Why martingale EAs look profitable… until they aren’t (especially on gold)

New comment
 
Over the years, I’ve tested and observed many Expert Advisors, especially on XAUUSD.
And there’s something that keeps repeating itself again and again.
Most EAs that show amazing backtests on gold rely on some form of martingale or grid logic. At first glance, everything looks perfect: high win rate, smooth equity curve, fast growth.
The problem usually appears later.
Gold behaves differently compared to most forex pairs.
It can trend aggressively, stay volatile for longer periods, and ignore “mean reversion” much more than people expect. When that happens, recovery-based systems stop recovering.
I’ve personally seen accounts survive for months and then get wiped out in a single bad week. Not because the strategy was unlucky, but because the risk model was fragile.
Lately, my focus has shifted toward much more conservative ideas:
no position stacking
no recovery systems
accepting fewer trades if conditions are not favorable
adapting stops and targets to volatility instead of fixed values
What surprised me the most is that when you remove the pressure to “always be in the market”, the overall behavior becomes much more stable — even if trading frequency drops significantly.
I’m curious to hear other perspectives here.
Have you had better long-term experiences with aggressive systems on gold, or did you also end up moving toward more conservative approaches?
 
Lucas Leguisamo Mallo:
What surprised me the most is that when you remove the pressure to “always be in the market”, the overall behavior becomes much more stable — even if trading frequency drops significantly.
I’m curious to hear other perspectives here.
Have you had better long-term experiences with aggressive systems on gold, or did you also end up moving toward more conservative approaches?

the mgs that have little points in profit, will disguise, counter, and hide the losing trades. This is why. This is same reason why backtests with eas with mgs are virtually ALWAYS most profitable on backtests.

not saying mgs are always bad; my best profitting strategies use them. But if the strategy can not get through a drawdown period without an mg, then, that strategy is probably not profitable in long run.

 
Michael Charles Schefe #:

the mgs that have little points in profit, will disguise, counter, and hide the losing trades. This is why. This is same reason why backtests with eas with mgs are virtually ALWAYS most profitable on backtests.

not saying mgs are always bad; my best profitting strategies use them. But if the strategy can not get through a drawdown period without an mg, then, that strategy is probably not profitable in long run.

I agree with you on that.
Martingale itself isn’t “evil” by definition. Used carefully, it can be part of a profitable framework — especially when it’s controlled and not used as a blind recovery mechanism.
The issue, at least from what I’ve seen, is when martingale becomes the core of the strategy instead of a tool. If a system can’t survive normal drawdown phases without constantly relying on position size escalation, then the edge is probably not in the strategy itself, but in the temporary behavior of the market.
Backtests amplify this problem because they often reward anything that postpones losses long enough. As you said, small profit targets plus MG logic can make almost any system look great — until the conditions change.
Personally, I’ve become more comfortable working with systems that can remain structurally profitable without MG, even if that means slower growth and fewer trades. If martingale is added, it should enhance an already robust logic, not compensate for a weak one.
Good point, and I think that distinction is often overlooked.
 
Lucas Leguisamo Mallo #:
Martingale itself isn’t “evil” by definition. Used carefully, it can be part of a profitable framework — especially when it’s controlled and not used as a blind recovery mechanism.

IMHO... Martingale is, by definition, a blind recovery mechanism. It was invented by a gambler for use with a roulette wheel in a casino--where no sophisticated matrix is applied.

The somewhat similar but limited strategy component that you describe is commonly known merely as "scaling in"--a valid component.

 
Ryan L Johnson #:

IMHO... Martingale is, by definition, a blind recovery mechanism. It was invented by a gambler for use with a roulette wheel in a casino--where no sophisticated matrix is applied.

The somewhat similar but limited strategy component that you describe is commonly known merely as "scaling in"--a valid component.

That’s a fair distinction, and I get your point.
Pure martingale, in its original sense, really is a blind recovery mechanism — increasing exposure purely based on loss, without any structural market logic behind it. In that form, I agree it’s closer to gambling than trading.
What often creates confusion is that many strategies label very different things as “martingale”. Some are exactly what you described: loss-driven position escalation with no real edge. Others are closer to controlled scaling, where size adjustments are tied to context, volatility, or predefined limits rather than just “price went against me”.
Personally, I try to separate the concepts as well. Blind recovery is something I stay away from. Limited scaling, when it’s constrained and not responsible for saving the strategy, can be a valid component — but only if the system has an edge on its own.
So yes, terminology matters, and mixing those ideas is where a lot of problems start.
 
Michael Charles Schefe #:

But if the strategy can not get through a drawdown period without an mg, then, that strategy is probably not profitable in long run.

imo and experience -- the only good mg is one that has the same signal conditions as the trade that is in drawdown. ie all trades have the same signal conditions. Any deviation from the original strategy is breaking that strategy.

If the original strategy cannot get "itself" out dd -- then that strategy is doomed. an mg is an aid -- or tool as you said.

 

Getting back to gold for second...

On a high timeframe, gold-tracking instruments have basically gone in one predominant direction in the recent past--up. I really haven't seen a valid short entry in years. To me, it would be like shorting the S&P 500. If anyone wants to scale into gold, just scale into the dips... for the foreseeable future at least.

 
Ryan L Johnson #:

Getting back to gold for second...

On a high timeframe, gold tracking instruments have basically gone in one predominant direction in the recent past--up. I really haven't seen a valid short entry in years. To me, it would be like shorting the S&P 500. If anyone wants to scale into gold, just scale into the dips... for the foreseeable future at least.

Gold is a bit of a special case, that’s true.
On higher timeframes the bullish bias has been hard to ignore, and structurally it doesn’t behave like most FX pairs. In that sense, I get the comparison with something like the S&P, fading that kind of move long-term is usually a bad idea.
Where I’ve seen people get into trouble is assuming that macro direction automatically translates into clean execution. Even in strong bullish phases, gold can spend weeks chopping, pulling back hard, or expanding volatility in ways that punish “just buy the dip” logic if risk isn’t very well defined.
From an automated trading point of view, I’m less worried about whether gold is bullish or bearish overall, and more about when conditions are actually favorable to participate without overexposing the account.
So yes, the direction might be clear, surviving the path is the tricky part.
 
Martingale and grid are both in the realm of gambling. If you grew a small 500 dollar account to 2000 dollars, you win, and then you remove the gambler from the chart.
 
Conor Mcnamara #:
Martingale and grid are both in the realm of gambling. If you grew a small 500 dollar account to 2000 dollars, you win, and then you remove the gambler from the chart.

Again, let's be more specific with terminology. Martingale, scaling in, and grid are 3 different terms:

  • Martingale - a blind recovery strategy (without technical nor fundamental analysis) that doubles down (factors up) on every losing position.
  • Scaling in - a filtered recovery strategy component that is not activated unless the technical and/or fundamental analysis of the filter continues to align with the initial trade entry.
  • Grid - horizontal and equidistant lines on a chart that can be used in a multitude of ways, e.g., stoploss levels, takeprofit levels, round levels, simulated Renko levels, in-profit trade stacking, scaling in, Martingale, etc.*
         * A true grid obviously includes vertical lines as well but that's outside the scope of this thread.
     
    Let's do some advanced thinking , Martingale fails time and time again ,stuck in the death spiral  heavy trend markets ,so just reverse the logic to small loss small loss small loss small loss .BOOM big win , BOOM big win .Be willing to accept drip losses until you know the BIG hits are coming . Even better run the normal crappy martingales on a demo when it starts losing 2/3 step in with your real account . 
    12
    New comment