Why most EAs perform better in backtesting than real trading?

I have tested a number of EAs on the MQL5.com/Marketplace which show strong backtesting ability but fail in forward testing, these were run on tick charts. Of course the sellers present their backtesting results which look spectacular. Can anyone explain what is going on here? Is this a case of deception by the sellers or just a simple oversight? Thanks for any comments.

These were run with a very low spread of 2.

You have your answer.

Please let me tell you a fact from my point of view as a programmer …

I did hundreds of EAs since the last days of MT3 … Few months ago, I did 4 or 5 EAs that use price breakout with different ideas.

They give me millions on the back tester, but useless on the forward testing. Not sure if I can make any one of them to work well

but at the moment all of them on the shelf. I can’t show people here just excellent back tests for EAs that could not make any profit

on demo or live accounts as a forward test.

Don’t rely on back test results. I even don’t trust 99% modelling quality. It gives to me some good feeling towards the EA (some hope)

but not trust it until try it on forward testing.

I believe that they should enforce every seller to link his EA to at least a demo account for a month to give him the license

to sell his product, once his account failed, the EA should be stopped. It is just an opinion !!!.

Thanks to both of you for the insights.

Strategy Tester is not reliable. The best way is test it is using demo accounts. Its the most close to the real account even demo account conditions sometimes are better than real for show trading more attractive. About EAs, can be profitables in a period of time but I think nobody will sell a EA profitable in long term because if you have one, you will earn enough for dont need sell anything :).

That don’t means you can’t make money with EAs but you have think about them like food. They have expire date ( when start to be not profitable ) and buy price have be proportional to profits.

Because a backtest doesnt take care of variable spread, order execution time, curve fitting, missed operations, delays, etc, and this affect the performance of an EA in real market conditions.

I believe that they should enforce every seller here to link his EA to at least a demo account for a month to give him the license

to sell his product, once his account failed, the EA should be stopped. It is just an opinion !!!.

I agree strongly

You have a good point here! I had the same thought for years but…still people rely in strategy test reports. I think at least a 3 month trading period in demo or real accounts before an EA should be treated as functional and be published. I’m one of those who believe in this and doing it myself.

You should move to Metatrader 5.

Backtest on Metatrader 5 is real history result , because when doing backtest on mt5 it follow real time spread. Every tick based on real ticks.

Use Mt5 real-ticks for tests,you will be surprised how many EA’s fail.Also helps determine broker suitability for strategies

There could be many issues, most of the time its the spread/scalping algos