Machine learning in trading: theory, models, practice and algo-trading - page 2419

New comment
 
Maxim Kuznetsov:

- I want a car to win the race for sure!

- here's a Ferrari for you!

---

- your car is crap and it sucks. I didn't make it to the finish line, the car's wrecked, there's bodies everywhere... I could hardly get away from the police to say "the car sucks!"

- idiot! Tour de France ...

PS / only read the thread, but here I could not help but say

Maxim, the joke is funny, but not on the subject)

 
Renat Akhtyamov:

the algorithm can not break, the algorithm can only become a drain over time, because it has such a fate

so if the question of force majeure stability and stability has not been solved, it is better not to implement it

And about the fairy tale you wrote above is correct, over time you will understand the depth of those words

Renat, you are probably right.

I have a question for you: you have a huge branch of thousands of pages on machine learning, but in practice all the robots with machine learning are lost immediately on the forward test, that is, where it is trained, it is a god, only in battle, the algorithm stops working. Why so?

 
YURY_PROFIT:

Renat, you're probably right.

I have a question for you: you have a huge branch of thousands of pages on machine learning, but in practice all the robots with machine learning are lost immediately on the forward test, that is, where he is trained, he is a god, only in battle, the algorithm stops working. Why is that?

Great question, fundamental. I look forward to getting answers from the experts.)
 
YURY_PROFIT:

Renat, you're probably right.

I have a question for you: you have a huge branch of thousands of pages on machine learning, but in practice all the robots with machine learning are lost immediately on the forward test, that is, where he is trained, he is a god, only in battle, the algorithm stops working. Why so?

This is not my thread, this is a forum.

Any statistical strategy / theory, and this is what it is, uses a small finite section of historical data for one instrument, a sample in simple terms

what is statistics?

The development of a pattern

so the answer is simple - there is no pattern

based on this - what do we find?

A certain pattern at the selected part, which is not always repeated, the real fiasco.

But there is a pattern in the market

Buying equals selling, but not on one currency pair, but in a portfolio of currency pairs in at least three

when the number of currency pairs increases, the same thing happens.

Search there.

 
Renat Akhtyamov:

This thread is not mine, this is a forum

Any statistical strategy/theory, and this is what it is, uses a small finite section of historical data on a single instrument, a sample in simple terms, for analysis

what is statistics?

The development of a pattern

so the answer is simple - there is no pattern

based on this - what do we find?

A certain pattern at the selected part, which is not always repeated, the real fiasco.

But there is a pattern in the market

Buying equals selling, but not on one currency pair, but in a portfolio of currency pairs in at least three

when the number of currency pairs increases, the same thing happens.

dig there.

Renat, I just wanted to say that there is no machine-trained Expert Advisor that would not fail beyond the trained period.

There are profitable Expert Advisors using other principles, but there are no profitable ones on the basis of machine learning.

If I remember anything about Maxim, the programmer I was quarreling with, his Expert Advisor is machine learning and backtesting very good, on forward testing it lasted for a month and a half with good profit statistics, now it is losing without even pauses).

In his article he has a lot of useful information in his articles on machine learning. It is really clear that the man is deep in the subject.

And nothing, nothing.

Maybe he should admit that machine learning in trading is just a dead-end way of development and not to mislead neither beginners - programmers, nor traders.

 
YURY_PROFIT:

Renat, ....

Maybe we should admit that machine learning in trading is just a dead-end way of development and not to mislead either novice programmers or traders.

You do not understand my post and I understand yours

The standard situation in the perception of information in the forex market.

the test is not a real trade in the market and never will be

First find out for yourself the answer to the question - what is the difference between the real and demo trading

 
Renat Akhtyamov:

You didn't understand my post, I didn't understand yours

the standard situation of perception of information in the forex market

the test is not a real trade in the market and never will be

First find out for yourself the answer to the question - what is the difference between the real and demo trading

Renat, I understand you very well.

I mean, what is the point of those thousands of pages of clever text about machine learning, if it's only a misleading for novice programmers and consumers of your product.

In practice, in real trading, there is not a single example of profitable Expert Advisors on this basis.

You are not doing this just for fun, but for the final result, which is a profitable Expert Advisor.

This programmer, who is a newcomer, reads your clever words and spends thousands of hours, his brain seems to be developing and working, but the result is nothing.

It's time to create a branch called "Machine learning in trading - is it the future or a dead-end way of development?

Renat, I'm sure you are a very experienced programmer.

Please give me an example of a profitable Expert Advisor using machine learning, which worked in the real market for at least 3 months in profit without re-training.

 
YURY_PROFIT:

Renat, I understand you very well.

I mean, what is the point of these thousands of pages of clever text about machine learning, if it is only misleading for novice programmers and consumers of your product.

In practice, in real trading, there is not a single example of profitable Expert Advisors on this basis.

You are not doing this just for fun, but for the final result, which is a profitable Expert Advisor.

This programmer, who is a newcomer, reads your clever words and spends thousands of hours, his brain is developing, but the result is nothing.

It's time to create a branch called "Machine learning in trading - is it the future or a dead-end way of development?

Renat, I'm sure you are a very experienced programmer.

Please give me an example of a profitable Expert Advisor using machine learning that worked in the real market for at least 3 months in profit without retraining.

Put it this way

try it on the EURCHF or USDCHF, on the demo for a couple of months to start.

if you don't like it - just throw it in the trash and that's it

you can't really discuss it, the rules are like this

I do not have a neuron;)

 
And it's right that there is no neuronics, it's a dead end. If there is no trading algorithm inside the Expert Advisor, just training on historical data is a dead end and a waste of time!
 
You're at the wrong end of the spectrum. The discussion here is about what's interesting, including trading related topics. Standard TS are stable one in a thousand or more, there are relatively few neural networks. So it's a survivor's mistake to claim that MO doesn't work. If this thread didn't exist, there would definitely be nothing to read on the forum.
Renat is on the side here, selling monkeys.
1...241224132414241524162417241824192420242124222423242424252426...3496
New comment
Reason: