Machine learning in trading: theory, models, practice and algo-trading - page 3411

New comment
 

By the way, I can't find a solution to one thing in kozul. How to distribute the error between meta-liners. Thought about it a lot, but so far I have not been able to do it.

I decided to make an experiment and try to do it through LLM. I'll post later what I got.

If it can do it, I will recognise their power.

 
Maxim Dmitrievsky #:

By the way, I can't find a solution to one thing in kozul. How to distribute the error between meta-liners. Thought about it a lot, but couldn't find it yet.

I decided to make an experiment and try to do it through LLM. Then I'll write what I got.

If it can, I will recognise their power.

Alas.... )

 
Maxim Dmitrievsky #:

Alas.... )

Predictably
 
mytarmailS #:
Predictable.

Somewhere around the edges, maybe I'm dumbing myself down.

 
Maxim Dmitrievsky #:
Listen, what is the probability of selecting working models by kozual. Well, for example, out of 100 models that passed OOS, how many will work on real data?
 
mytarmailS #:
Listen, what is the probability of selecting working models by kozual. For example, out of 100 models that have passed OOS, how many will work on real data?
They still have a low evidentiary power, even though they pass OOS. Because although the bias on new data and smooths out, but it is still unclear what is behind such a trade, what grounds that it will not break tomorrow. At your own risk, as they say.
But if you train many pieces in a package and most of them pass the OOS, the chances increase.

That's why I'm not completely done with them yet.

 

Maxim Dmitrievsky #:

I decided to do an experiment and try to do it via LLM. I'll post later what the results are.

If it can do it, I will recognise their power.

Ask the AI to draw a picture, sing a song or compose a symphony on the topic - they may inspire you to solve the problem) There is a thread on the forum, whose topikstarter considers it the highest manifestation of intelligence).

 
Aleksey Nikolayev #:

Ask the AI to draw a picture, sing a song or compose a symphony on the topic - maybe they will inspire you to solve the problem) There is a thread on the forum, the topikstarter of which considers this the highest manifestation of intelligence)

Well, there is already plenty of this infomusor in phones and laptops :) That's why people like storises and shorts in a few seconds - you don't have to think and it's like seeing something new.
 
Maxim Dmitrievsky #:
They still have a low evidentiary value, even though they pass the CB. Because even though the bias on new data is smoothed out, it is still unclear what is behind such a trade, what are the grounds that it will not break tomorrow. At your own risk, as they say.
But if you train many pieces in a package and most of them pass the OOS, the chances increase.

That's why I'm not completely done with them yet.

We can't know anything for sure, yes, but we can calculate the probability, but I wonder how likely it is to work.


I also have a method that adds confidence that the TS will work, it is to look at confidence intervals of Arima forecasts, there is also a method. Prado's Pbo, which also adds to the odds.

I'm all this to what... We can create a classifier that will say whether the TS will work on new data or not, and as signs of a bunch of different of our methods, together they are stronger than one.


We can add a lot of attributes, if the probability is more than 60-70%, it's already a grail, provided that we can generate a lot of TCs >=50.
 
mytarmailS #:
Listen, what is the probability of selecting working models by kozual. Well, for example, out of 100 models that passed OOS, how many will work on real data ?

None.

Confirmed by his personal experience.

Even his left OOS, with his knees backwards, does not guarantee against looking ahead.

Everything that Maxim writes is just advertising, which is not based not only on real trading, but even on trivial testing in the MT5 tester.

1...340434053406340734083409341034113412341334143415341634173418...3515
New comment
Reason: