Are MT5 EA capable of history read in backtests as in MT4?

New comment
 

I know there are a lot of MT4 EAs that read the tick history thru a DLL, then those EAs shows impressive results on backtests (fake , manipuled results)

I think it was not a problem in MT5 but I detected one MT5 EA that shows excellent results in backtests but completely different results in demo or real life.

I compared the life trades it took with the ones on backtest and I notice that the losing trades were not recorded on the backtests, only the winning trades.


To any experienced user/developer:

  • What is your opinion about it?
  • Has you also found the same situation?
  • Can we still trust in MT5 backtesting?


Thanks for your inputs

 
It would be of great help if you report such EA to Support. With software, anyone can create such fake tests. My opinion is that you better get EA from MQL as it will not allow any person to load such EA in market. 
 
Banyone Kapengele Chandala #:
It would be of great help if you report such EA to Support. With software, anyone can create such fake tests. My opinion is that you better get EA from MQL as it will not allow any person to load such EA in market. 

The EA I am talking about is published in MQL marketplace

 
I think that in no test platform should the Expert Advisor ever be allowed to be able to read future bars. If there are any such flaws in the MT4 and MT5 platforms, then the developers should put a serious fix in place. At the very least create a visual message for the end user that the EA has read the historical data. This does not necessarily mean that the EA is a scam, but the end user needs to be able to know.
 
Benjamin #: then the developers should put a serious fix in place.
Already is. DLLs are not permitted.
 
William Roeder #:
Already is. DLLs are not permitted.


Not enough. In fact, I still see so many EAs doing early bar reading during the BT.

 

It's called over-optimization. You don't need cheating to perform impressive backtest curves. You could easily add conditions(either rational or non-sense) to filter out bad trades in the backtest. And add new entries when the successful trade is not enough to make good profit. And that's how traders thought it would work during the early trading-bot programming stage.  But most entries will generate bad result when tested with new data, so you need months to test and refine the EA, months again to re-test it.

The historical trading data is vast that you could pick tiny little portion of winning trades. And some of them may narrowly escape from being stopped out. 

It would be easier if the developer already knew some proven profitable strategies and is able to implement them in the EA. 

If the EA has not been modified/updated for sometime and the developer is willing to show you the backtest result starting from the last modification time.  Then the EA is much more reliable.

New comment