How Long Should You Test an EA Before Deciding It Doesn't Work?

[Archimedes18]

I've noticed something over the years.


Many traders expect an EA to prove itself within a few trades.


One win...


"It's amazing."


One loss...


"It's garbage."


That doesn't make much sense.


No professional trading strategy is built around one trade.


It's built around probability.


If a strategy has an edge, that edge appears over a meaningful number of trades—not after one or two positions.


This raises an interesting question.


How many trades do you personally need before deciding whether an Expert Advisor is worth keeping?


Twenty?


Fifty?


One hundred?


I'd like to hear how other traders approach this.

 

[Isabelletrem]

An EA shouldn’t be judged after just a few trades. Give it 3–6 months of forward testing and enough trades to see how it handles different market conditions. Focus on drawdown, consistency, and risk, not just profit.

 

[EA ForexLAB]

The question isn't framed quite correctly. Drawing on my years of experience, the first step is to rigorously test the Expert Advisor in TDS using real spreads for MT4 or "every tick based on real ticks" data for MT5. Up to 80% of Expert Advisors will be eliminated at this stage alone.

The next important step is to verify the trades both in the backtest and on a live account over the same period, thereby confirming that the backtesting itself is relevant and meaningful.

 

[Magnolia_York]

I wouldn’t really judge an EA by a fixed number of months. First, I’d focus on the quality of the backtest, good data, and proper testing conditions. If it survives that, then I’d forward-test it and see whether the live results actually match the backtset.