Caelion Gold Pro v2 and Quantum Queen X on the same real ticks: one account was stopped out

Disclosure: I build and sell Caelion Gold Pro on the MQL5 Market. I do not sell Quantum Queen X (QQX). Every figure below comes from my own MT5 Strategy Tester runs on IC Markets XAUUSD data. They are simulations; none is a live trading result, and you can repeat them with the free demo of each product.

On 7 October 2026 gold moved hard during the Asian session, and on 16 October it did it again. Basket and grid EAs on gold are built to survive exactly this kind of day, and plenty of them did not. I wanted to know why one design survives and another does not, so I ran my own EA and QQX on the same ticks with the same account.

The test

• Strategy Tester, "Every tick based on real ticks", IC Markets XAUUSD, H1.
• 1 May to 16 October 2026, USD 5,000 deposit, leverage 1:100, 30 ms execution delay.
• Broker spread, commission and swap as recorded. Both reports show 100% real-tick history quality.
• Caelion Gold Pro v2 on its default preset. QQX v4.4 on its recorded automatic Medium configuration, with its drawdown action left at its default (off).

Until 7 October, QQX was ahead: its equity sat above mine for most of the five months. On 7 October at 09:02 tester time its margin level fell through the broker's 50% stop-out line, the tester closed its positions, and the account finished at USD 4,712.43 from a closed-balance peak above USD 11,000. A stop-out is not a zero balance, so the net figure of -287.57 understates how much was lost on the way down.

My EA took a hit too. On 16 October its equity fell to USD 8,752.65, about 25% from its peak, before the basket closed in profit. That dip is the whole of its 25.16% drawdown. I show it here because any basket EA trading gold will draw down like this at some point.

Why one account survived: two rules that apply every day

Caelion contains no hard-coded dates. Nothing in the code refers to a calendar day, a past crash or a named event. I mention this because a date filter is the easiest way to make a backtest look perfect. Skip the days you already know were bad and the curve looks clean, but it has learned nothing about the next bad day.

The default runs two rules that apply every day of the year:

• No new basket adds between 21:00 and 01:00 on the strategy clock. The strategy clock is New York time plus 7 hours, and the EA derives it from the broker's own clock automatically. This window covers the thin hours around the daily gold break, when spreads widen and a basket that keeps adding can grow too large for the account.
• A basket stop at 55% of balance. It is the last line of defence.

To see what these rules actually do, I ran the same engine over 1 January 2019 to 16 October 2026 (USD 5,000, 1:100, model 0) with every protection switched off. It made USD 0.91 million, was stopped out 9 times and showed a 119% relative drawdown. With the no-add window alone, the same engine finished at USD 11.41 million with 0 stop-outs, a 50.70% drawdown and a lowest margin level of 62.75%. The 55% basket stop never fired in those eight years: the session rule did the work. I keep the stop anyway, for the day this history does not contain.

I have not isolated which rule mattered on 7 and 16 October specifically, and a five-month window cannot settle that. I also make no claim about how QQX is built inside. This post only reports what its tester run recorded.

How it was tested

A single good backtest proves very little. These are the checks the default had to pass before I shipped it:

• Every variant on one leaderboard. I ranked 139 full-period MT5 runs of this engine family on net profit, drawdown and lowest margin level together. That includes older versions, earlier sweeps and rejected ideas. The default is the one no other variant beats on net profit. The other presets are the variants that trade profit for a lower drawdown or a higher margin floor.
• Margin level as a first-class number. Drawdown alone hides how close an account came to a stop-out, so every run reports its lowest margin level next to its drawdown.
• Exact parity checks. When I rebuilt the EA for v2, the old presets had to reproduce the old results trade for trade, down to the cent. One parity run came out at half its expected profit. The cause was my test runner silently inheriting a new default input. I fixed the runner and re-ran the check.
• Ideas that failed stay in the record. I tested a margin-level rescue that closes the basket when margin falls under 125%. It held margin above 100% for the whole history, but every one of its 11 rescues closed a basket at a 10% to 25% loss that the default run later closed in profit, often within the hour. Compounded, it cut the 2019 to 2026 result from USD 11.41 million to USD 1.84 million. It stays in the EA as an option, switched off, and the user guide shows the full table.
• Real ticks for the decisive window. The long run uses generated ticks for speed. The window that contains the October losses was re-run on real ticks.

Limits you should weigh

• The window was chosen after the fact. I picked these months because the October losses had already happened, and my default was chosen using results that include this period. This is not out-of-sample evidence.
• Two complete configurations. Sizing, filters and clock handling differ between the two products.
• Tester results only. Live spreads, slippage and requotes will differ from recorded ticks.
• Five and a half months is short. I do not project an annual return from it.
• Compounding inflates long-run totals. The USD 11.41 million figure comes from lots that grow with equity, so the late years dominate it. Its 50.70% drawdown matters as much as the headline.
• Tick gaps. The QQX journal logs four minutes on 29 June with missing real ticks, filled with generated ticks. I kept the run as it was.

Repeat it yourself

Both products have free demos that run in the Strategy Tester. Use your own broker's ticks with the dates and settings above. If your numbers come out different, please post them in the comments; differences between brokers are worth knowing about.

Caelion Gold Pro v2 is on the MQL5 Market for MetaTrader 5 (product 202210) and MetaTrader 4 (product 202202). Search for "Caelion Gold Pro" in the Market tab of your terminal.