QuestionHow do you choose a reliable XAUUSD Scalping Robot without falling for overoptimized backtests?

1
I’m new to forex robots and I’m especially interested in automated scalping on XAUUSD. I’ve noticed that many gold scalping EAs show very impressive backtest results, but I’m not sure how to tell whether the robot is actually robust or just overoptimized for historical data.

For those who have experience with XAUUSD scalping robots: what signs do you look for before trusting an EA on a real account? For example, do you pay more attention to live forward results, drawdown behavior, trade frequency, broker sensitivity, or how the robot handles spread and slippage?

I’m not looking for a “magic robot,” I’m trying to understand what separates a serious XAUUSD scalping EA from one that only looks good in Strategy Tester.

Re: How do you choose a reliable XAUUSD Scalping Robot without falling for overoptimized backtests?

3
PTR777 wrote: Wed Apr 29, 2026 11:53 pm The issue for gold scalping ea's (1m) are that live accounts has execution or order fill delays which make them useless in the real world

So if the scalping ea uses tight sl or trailing sl (which all do) the backtest will be inaccurate

Best is to check for live account track record of at least 6 months +
What if the robot traded well for 6 months, and I install it now, but in the seventh month it loses everything, for example, or simply starts trading at a loss because the market has changed or the trading conditions have changed? So there is still no guarantee, right?

Re: How do you choose a reliable XAUUSD Scalping Robot without falling for overoptimized backtests?

4
TimiStol wrote: Thu Apr 30, 2026 1:02 am What if the robot traded well for 6 months, and I install it now, but in the seventh month it loses everything, for example, or simply starts trading at a loss because the market has changed or the trading conditions have changed? So there is still no guarantee, right?
Yes correct - trading bots is (almost) a waste of time depending of how much money you have

Best road - Learn a real strategy - backtest - demo or prop trial - prop challenge - prop live

Re: How do you choose a reliable XAUUSD Scalping Robot without falling for overoptimized backtests?

5
TimiStol wrote: Wed Apr 29, 2026 8:58 pm
I’m not looking for a “magic robot,” I’m trying to understand what separates a serious XAUUSD scalping EA from one that only looks good in Strategy Tester.
First you need to be careful with metatrader

1) if you use MT4 from the get go, tick data will be extrapolated form M1 data, so it's totally unreliable

2) if you install data, it's better but still not optimal, because you don't really know historical spread of the broker
most data comes from Dukascopy which has a built-in spread very high.
so you need to install it with zero spread in mind. but with built-in MT4 spread it is not variable, so you can trade at 01:00 with a spread of 10 which is never possible in real life.

3) I would suggest using MT5 which is much stricter than MT4, and use "every tick based on history" that's the best data because it has historical tick data from the broker. so you need to be careful which broker you use. An EA can work on one broker but not on another one

4) it is very easy to create an EA and then optimize it to death. Most EAs nowadays (even those claiming using AI like Claude and Grok) are just martingales but with over fit data. So it performs really well in backtests, and when you test it live, it's a martingale so obviously it will "work" but in the end it is really dangerous.
Also lots of sellers on MQL5 Market cut martingale orders into smaller orders : for example instead of opening an order of 1.0 lot, and then an order of 2.0 lot, they divide the order into two orders of 1.0 lots, that way it doens't look like martingale

5) you need to look at ppt (profit per trade). You need at least 5 pips per trade (after fees) to work live.
OR
if you have insane winrate (liek 90%) maybe you can lower than amount

6) gold has been trending upwards for a long time, so of course backtests are gonna be good on Gold mostly...you need to be careful about that

Jef
Scalping the Century TimeFrame since 1999