Discussing the article: "Developing a multi-currency Expert Advisor (Part 2): Transition to virtual positions of trading strategies" - page 4

[Maxim Dmitrievsky]

Yuriy Bykov #:

Of course :) But, seriously, I started working on the articles only after the first time I managed to achieveresults similar to the results of the optimisationperiod for the forward period ofat leastone year. For example, when optimisation was performed for the period strictly till 2023, when running for 2023 I got something like this:

This inspires some optimism, but it must be handled very carefully to avoid self-deception.

About building a strong classifier from a few weak ones - that's the main idea, that's the hope, that you can achieve some useful results.

Yes, then the hard part starts - can we trust this forward, can we not. Whether he is obtained by chance or not by chance.

[mytarmailS]

fxsaber #:
How different perceptions of the same article can be, however.

Why is he wrong?

[mytarmailS]

Yuriy Bykov #:

As for building a strong classifier out of several weak ones, that is the main idea and the hope that it will be possible to achieve useful results.

I also like this idea, in fact, I voiced it and gave an analogy with RF classifier, and also posted pictures of simulations of growth....

An ensemble of ts is almost always better than a single ts, it's a fact. But provided that the ace is earning on average.

[fxsaber]

mytarmailS #:
Why is he wrong?

I don't think the quote and this question after it have anything to do with it.

[mytarmailS]

fxsaber #:

I think the quote and the question after it have nothing to do with each other.

Well, your quote was an answer to Maxim's remark.

You didn't just write it.

As I understood your vision is not the same as the writer above and you wrote about it in your quote.

So I wondered whether you wrote it because you disagree with Maxim or just because you are surprised that people see the same thing differently.

[fxsaber]

mytarmailS #:
Ну ваша цитата была как ответ на реплику Максима.

I read Maxim and Alexei. And then I wrote.

So I wondered whether you wrote it because you disagree with Maxim or just because you are surprised that people see the same thing differently

My perception of the article is not about TC (which is taken just for example), but about architecture.

Primitive architecture (MQ out of the box) makes it easy to write something superficial on your knee sometimes, but doesn't allow you to test deep things.

A complex/correct architecture makes it possible to write deep stuff, but it doesn't do it easily.

And then there are the technical decisions in implementing the architecture. This is an important but rarely covered point.

[mytarmailS]

fxsaber #:

Got it.

This is a really important and interesting topic

[Maxim Dmitrievsky]

It is possible to do not bagging (basket) of strategies, but boosting. This is when one strategy is optimised first, then its signals are substituted as a parameter for the second strategy, the second strategy is optimised, and so on. This will remove not only the bias, but also the scatter. I have not seen such implementations on mql.

[fxsaber]

Maxim Dmitrievsky bagging (basket) of strategies, but boosting. This is when one strategy is optimised first, then its signals are substituted as a parameter for the second strategy, the second strategy is optimised, and so on. This will remove not only the bias, but also the scatter. I have not seen such implementations on mql.

https://t.me/fxsaberDiscussion/3877

[mytarmailS]

Maxim Dmitrievsky bagging (basket) of strategies, but boosting. This is when one strategy is optimised first, then its signals are substituted as a parameter for the second strategy, the second strategy is optimised, and so on. This will remove not only the bias, but also the scatter. I have not seen such implementations on mql.

It is not a fact that in the market the complication of the model will work better than a basket of simple trades