Machine learning in trading: theory, models, practice and algo-trading - page 3689
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
In general, for some reason still no one does not do the fitting on examples of existing signals, although the idea was voiced a long time ago.
There are few signals with a sufficient number of deals on one instrument. Or these signals have grids and other averaging methods.
I do not have such a signal in mind.Few signals with sufficient number of trades on one instrument
That's what over-sampling/augmentation methods are touched on here for :)
Was re-reading the thread, over the last 2.5 years. I found that some of my posts look harsh, there is no context for it (or moderators have erased it). I apologise to anyone I may have offended. Apparently, this is my flaw - one of many.
When is it Forgiveness Sunday? )
And you will forgive us :)
And you'll forgive us :)
Who exactly do you represent? Give us the whole list, please :)
Ah, who exactly do you represent? Could you tell us the whole list, please :)
I'm picturing myself and my hordes of cockroaches.)
I forgive you, then.
This is a summed transformation of price using a kind of causal decomposition technique to transform price into more predictable (wave-like) components.
And this is using Maxim's propensity code on these summed series as input and all components as features, although it has yet to be tested in real time.
And this is using Maxim's propensity code on these summarised series as inputs and all components as features, although it has yet to be tested in real time.
Curious. Did you come up with it yourself or are there any papers on the subject?
Curious. Did you come up with it yourself or are there any articles on this topic?
This is not a scientific article, I derived it after years of endless searching for alternatives to inconvenient methods like Fourier transform and other decomposition techniques. The main problem is that most algorithms use data from the future and suffer from edge effects.
In practice, it's a very simple solution and whenever I tried to add more features/complexity, everything fell apart...
I'm willing to share this, if you want to collaborate on this and other things, drop me a message.
This is when using components as features, 50 thousand minute bars for training and 10 thousand bars for testing. It's not perfect, but the yield curve becomes much smoother than when using only moving average traits.
This is when using components as signs, 50 thousand minute bars for training and 10 thousand bars for testing. It is not perfect, but the yield curve becomes much smoother than when using only moving average traits.
What are the results when trading on a live account?