AMD or Intel as well as the memory brand - page 20

 
joo >> :

Something tells me that different quote values will have no effect on the result. Everywhere there will be numbers like x.xxxx. That is, the amount of processed data will be the same for everyone, except for a negligible difference in the number of bars.

On the subject of antivirus... You probably shouldn't joke about such things. We're all dealing with real money. ;)

But something tells me that different optimization results due to different history (depends on brokerage companies) - the number of deals, etc. - will, all else being equal, give different optimization results. - Other things being equal, different run times.

It's easy to check, though.

 
Svinozavr >> :

But something tells me, that different optimization result due to different history (depends on brokerage company) - number of deals, etc. - Other things being equal, it will give different run times.

But it is easy to check.

And why make an EA for a test, which is designed for real trading? I mean, you can open and close positions strictly at specified times, for example, without stops and takeovers. Then everyone will have the same amount of open and closed positions. We need to compare the total time of the test, not the total profit.

 
joo >> :

And why make an EA for a test that is intended for real trading? I mean, we can open and close positions strictly at specified times, for example, without any stops and takeovers. Then everyone will have the same amount of open and closed positions.

I wanted to make do with a standard EA (as originally intended). Then, don't forget that everyone has different spreads and different testing times. That's why I've been talking about offline and editing the files where it all lies.

>> OK. Now I will test one and the same Expert Advisor in two brokerage companies and check the time of optimization. If the differences are not critical for our purposes, then we will do so - without any troubles. Agreed.

 
Svinozavr >> :

What I meant was the same story for everyone - roughly speaking, the problem is creating your own symbol.

In the army, it may be ugly, but it's uniform.

Symbol, number of bars, etc...

What if we put it all into a csv-file?

and read everything from it...

The format is the usual one, for example, that saves the terminal:

2009.01.02 06:00,1.22020,1.22060,1.21980,1.22050,89,0
2009.01.02 06:30,1.22080,1.22260,1.22050,1.22120,325,0
2009.01.02 07:00,1.22130,1.22150,1.22120,1.22120,8,0
2009.01.02 07:30,1.22100,1.22220,1.22100,1.22170,75,0
2009.01.02 08:00,1.22190,1.22340,1.22100,1.22340,233,0

;)

That is.

The script, at first start-up, checks whether the required file exists.

If it does not exist, creates a new one, if it does, step 2 checks "quality", i.e. number of lines.

If it is not, it will create new one. If it is OK, continue...


Since this is solvable by means of µl, there should be no problems with operation.

 

Speaking of more complicated...

Is there an API function to root out information about current CPU load?

And at the same time, information about the CPU itself, what kind of CPU, manufacturer, etc...

 
kombat >> :

In the army: however ugly it may be, it is uniform.

Symbol, number of bars, etc...

What if we put it all into a csv file?

and read everything from it...

The format is the usual one, for example, that saves the terminal:

;)

That is.

The script, at first start-up, checks whether the required file exists.

If there is no it, creates a new one, if there is step 2 it checks the "quality", i.e. the number of lines.

If there is a problem, it creates a new one, if ok, we continue...


As it can be solved by means of µl, we shouldn't have any problems with exploitation.

Yes? And how do you get the tester to see it? What's the symbol?

 

Wouldn't it be faster to write your own tester than to think how to cheat Mql?

Actually, all the code I have that handles swaps, spreads, stops, takes

and opening trades took 19 Kb. Of course, with such a large volume there is no potiky

emulation and pending orders.

.

As it was a C++ solution, no problems :-).

.

I'll say even more - since when transferring from Mql

the indexes in the arrays are expanded (0 on the left, current bar on the right),

not so long ago I had to make a small class which unwrapped the indexing

back ;-) (old bars on the left, 0 on the right).

I ended up with a Close[0], Time[0], etc. construction.

 
Svinozavr >> :

Yes? And how do you get the tester to see it? What's the symbol?

Is it so crucial for testing what the figures in the prices will be?

;)

double БИД= считанная в данный момент цена из файла;
OrderSend(Symbol(),OP_SELL, lots, БИД, slippage, loss, profit,"",0,0,CLR_NONE); 
 

In short!

Let's take the Moving Average Expert Advisor. We pump up the history on the minutes. Then we set these parameters:

Parameters in the window below can be downloaded from t.set file or set manually.

Take off the flag of genetic algorithm! It is clear why.


This is it. The first screen will be a report with optimization time. I checked it on two brokerage companies: one has a 2 points spread and the other has 3 points spread with no difference.

EA (just in case) and setup file in the archive.

Go, brothers in madness!!!

Files:
t.rar  2 kb
 

Nah, it's better on all the ticks. There are only 510 variants, you can run each one longer. Or don't you want to think about tick modelling, which can introduce variation?