Dissecting 5,358 Real Trades: The Truth Behind the ×677 Martingale Gold EA — Why a 'Beautiful Upward Curve' Melts on Another Broker

This is the operator of FXEA365.

On the MQL5 Market there exists a Gold averaging-down (martingale) EA that advertises "$10,000 → ×677 (+67,600%)", attracting many buyers with its smooth, ever-rising balance curve. The flagship example is Quantum Queen (hereafter QQ).

This article is a record of dissecting all 5,358 of QQ's real trades, fully reverse-engineering its mechanics, then building a faithful clone of our own and verifying it on a different broker. Let me state the conclusion up front.

This is not a critique of any specific product — it is about the methodology for evaluating the martingale EA product category itself. QQ's numbers are genuine on its broker. The question is "will the same thing happen on your account?"

1. First, Let's Dissect What ×677 Is Made Of

From QQ's real trades (5,358 deals), we reconstructed 1,575 closures at the basket level.

Three facts can be read from this.

Fact ① Not a single losing basket (1,575 fights, 0 losses)

The basket win rate of every slot is 100%. This is the classic signature of martingale. A basket carrying an unrealized loss is never cut; it is averaged down with additional buys and held until it turns green (profitable). That is why the balance curve never shows a single loss.

Fact ② The risk is hidden not in "balance" but in "margin"

QQ's balance drawdown is only a few percent. But its margin (equity) drawdown is 79%. In other words, behind the "beautiful balance curve," the account's unrealized loss reached up to 79% of capital. It was walking a tightrope one step away from a forced stop-out (total loss), all while keeping a cool face on the surface.

Fact ③ Half the profit comes from "contrarian sells"

Surprisingly, 50% of QQ's profit comes from sell slots (T6 alone accounts for 42% of the total). 2020-2026 was a historic bull market in gold, yet QQ turned short baskets that fought against that rise into its single largest profit source by never cutting them and holding until a pullback. This is the exact opposite of the intuition that "it's a bull market, so just buy."

2. The True Identity of ×677 is "Compounding × the 2026 Spike"

Looking at the average profit per basket by year reveals where the number comes from.

QQ's Top 5 largest baskets were all in March-April 2026, sell baskets of 300-520 lots, each generating $200,000-590,000. This is the lot size that could only be placed once the balance had compounded from $10,000 to roughly $6.7 million, making the lots enormous. The bulk of ×677 is concentrated in the compounding spike of the final few months, while the early years were nothing more than a modest accumulation of $50-800 per basket.

3. We Built a Faithful Clone and Tried to Reproduce It on the "Same Account"

"If we know the parameters, we should be able to reproduce it" — with that thought, we built an EA that faithfully ported QQ's 7-slot structure (time zones, grid widths, lot coefficients, effective TP $89/lot) and ran a full-period backtest on the same MetaQuotes-Demo feed as QQ.

The result was unexpected.

• The same full 7-slot configuration as QQ → loses with PF 0.72 (the sell baskets burned out completely in our test)
• When we dropped the sells, expanded the buy strategy to 10 slots, and added a quality filter requiring "deep pullbacks (buy only $10 below the recent high)", it improved to ×15.62 / PF 5.49 / margin DD 32%.

Furthermore, even in IS/OOS (split-period) verification, it scored PF 5.97 in 2020-22 (including the COVID crash) and PF 5.42 in 2023-26 — PF above 5 across all periods. At a glance, it looked like we had completed "a robust aggressive EA that surpasses QQ."

——until we ran it on a different broker.

4. Cross-Broker Verification — And the Burn-Out

We ran the completed in-house EA (×15.62), without changing a single line of code, on Exness XAUUSDm (same gold, same period).

Same EA, same gold market, same period. Just changing the broker melted the account from $100,000 → $11,000.

When we lowered the lots to try to survive on Exness, the most that could survive was a combined lpk of 0.009-0.011. But at that point the performance was ×1.15 / PF 1.16 — essentially flat over 6.4 years, nowhere near a level that could be called an edge.

5. Three Checks Before Buying a Martingale EA

QQ, and my own clone, both fell into the same trap. From this experience, here is a checklist for purchase decisions.

1. Check the margin (equity) DD. Even if the balance DD is small, if the equity DD is 70-90%, it is a martingale that "simply hasn't exploded yet." Be wary if the sales page only shows balance DD.
2. Demand a track record across multiple brokers. Numbers from a single broker (especially the vendor's own demo) are likely specific to that feed. If they hold up on at least 2-3 brokers in live forward trading, they are genuine. Backtest only, or single broker only, is a danger signal.
3. Check whether there is a stop-loss. A 100% basket win rate and zero losing trades is not strength — it is merely "deferring losses as unrealized losses." An infinite martingale with no hard SL will inevitably end in one blow someday.

Conclusion

QQ's ×677 is a genuine record on top of that account. But by fully reproducing its mechanics and verifying on a different broker, we confirmed with real data that it is "single-feed overfitting," and there is no guarantee whatsoever that you will get the same result on your account.

The reason we at FXEA365 always verify on multiple brokers with a live tester when we release an EA is precisely to avoid this trap. Over a flashy multiplier, choose robustness that survives on any broker.

(The figures in this article are based on QQ's 5,358 publicly available real trades, and a MetaTrader 5 live backtest of our in-house clone [MetaQuotes-Demo / Exness, XAUUSD M5, 2020-2026, Model=1].)