Machine learning in trading: theory, models, practice and algo-trading - page 1975

 
Maxim Dmitrievsky:

The next week I'll rewrite it on the terminal and then I can put it to the tests. I will rewrite it for the terminal next week and it will be possible to put it for tests.

Do you have a chart with spread?

 

I added new predictors (all sorts of returns according to my technology, volatility, regression channel for the day), I think it turned out okay - not perfect - plausible.

But even with the old ones I'm on the plus side, 30 models (100%) in the green zone - nice, I don't know, maybe the CatBoost training settings have also had an effect.

 
Maxim Dmitrievsky:

The next week I'll rewrite it on the terminal and then I can put it to the tests. I will rewrite it for my terminal next week and I will be able to use it for tests.

Maybe I should use watch watch?

 
Maxim Dmitrievsky:

By the way, this is a bug in the packet analyzer. I fixed it this way:

The standard there is jedi in the settings. I installed Pylance and changed 'jedi' at the bottom to Pylance in the studio settings through the search. Now it sees all the fields of the package and doesn't give out errors.

It worked )

 
Aleksey Vyazmikin:
14. Quick R Assemblies

15. The JIT Compiler

Why do you need all this?

You have a working algorithm, but it lacks speed?

you have a super-duper idea but it lacks processing power?

Why do you need all that? Concentrate on the TS, everything else is just bollocks.

Mihail Marchukajtes:

Hear me, I've got the ECLIPS and you can do it. Wipe your drooling weakling and keep working :-)

You understand R-ka already? Prof.)

 
mytarmailS:

How about watchdogs?

I made a magic button that reduces/increases the number of trades

get into python, you'll test it... I'm already sick of so much coding)

 
Maxim Dmitrievsky:

I made a magic button that reduces/increases the number of trades

get to know python, you'll test it... I'm already sick of so much coding )

i don't know if i could do that, but if i wanted to, i could become a programmer in a day)

 

My Pipeline. The initial goal is to create a base for further research, the final goal is to beatthe "classic" system.

It is based on a forest, but can easily be replaced by a neural network. There is loading from file, renders, history depth setting (input size), importance of inputs, error on tray and validation, simple tester.

The plan is to add a zigzag, ma, to finalize the tester.

Criticism is welcome.

 
mytarmailS:

Why do you need all this?

Do you have a working algorithm, but it lacks speed?

You have a super-duper idea but it lacks computing power?

Why do you need all that? Concentrate on the TS, all the rest is rubbish.

I have a system, but not enough speed.

 
Rorschach:

My Pipeline. The initial goal is to create a base for further research, the final goal is to beatthe "classic" system.

It is based on a forest, but can easily be replaced by a neural network. There is loading from file, renders, history depth setting (input size), importance of inputs, error on tray and validation, simple tester.

The plan is to add a zigzag, ma, to finalize the tester.

Your critique is welcome.

This "system" is designed for 1 DTs with positive slippages.

And where is the system itself, what is it compared to?