Machine learning in trading: theory, models, practice and algo-trading - page 3200

 
You do a cross-validation and see how long you are willing to watch this mess. If you are ready, you put it into trade.

For ticks, I'll do it soon.
 
Ivan Butko #:

Magic is forbidden outside Hogwarts.
Enter point to point.

My post was just so that when designing signs in the MoD someone would first think -

What does my stochastic have to do with real market processes?

Is it possible to look at the market as a BP ?

Is the absolute price in the market important or should everything be normalised ?

...

..

.

 
mytarmailS #:

My post was just to make sure that when designing signs in the MoD, someone would first think about -

What does my stochastic have to do with real market processes?

Is it possible to look at the market as a BP ?

Is the absolute price in the market important ?

...

..

.

Write down where to get the volume history and that's it.
Levels are also based on BP values.
How long can you be in a state of permanent dumbness?)
 
Maxim Dmitrievsky #:
Yeah, but the fact that you can find them
.
If you give me a non-sb, I can find patterns there too.

Would you like to talk about that? What and how do you use to find patterns?

As for the non-SB sample - I've wondered about it myself, as I want to test my method on it.

There are samples without predictors, and I can't easily calculate my predictors on them - I have to get it all into the terminal somehow....

If you come across a sample with predictors, with at least 15k-20k observations, with a binary target, please let me know where I can download it.

 
Aleksey Vyazmikin #:

Would you like to talk about it? What and how do you use to find patterns?

As for the non-SB sample - I'm wondering about it myself, as I want to test my method on it.

There are samples without predictors, and I can't easily calculate my predictors on them - I need to get it all into the terminal somehow....

If you come across a sample with predictors, with at least 15k-20k observations, with a binary target, please let me know where I can download it.

Modified approach from the last article, with kozul elements
 
mytarmailS #:

My post was just to make sure that when designing signs in the MoD, someone would think first -

What does my stochastic have to do with real market processes?

Is it possible to look at the market as a BP ?

Is the absolute price in the market important or should everything be normalised ?

...

..

.

I would add that all these questions must be asked, realising that the goal is to create a car that has a sufficiently large number of parts that are critical to the car. And it is enough to attach exactly one foreign part to the car, so that the car stops driving. It takes months to carefully select a large enough number of parts and assemble them into a single working mechanism with less than 20% classification error outside of the training file.

You had an excellent post with PCA, which was negligently pointed out by one Maxim, although it is a classic for the application of statistics: PCA not working and working if the variance of all vectors is on the same scale. It would seem that what does the variance have to do with it? Didn't normalise the variance and the whole car doesn't go, again a 50/50 classification error.


On the thread there is NO understanding of the systematicity of the problem of MO application

 
Maxim Dmitrievsky #:
Write where to get the history of volumes and all
.
Levels are also based on BP values.
How long can one be in a state of permanent dumbness?)

Maybe try to read posts with a desire to understand what a person writes about?

 
СанСаныч Фоменко #:

Maybe try to delve into the posts with a desire to understand what the person is writing about?

You always do a great job of that
 
Maxim Dmitrievsky #:
1. Write where to get the volume history and all
.
2. Levels are also built by BP values.
3. How long can you be in a state of permanent dullness?)

1)

I don't use volumes as initial signs of ohlc and sma and different TFs (they are useless in themselves), further more complex things are built on their basis....

But I could not formalise it into a standalone code, it's too complicated.... overshooting is impossible (there are trillions of features)


2)

Someone who is fascinated by BP theory will never build levels...

He will build arims, arches, garchies, seasonalities, etc... or stochastics, mugdies, etc....

It's a different kind of thinking.


3)

Hz...

"I only know that I know nothing, but others don't know this either."

Socrates.

 
If I write an article on the new approach now, the whole market will be flooded with such bots, because I write thoroughly, a complete implementation that even a child can run. And then all the bumps will fall on me again? I don't want to :)