Machine learning in trading: theory, models, practice and algo-trading - page 2596

 
Maxim Dmitrievsky #:
Well, like train on the last history on different tray lengths and compare how the models worked on an earlier history, if they stop working at the same time somewhere, then here it is "the horizon of applicability in the past", taking into account the last history. And how it will be in the future is unknown, but you can determine the criteria for stopping the bot - if it started to trade worse than on the history, for example.

In a long sample, the general pattern that appears in the subsamples may disappear. If the patterns from the subsamples contradict each other, then the only thing left to learn is noise, which successfully happens in most cases :)

Yeah, that's probably it)

 
Maxim Dmitrievsky #:
Well sort of train on the last history on different tray lengths and compare how the models worked on an earlier history, if they stop working at the same time somewhere, then here it is "the horizon of applicability in the past", taking into account the last history.

This is what is called the "brute force method" or the "rule of thumb" method...

As you all understand, this is the worst possible option in terms of effectiveness...

Normal guys begin with the development of THEORY ... Even then go: the development of strategy, optimization and testing ...

 
Serqey Nikitin #:

This is what is called the "trial and error" or "gut feeling" method...

As you all understand, this is the worst possible option in terms of efficiency...

Normal guys start with working out THEORY... Only then come: strategy working out, optimization and testing...

Many theories (probably most of them) were developed just based on the results of practical experiments.
 
elibrarius #:
Many theories(perhaps most) were developed just based on the results of practical experiments.

Yes, you're right! But at that time (when "most") there were no effective problem-solving methods, such as TRIZ...

We had to grop for the best ways to solve the problem...

 
Serqey Nikitin #:

This is what is called the "trial and error" or "gut feeling" method...

As you all understand, this is the worst possible option in terms of efficiency...

Normal guys start with working out THEORY... Only then come: strategy working out, optimization and testing...

It's called "insert unobtrusive advertising on the sly". Normal guys don't do it 😂
 
Maxim Dmitrievsky #:
This is called "insert unobtrusive advertising on the sly".
What kind of advertisement? The TRIZ method? So it's necessary first of all for you, who don't know how to solve elementary problems...
 
Serqey Nikitin #:

Normal guys start by developing THEORY...

So start by studying the theory behind machine learning, not with demagogic picking on words.

 
Serqey Nikitin #:
Advertising what?... The TRIZ methodology?... So it is necessary first of all for you, who do not know how to solve elementary problems...
Confused with another comrade, the acronym is similar :) There were a lot of suggestions. DSP and kill yourself on the wall, now TRIZ...
 
Serqey Nikitin #:
Advertising what?... The TRIZ methodology?... So it is necessary first of all for you, who do not know how to solve elementary problems...

TRIZ is meaningless pseudoscientific nonsense which is not applied anywhere in practice.

What you advertise is your mentorship.

 
Aleksey Nikolayev #:

So start by studying the theory behind machine learning, not by picking on words with demagogy.

Have you thought about the fact that all this mumbo-jumbo hasn't given you any results...?

So why study something that doesn't produce any real results ...?

Are you relying on your uncle, who will tell you the best way of development...?