Machine learning in trading: theory, models, practice and algo-trading - page 1755
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
that you look at the local trends, they are different in time, how do you gradate, at least the logic of how? That's certainly how it is. The answer is in the test on all symbols from 70 to 20 tickwise )))) and if the profit in 90% of the series ))))) Although damn probability is a no brainer... and 10% is always .... Gotta educate always and opt out too get it....
Everything is pretty dumb. For example the rsi indicator is smoothed by polynomial regression in a sliding window of a given length. Subtracts the first value, we get a local trend from zero. Then we try to sample trades with different frequency, from different distributions and choose the variant, that is stable on the new data.
Like this
https://www.mql5.com/ru/articles/4777
Are you drunk? ))
All right, the belief and understanding of the problem are too different things. if we go public, then the same problem on different conditions, no more. your proposal is too broad)))))
Everything is pretty dumb. For example, the rsi indicator is smoothed by polynomial regression in a sliding window of a given length. The first value is subtracted, a local trend from zero is obtained. Then we try to sample trades with different frequency, from different distributions and choose the variant, that is stable on the new data.
Like this
https://www.mql5.com/ru/articles/4777
that's what I thought.... sad, although the algorithms are 20 30 years old.... the answer, we are not the first. Averaging in the current filter, in the algorithm to hell... I can't explain it, but after looking at how old the algorithms are, you come to the conclusion, their results are random, something is wrong. I have no idea what and how and I don't know what to do yet.
Marakovuyu how to take into account all the TF. and make the right conclusions. Past leaders simply did not have such an opportunity.
kinda figured that way.... sad though the algorithms are 20 30 years old.... the answer, we are not the first. Averaging in the current filter, in the algorithm to hell... I can't explain it, but after looking at how old the algorithms are, you come to the conclusion, their results are random, something is wrong. The analysis today, I don't know yet what and how and the past additions to the filter.
It was just an example, sometimes you still need to smooth the series
All right, the belief and understanding of the problem are too different things. if it is public, then a homogeneous problem on different conditions, no more. Your sentence is too broad)))))
What the fuck faith? What kind of nonsense are you all talking about? What's wide in a single dataset for everyone ? How does it work?
What the fuck is wrong with you people? What are you all up to?
All right, I get it, I'm the only one.
I'm trying to figure out how to use one algorithm to account for all the timeframes and make the right conclusions. The previous leaders simply didn't have this ability.
Spectral analysis
It's been invented 100 years ago)
What if we create a dataset (unified for all) with target and prices + various useful indicators and post it here, make a test and a trace and also "test2" for full OOS check of already trained model.
People will download the dataset and try to improve the quality of the classification, if something works, it will be added to the dataset as a chip/indicator.
This requires a sufficient level of understanding of people's goals, which is difficult to provide in a public. People instead of target actions will do not target actions, ask questions and waste time and your time too)))))
requires a sufficient level of understanding of people's goals, which is difficult to achieve in a public forum. People will do non-targeted actions instead of targeted actions, asking questions and wasting time and your time too)))))
What other purposeful actions? What are you talking about?
target is a target variable in the dataset! or class label or target, label, class whatever you want to call it..... This is not some subjective something, but a clear often binary vector !
It requires a sufficient level of understanding of the purpose of the task, which is difficult to provide in a public domain.
It's just as simple, and succinct.
If you can not reduce the error in the dataset, it means there is no understanding, if there is no understanding, then sit down and do not write! And he reads clever people, and clever people are those who managed to reduce the error!
Everything is clear and not contradictory!
The smart ones become smarter because they work in a pseudo team because they work on one dataset.
Fools get smarter because they watch what the smart ones do.
Dataset grows with new features which better and better predict the target.
All win!, but damn, you should understand it, but who does not ((
Spectral analysis
Spectral analysis has been around for about 100 years))
I understand it, but it's for one averaging, or rather thinning, the question is how to make the same algorithm for all timeframes and how to decide if the target timeframe is 15 minutes instead of 5 minutes, then we look at the hourly, then the 1 hour is over and we go back to 1 minute, then to 5 minutes. Limiting the minimum spread to the trend is understandable, but it's just breakeven, not maximizing the profit. How to quickly calculate/estimate on what TFs the profit will be maximal? And at the same time the expectation risk will not be exceeded to breakeven.
What other targeted actions? What are you talking about?
Targeted action is what we assume people, performers or friends should and will do when we do something together. But sometimes you say one thing, and they don't understand you the way you assume. In one-sentence tasks, this is easy to fix. In complex ones it's harder.