Machine learning in trading: theory, models, practice and algo-trading - page 134

 
Andrey Dik:
Did you try to read my article? There just such a ZZ is discussed as an example of a complex optimization problem.

I remember her. But since I know what GA and its variations are, I didn't read into it, honestly.

There's just going to be 2000000 chromosomes in the genome, if anything...

 
Alexey Burnakov:

I remember her. But since I know what GA and its variations are, I didn't read into it, honestly.

There's just going to be 2000000 chromosomes in the genome, if anything...

You should have not read it - not chromosomes in the genome, but genes in the chromosome.

And you shouldn't be so dismissive, there are no other variants (more acceptable in time).

 
Alexey Burnakov:
I used an indicator that searches for price patterns by Euclidean distance and draws continuation into the future with confidence intervals. But it didn't have x-axis scaling and was not very good.
What's wrong with just the average?
 
Andrey Dik:

You should not have read it - not the chromosomes in the genome, but the genes in the chromosome.

And you shouldn't be so dismissive; there are simply no other options (more time-appropriate) for a solution.

Okay. Thanks for the correction.

And if we just optimize the value in pips to build the breakpoint - it would not be an approximate ideal solution? To enter trades then with equal volume at inflection points.

PS I myself wrote GA completely from scratch, so I know how it works. And I used it a lot.

 
Alexey Burnakov:

Okay. Thanks for the correction.

And if we just optimize the value in pips to build the inflection point, it will not be an approximate ideal solution? To enter trades then with equal volume on inflection points.

PS I myself wrote GA completely from scratch, so I know how it works. And have used it a lot.

That is exactly what you should do - look for the maximum value of points, which will be obtained if you pass through the obtained peaks of ZZ. This is exactly the example in the article.

However, I do not insist, do as you see fit. Once again I see that nobody needs good advice, everyone prefers just his own rake...

 
Andrey Dik:

This is exactly what you should do - look for the maximum value of points, which will be obtained if you pass through the vertices of the obtained ZZ. This is exactly the example in the article.

Yeah. I've already read the practical example in the article.

But there you also found that the most ideal entry and exit points do not lie on the most ideal ZZ (if you optimize it only). That is, you can find the most ideal points through optimization "on every bar".

 
Alexey Burnakov:

Yeah. I already read the practical example in the article.

But there you also discovered that the most ideal entry and exit points do not lie on the most ideal ZZ (if you optimize it only). That is, you can find the most ideal points through optimization "on every bar".

An ideal PZ is not the one built by the indicator, but it is the one found by points on each bar. The article shows that the classical indicator is inferior to the pointwise one.
 
Andrey Dik:

This is exactly what you should do - look for the maximum value of points, which will be obtained if you pass through the obtained vertices of ZZ. This is exactly the example in the article.

However, I do not insist, do as you see fit. Once again I get convinced - nobody needs good advice, everyone prefers just his own rake...

And what are you suggesting? I have not understood. Suggest again. I was just making a point.
 
Andrey Dik:
The ideal PZ is not based on the indicator but on the points on every bar. The article shows that the classical indicator is inferior to the pointwise one.

There you go! That's what I understood and that's what I asked here in the thread.

And what do you think, Andrey, is it possible to find a good value while doing optimization on millions of bars?

 
Alexey Burnakov:

There you go! That's what I understood and that's what I asked here in the thread.

Andrey, do you think it is possible to find a good value when optimizing on millions of bars?

For how long?