Discussion of article "Multiple Regression Analysis. Strategy Generator and Tester in One"

 

New article Multiple Regression Analysis. Strategy Generator and Tester in One is published:

The article gives a description of ways of use of the multiple regression analysis for development of trading systems. It demonstrates the use of the regression analysis for strategy search automation. A regression equation generated and integrated in an EA without requiring high proficiency in programming is given as an example.

Fig. 10. The equation includes the significant parameters only


Author: ArtemGaleev

 
I don't know, I don't know...
According to the charts, the ratio of profitable trades to winning trades is very low, and there are large drawdowns.
Also, it looks like small stops and large profits (or fast closing of losing trades and overholding of profitable ones).
Now in demand are point advisors, which open rarely but aptly.


Although the topic is quite interesting, in principle, I used to train a neural network in the same way.
Only the neural network trained itself, although it consumed a lot of resources. And also to understand the principles of neural network behaviour on estimation of weights,
is many times more difficult than to look at a beautiful table that Statistica gives out. :)

And the saddest thing is that almost any super clever method of analysis is easily outperformed by an Expert Advisor with two or three mashka and some bollinger.
And both in the short and long term....

 
It should be noted that MQL5 supports output of only 64 indicators to a file in one line.

And if you glue a string with the necessary data set separated by the required delimiter (for example, "asd;qwe;zxc[....]bnm") by yourself and pass it to FileWrite...?

 
I also see some advantage of "strategy generators". It remains to learn how to use them correctly
 

komposter:

It should be noted that MQL5 supports output of only 64 indicators to a file in one line.

What if you glue a line with the necessary set of data separated by the necessary delimiter (for example, "asd;qwe;zxc[....]bnm") and pass it to FileWrite...?


I didn't understand from the text what this phrase is referring to at all?

But to refute it, I'll tell you that you can fit up to ~32000 letters into one string without any problems, and using the "\r" delimiter you can fit the whole file into one string.

This StringConcatenate() function has restrictions on parameters, but nobody forbids to simply add to the existing string, or to use StringConcatenate() repeatedly .

 
Well, our regiment has arrived. It remains to switch from statistics to EViews - full regression analysis will be available, including residuals and stability analysis
 
faa1947:

Well, our regiment has arrived. Now we need to switch from statistics to EViews and full regression analysis, including residuals and stability analysis, will become available
Analysis of residuals and stability deserves a separate publication. Here I have only mentioned it in passing. As for statistical programs, there are many different ones. Since regression analysis is one of the basic analyses, it is included in many programmes.
 
ArtemGaleev:
Analysis of residuals and stability deserves a separate publication. Here I have only mentioned it in passing. As for statistical programmes, there are many different ones. Since regression analysis is one of the basic analyses, it is included in many programmes.
EViews is my personal choice in terms of some knowledge. Started a thread here based on two articles. You are welcome to participate.
Эконометрика: прогноз на один шаг вперед - MQL4 форум
  • www.mql5.com
Эконометрика: прогноз на один шаг вперед - MQL4 форум
 

Из статьи:

Interestingly, the training data did not exhibit over-optimisation, as is often the case in the tester. This probably indicated the absence of overoptimisation.

fat plus to the author's karma for such optimism)
 
Urain:

I don't understand from the text at all what this phrase is referring to?

But to refute it, I'll tell you that you can fit up to ~32000 letters into one string without any problems, and using the "\r" delimiter you can fit the whole file into one string.

This StringConcatenate() function has restrictions on parameters, but nobody forbids to simply add to the existing string, or to use StringConcatenate() repeatedly .

This is probably what FileWrite(h,1,2,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,64) is about;

It's not even 64, but 63. There may be 64 parameters of the function in total.

 
Integer:

Probably about this FileWrite(h,1,2,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,64);

Not even 64, but 63. There may be 64 parameters of the function in total.

Well, you can use "+" instead of "," though this is probably an answer not to you but to the author.