Veteran Army Gold review: independent test of the 3 presets on MT5

Veteran Army Gold is an MT5 expert advisor released on 1 August 2026 on the MQL5 Market by Seckin Erkut. At the time of the test it sells for 899 dollars, or 129 dollars a month to rent, and the page shows version 1.13, updated on 3 August. It trades gold only, XAUUSD, on H1 and H4, and requires a hedging account.

The idea behind it fits in one sentence: you do not pick a return, you first pick the loss you accept. There are three risk presets, Conservative (the default), Moderate and Aggressive, each with a drawdown range from a Monte Carlo simulation and a historical annual return from his backtest. All three presets take the same trades; only the position size changes. He says so himself, and it can be checked.

The page makes several precise, therefore testable, claims: no grid, no martingale, no averaging down; a fixed stop on every trade from the moment it opens; about a hundred trades a month; and, over twenty years of testing, never more than 26 open positions at once. At startup, the robot logs what it loads: 66 strategies out of 66, a single symbol, a hedging account. It is a portfolio of strategies on one market, not a single strategy.

One input matters more than the others: ReferenceCapital, set to 20,000. Published lot sizes apply to a 20,000 dollar balance and are scaled to the real balance. Section 8 explains why that is where a buyer's real risk sits.

One detail of method matters. At IC Markets, real gold ticks only go back to April 2026. Over three and a half years, a "real ticks" backtest would be mostly generated by the terminal without saying so. The long window therefore runs in 1 minute OHLC mode, and real ticks serve as a control from April to September 2026, where both modes exist.

That is also the mode of his gallery. His Conservative gallery report shows 5,067,247 ticks for 21,180 bars, about 239 ticks per one hour candle. That is the exact signature of 1 minute OHLC: four prices per one minute candle, sixty minutes per hour, 240. Real ticks give tens of thousands of ticks per bar (31,853 in our June 2026 control). My report shows 5,062,735 ticks for 21,157 bars: the same modelling, within 0.1%. The product page describes "every tick based on real ticks" modelling; the gallery report does not show that. Section 7 measures what the difference changes.

If these modelling modes are new to you, our guide on how to backtest an EA in MetaTrader 5 explains them one by one.

First check: the Conservative preset, over his window, with his 10,000 dollar starting balance. Here is his report, as published in the product gallery, then mine.

Fifty dollars apart out of almost thirty thousand, at another broker with a different price history: that is the strongest result of this test. A backtest that reproduces elsewhere guarantees nothing about the future, but it shows the published figures do not depend on one particular price feed or a hidden setting. On Moderate and Aggressive I get about 10% less profit and a slightly deeper drawdown, nothing that changes the meaning of the result.

Over his three and a half year window, none of the three presets exceeds its range. The ranges come from a simulation, and the page itself warns that future drawdowns can exceed them: one backtest path staying inside is consistent with the range, not proof it will never be crossed. And, as section 8 shows, these numbers only hold at the intended capital.

I compared the trade lists of the three presets entry by entry: same time, same direction, same price, on 4,353 trades out of 4,353. Only the size changes.

In the orders table of the MT5 report, every entry carries a stop at the broker from the moment it opens: 4,353 entries out of 4,353. The only orders without a stop are the closing ones. You can also see how it works: it places stop and limit orders at price levels, hour after hour, and removes them if price does not come. In June 2026 alone, 1,452 buy stop orders were placed and cancelled that way.

His page raises the question itself: gold is expensive to hold, and the swap is heavily one-sided. Positions last just over a day on average, so the swap is paid or received most nights.

The swap is indeed one-sided: buys pay, sells receive. His page says holding a buy costs several times what holding a sell pays; at IC Markets, over this window, buys pay about a third more than sells earn. Over his window, costs do not eat the edge. One limit: the MT5 tester applies today's swap to the whole history, and every broker has its own swap and commission table. At yours, this weight may differ.

From April to September 2026, when real ticks exist at IC Markets, I ran the three presets twice on 10,000 dollars: on real ticks, then on 1 minute OHLC.

Same month by month profile, same sign every month. But real ticks give 15 to 18% less profit and a drawdown about one and a half points deeper. One minute mode flatters a little; it reverses nothing.

Also note what the window flatters in his favour: over his 2023 to 2026 window, my Conservative run makes about 47% a year against the 23% he states, because 2025 alone brought +77.7%. His page does not use that number. It shows the lower one and warns that a quiet year for gold can end flat. That is the right reading. To spot the year that carries a whole backtest, see how to read a backtest report in MT5.

The robot's lots are sized for 20,000 dollars and scaled down on a smaller account. At startup the log says it plainly: balance factor 0.50x on 10,000 dollars, 1.00x on 20,000. But a lot never goes below the broker's minimum of 0.01. On a small account, the Conservative lot falls under that floor and is rounded up to 0.01.

From April to September 2026, on real ticks, I ran Conservative on 20,000 dollars, his recommended capital, then on 10,000. Result: the same 561 trades, the same 0.01 lots everywhere, so exactly the same gain in dollars, +2,165.88. But the equity drawdown goes from 11.96% to 23.23%. Double, on the most cautious preset.

That is exactly what his own minimum capital table anticipates: 20,000 dollars for Conservative, 15,000 for Moderate, 10,000 for Aggressive. Below those amounts, the preset you pick is not the one you get. The same effect, weaker, hits Moderate: 19.59% drawdown at 15,000 dollars against 24.20% at 10,000 over the same months.

On his long window at 5,000 dollars, the starting balance of his own live account, Conservative only has 0.01 lot trades from first to last, and its relative drawdown rises to 13.65% against 8.73% at 10,000. One caution of method: the effect shows on a fresh start, like a buyer's. Over the long window the balance grows and lots grow with it, which blurs the comparison.

At the recommended capital, Conservative gains almost 10% in August, gives back 6.64% in September, and its drawdown reaches 7.68%, below its range. On 10,000 dollars, the same trades make +19.76% then -12.18%. Same lesson: capital sets the risk.

He also publishes a live account at Pepperstone since 1 August, which he himself presents as a stress test: the Aggressive preset with doubled lots, on 5,000 dollars. I quote it as an illustration, not as a test result. Replaying exactly that setup over the same weeks gives +46.48% in August and -32.29% in September: the same shape, a very good month then a deep pullback, with a balance drawdown of 33.18% in the backtest against 36.28% on his account. The August gap is at least partly down to the broker's different price feed.

Over these eight weeks that setup ends at -40.71 dollars: +23.30 before costs, then 16.08 of commission and 47.93 of swap. On a flat period, costs decide the sign.

His gallery figures reproduce at another broker, within 50 dollars on Conservative. At the intended capital, drawdowns never exceed his ranges. Costs do not eat the edge over his window. And his page does not show the most flattering reading of his own window, it shows the most cautious one. The testable claims of the page all hold, with one nuance on the gallery's modelling mode, whose measured effect is limited.

This test does not say what the robot will do tomorrow: a backtest, even reproduced, describes the past, and eight weeks after release is a short sample.

No recommendation to buy or use this robot is made here. This page reports what measurements gave over past periods.