MetaTrader 5 Strategy Tester: bugs, bugs, suggestions for improvement - page 19

 
Irql is iron
 
Renat Fatkhullin:
Irql is hardware.
On 6 threads everything worked in the end. Judging by the fact that every time different errors, most likely, the tester ate all the resources, especially on memory and started crumbling.

PS. The memory problem will be solved if you make general ticks with mapping for agents plus cache. There will be lapses of course, but the more threads, the more hits.

I tested the hardware, by the way. No errors.
 
Hello! Faced with this problem - the result of optimization from the old version of MT5 - cache file, does not fit the new version from 4.10.2019 - who mtalked with this - how to cure? Just nothing happens - no cache file is loaded.
 

Is there a way to pass parameters during optimisation to a remote agent? Or a file or a string? Local or network agent.

 
Another problem I've encountered is that the opti-me gets stumped before half of the initial number of rebounds is reached. What affects this? How to fix it? Observed on several pairs - EUR AUD and USD CAD at least
Files:
jzoh912.JPG  68 kb
 

The old tester doesn't hang the system... At all)
Came across this article on hubra:https://habr.com/ru/post/402551/

Maybe the new tester uses new instructions, I just have Ryzen 5 2600, I will try to update the microcode.

PS. Yay!!! Everything worked after updating mother's bios (microcode goes there as well). In 12 threads the calculation goes without glitches and hangs
 
Please do not check the optimisation flag in the settings when selecting the optimisation cache.
 

A very handy solution for input variables is to specify their name via a comment.


I would like to improve the functionality as follows

input int inRange1 = 23; // Количество пунктов //  0||1||40||Y
input int inRange2 = 8;  // Количество пипсов  // -9||3||12||N

Highlighted code sections specify a default optimization range for each input parameter. In the settings tab of the tester, you can select this default range defined by the author of the Expert Advisor for each parameter.


Now, the Market Advisors do not have default ranges. In order to optimise them, we have to ask the author about the range for each parameter.

With the same solution, everything would be much easier even for the authors of Expert Advisors when writing them for themselves.


Let me know if this functionality is useful, or it's one of my strange tricks.

 
fxsaber:

A very handy solution for input variables is to specify their name via a comment.


I would like to improve the functionality as follows

Highlighted code sections specify a default optimization range for each input parameter. In the settings tab of the tester, you can select this default range defined by the author of the Expert Advisor for each parameter.


Now, the Market Advisors do not have default ranges. In order to optimise them, we have to ask the author about the range for each parameter.

With the same solution, everything would be much easier even for the authors of Expert Advisors when writing them for themselves.


Please let me know if you think this functionality is useful, or it is just one of my strange side effects.

1. What is there to handy for multilingual EAs? It's better to set the names by separate methods, depending on the language.

2) This format is not convenient at all, especially if something needs to be changed in a range. I just save the settings and read them later - this is really fast and convenient.

3. Why do you want to optimize Expert Advisors in Market? I think that's what the Expert Advisor author should do.

PS. I'm not picky. My opinion: everything should be simple, convenient and logical, not "convenient for a developer's temporary solution in a particular place".

 
Am I correct in assuming that restarting the GA adds its passes to the existing opt file?
Reason: