Machine learning in trading: theory, models, practice and algo-trading - page 154
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
Hoping that a convolutional or recursive network will do everything by itself is a naive view, just as we used to think about graal turkeys.
It's not a naive view. It's the direction of the search. And it's not necessarily a neural network. The thesis is this: from past values of a time series of prices, you can pull information sufficient for profitable (overcoming costs) trading regardless of the time horizon of the actual forward test.
I will also post a couple of charts on this topic. I am currently preparing material for an article.
PS: Personally, taking into account all the factors leading to overtraining, and trying to take the most conservative and reliable model condition, it turns out that more than 30-40% per year (with max drawdown 25%) is not to squeeze out. But it already exceeds the median yield of hedge funds. All other cosmic interest supposedly received in the long term purely on the technical analysis of time series - this is a lie.
It's been on MQL for years.
where to look?
=======
oops didn't see it :)
where to look?
1) Your thinking is correct as a start, but you need to use not only currency pairs, but also raw materials, stocks, futures, options,
2) Lao Tzu was an algotrader saying, "He who knows - keep silent; he who speaks - does not know")))
I think that in order to do this I need to understand the mechanics of the market and be able to formalize the unformalizable, when I will learn, I will share my experience...
I understand the mechanics, what's left to formalize
2) a lot...
Above gave the link.
I didn't find anything about cross correlation and the approach I described, just portfolio trading, that's all.
Now I get it.
This is consistent with my observations. I call it mirroring (I'll probably post a couple of graphs on this topic later). For predicting a n-minute price increase, looking backwards for the same n-minutes shows the best explanatory value.
But beyond that, I have very extensive data on which horizon the predictive power is greatest.
Where do the 4-5% numbers come from? How are they counted? Significance of predictors, R^2, mutual information?
This is an empirical estimate of the gain in classification quality with - without this factor, it's simple, mutual information and determinacy in nonlinear multifactor systems do not work reliably. And figures 4-5% is certainly not a dogma, you just need to understand that using "all markets" and information flows without price dynamics of the instrument you can predict its future for a certain horizon <5% worse, that's all. That is, if you have a one minute probability of predicting the future rise in front of the asset, for example 70%, then excluding the price of the predicted series from the data for analysis you get 70 - (70-50)*0.5 = 69%, almost within the limits of noise. Well, of course if you have real-time data from all world markets and not only markets, but without insiders, and if only one instrument's price... whatever AI you have, it's easier to create a terminator than to beat a market with such data.
I did not find anything about cross-correlation and the approach I described, just portfolio trading and that's all...
The covariance matrix followed by the correlation matrix was found somewhere on MQL here on the resource.
I only remember that this matrix was calculated very quickly when shifting the window to BP. Perhaps this is also done in R.
And the deviations you were talking about were shown directly in the terminal. Ask around, someone will tell you.
The covariance matrix with subsequent transition to the correlation matrix was found somewhere on MQL here on the resource.
I only remember that this matrix was calculated very quickly when shifting the window to BP. Perhaps this is also done in R.
And the deviations you were talking about were displayed right in the terminal. Ask around, someone will tell you.
yes of course done :) but Covariance matrix is not cross-correlation :)
You can use any tools you want. If you want, do it through cross-correlation head-on.
It was about finding "outliers" to then trade them towards "balance".
yes of course done :)