Your Gold EA Survived a 5× Spread Test. That Doesn't Mean It's Safe.
Recently I was stress-testing one of my Gold systems and started with the obvious thing:
spread ×1 → ×2 → ×3 → ×5
The EA survived.
Nice.
Then I realized I was asking the wrong question.
Because during CPI, spread is only part of the problem.
The real problem is execution.
And that's much harder to backtest honestly.
Your EA can be right — and still lose
Imagine your EA correctly predicts that XAUUSD will go up.
But the market moves before the order is filled.
Your entry is worse than expected. The stop gets hit during the first violent move. Seconds later, Gold goes exactly where the strategy predicted.
Signal: right.
Trade: lost.
That's the part a beautiful backtest can hide.
A 5× spread test is useful, but let's be honest: it doesn't recreate a real news release.
During major events we can get widening spreads, slippage, execution delays and very fast changes in liquidity.
So I now see the spread test for what it really is:
a simple way to start looking for fragility.
I stopped trying to prove that my EA works
This has probably been the biggest change in my own development process.
When I was working on my XAUUSD EA, Golden Royal, I caught myself spending too much time looking at the good side of the backtest.
Profit Factor. Recovery. Equity curve.
The numbers we all enjoy looking at.
Eventually I started doing almost the opposite:
trying to break my own system.
And I learned more from that than from another optimization run.
Spread.
Execution.
Volatility.
Different periods.
Bad sequences of trades.
Different market regimes.
The question changed from:
“How good is this backtest?”
to:
“How little does it take to destroy it?”
That is a much less comfortable question.
But I think it's the more useful one.
Not every EA breaks the same way
This matters.
For a scalper, spread and execution delay can destroy the edge very quickly.
For a trend system with wider targets, the same increase in spread may barely matter.
And for a grid or averaging system, the spread test can actually give false confidence.
A grid might look healthy even at spread ×5 — right until one violent directional move expands beyond the range it was designed to survive.
So there isn't one universal stress test.
You have to understand where your particular EA is fragile.
That's also one reason I personally moved away from Martingale, Grid and Averaging when building my current Gold system. I wanted risk to be visible rather than hidden inside the next recovery trade.
Not because that approach is perfect.
Nothing is.
I simply sleep better when I understand where the risk is coming from.
Profit Factor isn't enough
Another lesson: an EA can remain profitable and still become unacceptable.
Imagine your normal Max DD is 12%.
Under slightly worse conditions it becomes 35%.
PF is still above 1.
Technically profitable.
Would you be comfortable running it?
I wouldn't.
That's why I increasingly look at degradation from baseline, not one magic number:
Max Drawdown
Recovery Factor
Largest Loss
Consecutive Losses
Equity behavior
If a small deterioration in trading conditions completely changes the character of the system, I want to know why.
And backtesting has limits
We should be honest about this too.
Strategy Tester is an incredible tool.
But it isn't the future.
We don't know our future liquidity.
We don't know our future slippage.
We don't know the exact path price will take during the next CPI release.
So I wouldn't use stress testing to say:
“My EA is safe.”
I'd use it to discover:
“My EA is more fragile than I thought.”
That's valuable information.
Especially before real money is involved.
One last thought
There is a trade-off here.
We can make an EA so defensive that it barely trades.
Robustness isn't about surviving absolutely everything.
That's impossible.
It's about understanding where the edge ends and deciding whether the return is worth the fragility.
That's the philosophy I've been trying to follow with my own Gold development.
For anyone curious, I've put the current implementation and test information for Golden Royal here. Don't take the backtest numbers on faith — I'd much rather you download the demo, stress-test it yourself and try to find where it breaks.
Seriously.
Try to break it.
And if you find something interesting, tell me.
That's more useful to me than another “nice EA” comment.
Because after enough years around trading systems, I'm becoming less impressed by an EA that looks great under perfect conditions.
I'm much more interested in what happens when conditions stop being perfect.
What test has exposed the biggest weakness in one of your EAs?
Spread? Slippage? Volatility? Execution delay? A bad market regime?
I'd genuinely like to hear it.
The ugly tests are usually the useful ones.
