A little surprised :) Thought I'd share and ask a NOT rhetorical question. - page 2

[Renat Fatkhullin]

Academic:

In a word, the merits of MT-optimizer, the optimizer in particular, are not obvious to me.

But don't lie so blatantly and pretend to be incomprehensible.

Do not go too far, please.

[Academic]

Renat:

But don't lie so blatantly and make it sound like you don't understand.

Don't go overboard.

I'm sorry, I've already noticed you're a bit paranoid. I am not overreacting - I do not understand, I cannot even guess what you are imagining about me. I'm being completely honest - for me personally it is not obvious - as there is no way to overtake the digitizer. That's why I do not see it. But this I personally do not see, and what they see ( and probably there are ) others I do not know.

I do not understand what you think I may be pursuing a secret purpose, quite. :)

[Renat Fatkhullin]

You are not the first to make a simple loop of your own for(i=0; i<bars; i++) and come to the forum with a question "why do I need a tester if I am making N million bars per second here". Of course, no one will ever see, use and have no chances to apply this solution anywhere. And history is constantly repeating itself.

In a serious discussion (and for us this is serious business for the last 10 years) attempts to argue on the level of "I'm being completely honest - for me personally it's not obvious, I don't understand it" look frankly ridiculous. Maybe in another society you can play the incomprehensible.

What the MetaTrader 5 tester has:

• multicurrency
• automatic download and synchronization of the entire history
• several modes of bar simulation
• access to the entire market environment
• access to indicators
• use of protected and secure (EX5 format) Expert Advisors and indicators
• multi-threaded testing with support of remote agents
  
• developed and tested infrastructure
• real processing of all trade operations, modeling of trading errors and requotes
• genetic optimizer that allows to complete the optimization in a reasonable time
• reporting test results
• visualization (1D, 2D, 3D) of results, exporting results
  
• well-documented and extensive tutorials
• availability of 64 bit versions
• MQL5 programs execution speed in the native 32/64 code
• 10 years of experience in development of automated trading language (MQL appeared in 2001)
• etc.

and against it you should:

• loop for(i=0; i<bars; i++)

[Academic]

Renat:

You are not the first to make a simple loop of your own for(i=0; i<bars; i++) and come to the forum with a question "why do I need a tester if I am making N million bars per second here". Of course, no one will ever see, use and have no chances to apply this solution anywhere. And history is constantly repeating itself.

In a serious discussion (and for us this is serious business for the last 10 years) attempts to argue on the level of "I'm being completely honest - for me personally it's not obvious, I don't understand it" look frankly ridiculous. Maybe in another society you can play the incomprehensible.

So it's funny and OK.

But it became clear to me that I must not speed up the tester, but use a separate optimizer. And that's all. It seems to be an obvious thing, but it became 100% clear. So I think ( honestly ) maybe I have missed something? Although I don't think so. But the point is not I have noticed as you have not called for anything, nor am I asking. I am happy with everything. And you God forbid, I do not want to change your mind or clarify anything, what I see that I understand. Nope. It's fine. I'm sorry. No... there's really no sarcasm or backbiting. :)

[михаил потапыч]

Academic:

that you should not speed up the tester but use a separate optimiser. That's all. Seems to be an obvious thing,

Would you mind explaining it to me in more detail, for a dummy?

[Academic]

Mischek:
Would you care to explain in more detail, for a dummie?

There's nothing there - everything is obvious and banal - the optimizer needs to be very fast. And the tester needs to be realistic.

Don't really pay attention. I mean that it's easier to write a numeric grinder in C++ to search for the optimum parameters and that's all. Nothing too abstruse. But it's not so simple - or rather simple if you program quickly and without problems. Like in your mother tongue. Apparently, this is the only case where it makes sense to do something like this. Otherwise you can get bogged down in errors and so on. Then it's easier to use a MT tester.

[михаил потапыч]

Academic:

There's nothing here - everything is obvious and banal - the optimizer - you need it very quickly. And the tester needs to be realistic.

So "realism" for the tester and for the optimiser may be different?

In the tester "by ticks"? and in the optimizer by"opening prices" for example?

[Roman Zamozhnyy]

Once, about five years ago, I wrote my tester in pure C (which I honestly admitted to https://www.mql5.com/ru/users/edit). I gave it up. I'd rather spend time generating new ideas than writing super-fast code.

[Academic]

Mischek:
So "realism" for the tester and for the optimiser may be different?

The optimiser is a search, and the tester is a test.

Optimizer says - it will be a plus for so-and-so parameters, and tester confirms - yes it's right. :)

[Academic]

Mischek:

So "realism" for the tester and for the optimiser may be different?

In the tester "by ticks"? In the tester, and in the optimizer by"opening prices" for example?

No, this is exactly what we should not do. And we should not imitate work with the history completely.