Different result from non visual to visual mode.

[Yerim Rayann Gon Diarra]

Hi fellow dev!

I stumble apon something quiet disturbing concerning the strategy tester. 

I test the same EA with the same parameters twice. First test was with visual mode and the second was without visual mode. 

Both result where completely different. 

I made an hypothesis, since my strategy use graphical object created with ObjectCreate(), it may the  causes. Because those object my be non existent in non visual mode. 

But if this is true, how many others flaws like that exist? 

This is alone put in question all those strategy I tested throughout the years. Were they really losing money? Or the tester was just to bad? 

Please I need answer!!! 

[[Deleted]]

Yerim Rayann Gon Diarra: I stumble apon something quiet disturbing concerning the strategy tester. I test the same EA with the same parameters twice. First test was with visual mode and the second was without visual mode. Both result where completely different. I made an hypothesis, since my strategy use graphical object created with ObjectCreate(), it may the  causes. Because those object my be non existent in non visual mode. But if this is true, how many others flaws like that exist?  This is alone put in question all those strategy I tested throughout the years. Were they really losing money? Or the tester was just to bad? 

There is no "hypothesis"—it is a documented fact, which you have overlooked.

EAs dependant on Graphical Objects cannot be back-tested normally, nor optimised, and this is stated in the documentation:

• Documentation on MQL5: MQL5 programs / Testing Trading Strategies / Graphical Objects in Testings

[Yerim Rayann Gon Diarra]

Fernando Carreiro #:

There is no "hypothesis"—it is a documented fact, which you have overlooked.

EAs dependant on Graphical Objects cannot be back-tested normally, nor optimised, and this is stated in the documentation:

• Documentation on MQL5: MQL5 programs / Testing Trading Strategies / Graphical Objects in Testings

Thanks you Fernando!

So If go through all this section of the documentation, will I be able to see all pitfalls I missed? 

[[Deleted]]

Yerim Rayann Gon Diarra #: So If go through all this section of the documentation, will I be able to see all pitfalls I missed? 

For the most major part yes, but there are also the minor cases to consider—logic bugs in the code that can also cause inconsistency.

However, that is the responsibility of the coder to make sure their code is properly written and that they have fully debugged their code to prevent it.

[[Deleted]]

Yerim Rayann Gon Diarra:
Hi fellow dev!

I stumble apon something quiet disturbing concerning the strategy tester. 

I test the same EA with the same parameters twice. First test was with visual mode and the second was without visual mode. 

Both result where completely different. 

I made an hypothesis, since my strategy use graphical object created with ObjectCreate(), it may the  causes. Because those object my be non existent in non visual mode. 

But if this is true, how many others flaws like that exist? 

This is alone put in question all those strategy I tested throughout the years. Were they really losing money? Or the tester was just to bad? 

Please I need answer!!! 

I have found that EAs which use AI may indeed show significantly different results depending on whether you have the visual mode enabled or not.

It may indeed be an issue/mistake in the EA's code or with the tick data, but the visual mode seems to be closer to reality.