain description of how it trades: what makes it enter, where the stop goes, how it exits.
- Check it always uses a stop loss. An EA that trades without one can lose far more than you expect on one move.
- Vague words like "AI", "neural" or "secret algorithm" tell you nothing about the risk.
If you can't say in one sentence how it decides to trade, you can't judge its risk yet.
Not seeing it?
The listing doesn't explain the strategy. Ask the seller. A clear answer is a good sign; no answer is a reason to wait.
2
Martingale, grid and averaging
The three that hide risk.
- Martingale: after a loss it trades bigger, to win the loss back.
- Grid: it opens more trades at set distances as price moves.
- Averaging: it adds to a losing trade to improve the average price.
- All three can show many small wins in a row, then one loss big enough to wipe the account out. A short test may never reach that day.
Watch the open trades in the Strategy Tester's visual mode. Several trades on one side at once, or lot sizes growing after losses, are the tell.
Not seeing it?
The listing says 'no martingale'. Check it yourself in the tester. Look at the lot size after a loss and how many trades it holds at once.
3
Read the drawdown properly
Equity, not just balance.
- Balance drawdown counts closed trades only. Equity drawdown includes the open loss while trades are running.
- A grid or averaging EA can show a small balance drawdown and a very large equity drawdown, because it holds losers open until they come back.
- In the Strategy Tester report, read the maximal equity drawdown and ask: could my account survive that, or more?
The worst drawdown in a backtest is not the worst that can happen. It is only the worst that happened in that test.
Not seeing it?
The listing only shows balance growth. Ask for the equity curve, or run the test yourself and read the equity figures.
4
Spread and costs on gold
Small numbers that change everything.
- Gold's spread widens around news and the daily rollover. An EA with small targets can lose most of its edge to the spread alone.
- Test with Every tick based on real ticks in the Strategy Tester, on your own broker's data, so the spread is the one you would really pay.
- Add commission and swap. Hold-overnight systems pay swap every night.
Run the same test on two brokers. If the result changes a lot, the EA is living on the spread.
Not seeing it?
Real ticks aren't available. Your broker may not provide tick history for that period. Test a shorter recent period with real ticks instead.
5
In-sample and out-of-sample
Was it tuned to the past?
- An EA's settings are usually chosen on one stretch of history: the in-sample period. It will look good there by design.
- The honest test is a period it was not tuned on: the out-of-sample period.
- Ask which dates were used to choose the settings, then test a different period yourself.
A smooth curve on the tuning period and a poor one elsewhere means the settings fit the past, not the market.
Not seeing it?
The seller won't say which dates were used. Then treat every backtest you are shown as in-sample.
6
Forward test it yourself
Demo first, then small.
- Run it on a demo account with your broker for several weeks, ideally through news days and a quiet spell.
- Compare the demo trades with what the backtest did over the same days. Big differences need an answer.
- If you go live, start at the smallest size and only grow it if the live results look like the demo.
A live track record on MQL5 Signals shows real trades over time. Look at how long it has run and how many trades it holds, not only the growth figure.
Not seeing it?
It hasn't traded much on demo. Some systems trade rarely. Wait for enough trades to judge, rather than going live early.
7
Red flags in a listing
Any one of these is a reason to wait.
- Promises of guaranteed returns, "no losing months" or "safe" profits.
- A very high win rate with no drawdown shown, or only backtests and no live record.
- A live record of only a few weeks, or one that started just before the listing went up.
- Pressure to buy now: a price that "goes up after the next few copies" or a countdown.
- Results you can't check yourself, such as screenshots instead of a linked live account.
None of these prove an EA is bad. Each one is a reason to slow down and test more.
Not seeing it?
The listing looks clean. Good. Still run every check above on a demo before a real account.
Your safety nets
Three things to keep in mind.
1
No test is a promise
Backtests and demo results show the past. They do not tell you what will happen next.
2
Size for the worst day
Size so the worst drawdown you could face would not end your account, and keep a margin.
3
Protect the account
A drawdown limit outside the EA, such as the free PFX Equity Shield, stops the whole account at your limit, whatever the EA does.
!
Prop-firm accounts
Many prop firms restrict EAs, grids, martingale or holding over news. Read your firm's rules before you run any EA on a challenge or funded account.
All done
You know the checks.Now test before you trust.
Risk warning: Trading carries a high risk of loss and is not suitable for everyone. This guide is for learning and names no product. Expert Advisors can lose money, and past or simulated results do not guarantee future results. MetaTrader and MQL5 are trademarks of MetaQuotes Ltd.; PrecisionFX is not affiliated with or endorsed by MetaQuotes.