Machine learning in trading: theory, models, practice and algo-trading - page 299
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
If the forecast will be confirmed in most cases on the same history, then why do you think that this is nonsense?
It is for the reason that it is confirmed that the coolest pattern will be found, not the other way around. On the forward side, it will be a bummer.
Trading by patterns is a myth.
It is for the reason that it is confirmed that the coolest pattern will be found, not the other way around. On the forward side, it will be a bummer.
Trading by patterns is a myth.
When I draw results I'll post them on the forum or maybe even an article, but the approach is out of the box. And the biggest misunderstanding, that I faced with, is exactly the search of patterns through correlation, 50% of the quality of prediction is gone already at this stage, that's why I mentioned machine vision
such conclusions require evidence)
It is impossible to provide them, because no matter how scrupulous and complex the evidence, it is always possible to object "look for patterns wrong".
To refute my statement, of course, it is possible to counter-argument - robust TC on patterns. But in my opinion it is even less probable than the mythology convincing.
It is impossible to provide them, because no matter how scrupulous and complex the evidence, it is always possible to object "look for patterns wrong".
It is of course possible to refute my statement with a counter-argument - robust TC on patterns. But I think that's even less likely than convincing me of mythology.
I will try to refute in the form of a TS :) even purely for myself it is interesting
That is, by asserting the lack of possibility of a TS on patterns, we deny any intelligible TS on machine learning at all, and this topic could be closed altogether :)
That is, by stating that there is no possibility of using a ts on patterns, we deny any intelligible ts on machine learning at all, and this topic could be closed altogether :)
Yes.
Too strong and unsupported.
We take the machine learning algorithm "random forest" - randomforest. Run it on a sample of 5000 bars and try to find 500 trees - which are your patterns.
Result.
After a hundred trees, the error drops very little. And multiple increase of sampling does not lead to increase of trees. That is, we can assume that this algorithm finds less than 100 patterns (trees) and additional trees have little effect on the result
This is how it looks like
Too strong and unsupported.
We take the machine learning algorithm "random forest" - randomforest. Run it on a sample of 5000 bars and try to find 500 trees - which are your patterns.
Result.
After a hundred trees, the error drops very little. And multiple increase of sampling does not lead to increase of trees. That is, we can assume that this algorithm finds less than 100 patterns (trees) and additional trees have little effect on the result
Here's how it looks like
Do the same for a random BP.
Maxim Dmitrievsky:
I will try to refute in the form of TS :) even purely for myself it is interesting
That is, by stating that there is no possibility of a ts on patterns, we deny any intelligible ts on machine learning at all, and this topic could be closed altogether :)
fxsaber:
Yes.
So how do you trade then? By guessing?
Indicators are also means of MO, not parametric regression, but primitive...
And without statistical advantage in forecasting there is no point in trading as we know. But some people trade, make millions of trades and are consistently in the black, how is that? It is not an insider, because trades last a few seconds, a minute and there are thousands of them a day. It turns out that these guys predict somehow, and there is nothing better than sophisticated MO for this purpose.
That is the proof of MO's work in the markets, the very fact that HFT offices make a stable alpha with double-digit Sharpe.
So how do you trade then? By guesswork?
Indicators are also MO, not parametric regression, but primitive MO...
And without statistical advantage in prediction there is no point in trading, as we know. But some people trade, make millions of trades and are consistently in the black, how is that? It is not an insider, because trades last a few seconds, a minute and there are thousands of them a day. It turns out that these guys predict somehow, and there is nothing better than sophisticated MO for this purpose.
That is the proof of MO's work in markets, the very fact that HFT offices make a stable alpha with double-digit Sharpe
By guessing yes, sometimes it's funny, sometimes it's sad... I was just saying that it's possible to make market predictions based on patterns, but everybody needs proof, you can't do without it. Moreover, I think that it is possible to forecast anything, and soon mankind will achieve unbelievable results, up to forecasting and making maps of all mankind's life, and then it will do hara-kiri to itself at a certain day.
If you ask a philosopher or esotericist, he will tell you that there is no time and that the present, future and past are an illusion, maya, and the result of imperfect perception. So you only have to look at the past and know the future without much difficulty, for all is one.
That is the proof of MO's work in the markets, the very fact that HFT firms make a stable alpha with double-digit Sharpe