Machine learning in trading: theory, models, practice and algo-trading - page 2592

New comment
 
Maxim Kuznetsov #:

I don't like to interject, but just a remark about hundreds or other MA ...there is a limit to their reasonable number and it is no more than 1.386*ln(N) (where N is the whole observable history)

Disagree, love it)

Why not 1.387?)

 
Replikant_mih #:

If the model inputs both the period of the average and the value of the average, it won't care what you call a predictor and what you call a parameter.)

The period of an average is invariable in each line - we will get an uninformative column. We do not need it. We remove it to speed up calculations.
 
Maxim Dmitrievsky #:
It depends on what you are looking for. It implies a plateau on the global, as an ideal and a dream.
No, it is not.
 
Replikant_mih #:

Well, after adding ML I don't want to optimize anything further. The wind of overfitting blows from that side). Although if the speeds are normal, you can at least try it for sure. In general, yes, because of not the best integration speed I paid little attention to trader add-on over ML, if so integrated and in the tester conditionally-native can be tested and normal speeds, it certainly opens additional horizons of possibilities.


And in general the better speed (in comparison with my solution, I think the normal speed difference will be) is always good - both when there are a lot of robots and when timeframes are small and the speed is more critical.

For example the bot trades according to the signals of the model and opens and closes trades. Sometimes they may be too long which may lead to a potentially big loss. You can optimize the protective stops, based on your own preferences. And all sorts of other nuances to improve trading performance.
 
mytarmailS #:
No, it is not implied
Since we like to talk about non-stationarity (or lack of periodic cycles), finding a group of parameters or choosing a model via an optimization surface will have a short-term effect, just like anything else. It makes sense, but it's not that meaningful, but it is there.
 
Maxim Dmitrievsky #:
Since we like to talk about non-stationarity (or lack of periodic cycles), finding a group of parameters or choosing a model through an optimization surface will give a short-term effect, like everything else. It makes sense, but it's not that meaningful, but it's there.

Yes! found "correct" maximum will give short-term effect and AMO will work for some time, but found "any" global maximum will not work at once

opened a question on CV

 
Ahaha, Maximka heard clever ideas from me about optimization and quickly made up an Expert Advisor, even made a video)))
 
mytarmailS #:
Ahah, Maximka heard a lot of clever ideas from me about optimization and quickly set up an EA, even made a video)))
That project was probably a year old. On other principles, which have not been discussed anywhere. I wanted to write articles, but I'm not ready yet, because it all very quickly goes to the market almost unchanged. You can, of course, limit yourself to theory, but I do not like theory.
 
Maxim Dmitrievsky #:
This project is probably a year old. On other principles, which are not discussed anywhere. Don't flatter yourself 😄 I wanted to write articles, but I'm not ready yet, because it all very quickly gets to the market almost unchanged. You could, of course, limit yourself to theory, but I don't like theory.

your optimization video from yesterday is probably a year old too ))))

 
mytarmailS #:

your optimization video from yesterday must be a year old too ))))

It's not my fault you only recently discovered it. Read my earlier articles, for example. Only the lazy man is not optimizing his stops. Well in general there is no desire to fight paranoia
1...258525862587258825892590259125922593259425952596259725982599...3515
New comment
Reason: