Machine learning in trading: theory, models, practice and algo-trading - page 2070

 
Aleksey Vyazmikin:

There units only 24%, and the strategy is profitable for 7 years on minutes!

The CatBoost does not train by it at all, it has all zeros, only the genetic algorithm on R somehow catches the logic, but badly.

What kind of strategy is this?)
 
Aleksey Vyazmikin:

There units only 24%, and the strategy is profitable for 7 years on minutes!

The CatBoost does not train by it at all, it has all zeros, only the genetic algorithm on R somehow catches the logic, but badly.

Why all zeros? It seems to be giving out probabilities by classes. Trade if the probability of one is 30% or 40% and look at the balance.
 
Alexander Alexeyevich:
What kind of strategy is this?)

After the appearance of a new segment of ZZ, we wait for a pullback and stand in the direction of the vector of ZZ, stops at the point of appearance of a new vector of ZZ.

 
elibrarius:
Why zeros? It seems to give probabilities by class. Trade if the probability of one is 30% or 40% and look at the balance.

I thought to try it that way when I'll figure out the article, but I think it will make a lot of noise - such shifts seldom give stability.

 

So who will teach this strategy, he can earn, I think so :) I can fill in a sample - to try my forces, if you can teach and share the secret - I will share the strategy in the code!


Basically on 1 indicator.
 
Aleksey Vyazmikin:

So who will teach this strategy, he can earn, I think so :) I can fill in a sample - to try my forces, if you can teach and share the secret - I will share the strategy in the code!


I will use only one indicator.
Aleksey Vyazmikin:

So who will teach this strategy, he can earn, I think so :) I can fill a sample - to try my forces, if you can teach and share the secret - I will share the strategy in code!


Basically on a single indicator.

How do you teach without knowing the strategy in code)?

 
Alexander Alexeyevich:

how to teach without knowing the strategy in the code)?

By sampling, as everyone here teaches...

 
Aleksey Vyazmikin:

On sampling, as well as all here teach ...

I work with mt, and it teaches a bit differently) i can't put sampling there, it takes longer to train, but it has its advantages)))

 
Aleksey Vyazmikin:

So who will teach this strategy, he can earn, I think so :) I can pour a sample - to try my forces, if you can teach and share the secret - I will share the strategy in the code!


In essence on the 1 indicator.

Will sit in a flat for a year!)

 
Maxim Dmitrievsky:

will sit in the flat for a year )

This is not the worst possible option :)))

Reason: