The short answer
Look for gold EA reviews and you find three kinds of them: vendor backtests, buyer ratings and live accounts. Only the third says anything about what a gold EA does with real money at a real broker, and only if it is clear who picked the account. A backtest on XAUUSD shows what the EA would have made with ideal fills. Live, spread, commission and the fills of triggered stops come on top, and on gold every 0.01 USD of price difference costs 1 USD per lot.
Of the 48 gold EAs we review, 16 have a live account on record: an account of our own in our EA park, or at least one published by the vendor or by ATS Ad that is still listed. We do not print returns or drawdowns on this page; what each account shows is on the review page.
Disclosure: AlgoTradingSpace (ATS) and the brokers IC Markets Ad, RoboForex Ad, BlackBull Ad and Eightcap Ad that appear on this page are partners of ours. If you sign up through a partner link, we earn a commission. It changes no grade and no measurement.
Three kinds of review and what they are worth
- The vendor’s backtest. Does not count for our grade. It is tailored to historical data, often with a fixed spread and no commission, and the vendor picks the period and the settings. A backtest is a tool for you to re-run the EA on your broker’s data, not evidence.
- Buyer ratings. Say little about survival. On the MQL5 Market buyers mostly review in the first weeks, the phase in which a gold grid in particular almost always delivers. The Quantum Queen review shows this for an EA with a very large number of top ratings that had not yet lived through an extended gold downturn.
- Live accounts. Count, but not all equally. An account in our own EA park counts most, because we know the broker, the start and the settings. An account published by the vendor or by ATS counts; screenshots and screen recordings from videos do not. The rules are on the methodology page.
Why gold backtests flatter
Spread and commission. One lot of gold is 100 ounces. The average gold spreads the brokers in our gold comparison state for their raw accounts range from 0.05 to 0.12 USD, i.e. 5 to 12 USD per lot, plus commission (as of 1 October 2026). Around US data the spread widens. A backtest is only as realistic as the spread and commission it runs with: with a fixed, tight spread and no commission it underestimates the cost of every trade, and for a scalper with a small target that is quickly the whole difference between profit and loss. The cost per lot is worked through in gold scalping EAs.
Triggered stops. The strategy tester usually fills a stop-loss exactly at its level. Live, the broker fills it at the next available price, and on gold that can be well off in fast phases. In our evaluation of real MT5 accounts from the public ATS tracker, a triggered gold stop cost a median of 6.00 USD per lot at IC Markets, 6.00 USD per lot at RoboForex and 15.00 USD per lot at BlackBull (as of 2 November 2026). By default the strategy tester does not account for that. Method and limits are in the best broker for gold EAs.
A period the EA will not see again. Gold moves in long one-directional legs. A backtest optimised on the trend legs of recent years shows how the EA would have traded exactly those years. Even a long live account does not cover every market: the Gold Reaper’s track record is long, but not yet a full gold cycle.
The balance hides the floating loss. With grid and martingale EAs the balance curve rises as long as baskets eventually close in profit. What the basket loses in between shows only in the equity. How to read it is explained in the equity curve explained; what a gold basket costs per dollar of move in gold EA risk.
What our tests show
Four cases from our reviews, each with the date of the review page:
- Same EA, two brokers. ATS runs the scalper mode of Happy Gold EA live at Eightcap, and the account is clearly in profit, though with a deep realised drawdown (ATS statistics, as of 30 Sep 2026). An account with the same mode at RoboForex that we track ourselves, and that ran in a different time window, is in loss, with a deep drawdown of its own (as of 22 Sep 2026, a snapshot). Broker and time window both differ, and either can shift the result.
- Signal and own account. We verified the public live signal of The Gold Reaper. On our own account at IC Markets, on 6 October 2026 all eight sub-strategies opened a short within three seconds and closed it at a loss; afterwards the account sat deeper below its high than the signal had ever fallen up to 30 September. On top come reports from signal copiers about worse fills on fast gold moves.
- Fixed stop, deep decline anyway. Gold Scalper Pro sets a stop-loss and take-profit on every trade. The published ATS account still showed a very deep decline because many small losses added up, and the edge after costs was practically zero (ATS live statistics, as of 5 Oct 2026). The account is no longer listed.
- Promises without an account. For Gold X9 the vendor publishes no verifiable live signal, only screen recordings in a video (as of 5 Oct 2026). For Gold Matrix EA the marketing promised a very high win rate without evidence; the trade log of our account showed a grid adding positions against the move instead of the advertised momentum strategy. We took the account out of our tracker in September 2026.
These EAs are graded on the same criteria as every EA; the cases above explain why evidence carries so much weight. All gold EAs compared are in best gold EA 2026.
How to check a gold EA yourself
- Backtest on your broker’s data. Download the demo version if there is one and run it in the MT5 strategy tester on your broker’s data, with a gold spread like on your account and the commission per lot. If the result differs strongly from the vendor’s backtest, that is your first answer. The basics are in backtesting and forward testing.
- Read the live evidence, do not just look at it. Who picked the account, which broker does it run at, since when, and do other EAs run on it? An account at a broker you will never use says less than one at yours.
- Equity, not balance. For every gold EA, ask for the equity curve and the largest floating loss first, not for profit or win rate.
- Forward test on demo. Several weeks at the broker you plan to use, on the VPS you plan to use and with the settings you want to run live. A VPS close to the trading server is part of it, because every delay on gold stops costs money.
- Start small live. With half the suggested lot size or less, and let your own live drawdown decide whether you raise it. The formula is in position sizing and risk per trade.
Why EAs generally do worse live than in testing, gold or not, is explained in why EAs fail live. If you want to run a gold EA on a prop firm account, the rules and the daily-limit maths are in gold EAs at prop firms.
Risk warning: trading CFDs on margin carries a high risk of loss, on gold especially because of the contract size of 100 ounces per lot. Backtests are simulations on historical data, and an EA’s live results do not guarantee future ones either. Test on a demo account first and only put in money you can afford to lose.
AdLinks marked "Ad" are partner links: if you open an account or sign up through one, we receive a commission, at no extra cost to you. It does not change our ratings or what we write.