Mirror Test for Gold EAs: Is Your XAUUSD Robot Just Riding the Bull Run?

20 November 2026, 10:46
Jonathan Mboya Kinaro
0
13

Gold climbed almost without a break through 2024 and 2025, set a record near 5,600 USD in January 2026, and has since fallen about 30% to a 14-month low this week. Many gold EAs on the Market were built and backtested during that rally. An EA that mostly buys looked brilliant on that chart without knowing anything about the market, and it is exactly the kind of EA that struggles when the trend turns. So how do you tell an EA that reads price action from one that has memorised a bull run?

The simplest check I know is the mirror test. I ran it on my own EA, GoldRush Pro Xauusd, and below is exactly how it works, the results, and how you can run it on any EA you are thinking of buying.

What the mirror test does

You turn the gold chart upside down. Every price is replaced by its reflection around a fixed level (the pivot):

mirrored price = 2 x pivot - real price

With a pivot of 3,500, a real price of 4,000 becomes 3,000 and a real price of 3,000 becomes 4,000. Every rally becomes a sell-off of exactly the same size and speed. Candle ranges, volatility, gaps and timing stay identical, only the direction flips. Bid and ask swap places, so the spread stays the same size.

The result is a gold chart that never existed: the 2025 rally becomes a crash and the 2026 sell-off becomes a rally. A robust trend-following EA should simply trade it the other way round and produce a broadly similar result. An EA that is fitted to the uptrend, or that quietly prefers buying, falls apart.

How I set it up in MT5

  • A small script copies the real XAUUSD tick history into a custom symbol and writes the mirrored ticks (pivot 3,500).
  • Same EA version (v12.3), same Balanced preset, same 15,000 USD deposit, every tick based on real ticks, 1 January 2025 to 20 November 2026.
  • The EA's US dollar filter reads the real EURUSD chart, which is not mirrored. I switched that filter off in every run so it could not contradict the flipped gold chart.
  • As a second check, I moved the same real prices 364 days earlier (52 full weeks, so weekdays still line up). This catches EAs that depend on calendar dates, seasonal rules or hard-coded days.

Results

Real goldMirrored goldDate-shifted gold
Final balance (USD)21,77926,25721,914
Net profit+45.2%+75.0%+46.1%
Profit factor1.491.701.50
Positions282291283
Win rate49.6%51.2%49.8%
Worst equity drawdown10.90%9.96%10.90%

How to read this

The mirrored run made more money, but that is not the point and it does not mean the strategy is better on falling markets. A mirrored chart is not the same market, and this sequence of swings happened to suit the EA a little better. What matters is what did not happen:

  • The profit factor stayed well above 1 in both directions (1.49 and 1.70).
  • The number of positions, the win rate and the drawdown barely moved. The EA behaved the same way on a chart it had never seen.
  • The date-shifted run was practically identical to the real one, so nothing in the logic depends on the calendar.

What the test does not prove

Passing removes one common failure: an EA that only works because gold went up. It does not prove future profits. You still need to look at costs, drawdown on your own account size and forward results. For GoldRush I also ran a Monte Carlo analysis (10,000 resampled runs per preset) and a cost stress test with an extra 1 USD per ounce on every entry and exit. Those will be the next posts.

Run it on any EA before you buy

  1. Create a custom symbol from your broker's XAUUSD and import its tick history.
  2. Write the mirrored ticks: new bid = 2 x pivot - real ask, new ask = 2 x pivot - real bid. Pick a pivot above half of the highest price in your period so every price stays positive.
  3. Test the demo version of the EA on the real symbol and on the mirrored one with identical settings.
  4. Switch off any filter that reads another symbol, or that filter will fight the mirrored chart.
  5. Compare profit factor, number of trades, win rate and drawdown.

Red flags: the profit factor falls below 1 on the mirror, the trade count changes sharply, or the EA trades almost only one direction. Any of those means the backtest is telling you about gold's history, not about the EA.

GoldRush Pro Xauusd is a trend-following EA for XAUUSD on H4 with seven risk presets. I will add the product link here as soon as the new listing is live. If you want me to run the mirror test on a specific EA, ask in the comments.

All results are backtests on historical data. Trading gold on margin carries a high level of risk and past results do not guarantee future results.