Machine learning in trading: theory, models, practice and algo-trading - page 1229
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 fact, it is done so, but you can vary the degree of "ideal equity", because the more ideal it is, the more overtraining
error on the tray: 0, on the oos: 0.4.
An "ideal" trade, taking into account OOS (inside), loss-making trades are only 15%, which corresponds to the amount of OOS (here 20%). It's not hard to guess what will happen with the new data
Forum on trading, automated trading systems and testing of trading strategies
Machine Learning in Trading: Theory and Practice (Trading and Beyond)
Maxim Dmitrievsky, 2018.12.24 09:32
The difference between the "Expert Advisor" and the "Trader's Strategy" is not so much the same as the trading strategy.
I tried to show that it is a wrong approach to train "ideal" entries, moreover to assign the same probabilities to all outputs
If the NS has qualitatively remembered the patterns of "ideal" entries in the tray and intensively trades at a loss on OOS, it means that the patterns are repeated out of order, i.e. the set of predictors does not characterize the signals and it should be changed or expanded or too low signal detection threshold is chosen, the size of this threshold, in theory, should allow to reduce the number of losing trades up to the absence of any.
If NS has memorized correctly the patterns of "ideal" entries in a tray and intensively trades at a loss on OOS, it means that the patterns repeat out of order, i.e. the set of predictors does not characterize signals and it needs to be changed or expanded or the threshold for signal detection is too low, the size of this threshold, in theory, should allow to decrease the number of losing trades up to the absence of any.
And the errors already show that there is no chance, well, I'm just talking about playing with the probabilities of signals before changing the threshold, so that the error can be adjusted for the track and test using them... automatically, just automatically... I do not want to select a set of predictors myself, it's not a lordly thing to do)
I don't see any counterarguments why it shouldn't be done this way
i can adjust classes and make them correlate probabilities with return fics, i don't know if it will work on automata or not, since toxic and others write about super duper method of correlation
Later maybe, I'm not a marketer and not a "guru", demagogy is not trained, few people here can explain about the gross errors, and the subtle ones... In my opinion, this is a very subtle error that you need to be aware of it in your own practice.
Later maybe, I'm not a marketer and not a "guru", demagogy is not trained, few people here can explain about the gross errors, and the subtle ones... If you have not understood them, or have made them yourself, you can use your own experience to correct them.
Forum on trading, automated trading systems and trading strategies testing
Machine learning in trading: theory and practice (trading and not only)
toxic, 2018.12.24 14:33
Tree cuts recursively point cloud, for classification entropy is taken as partitioning criterion, for regression mean square error, these are the basics, ask Rev. Innocent or Martin Chigewara for MO lectures, they give out for the thirsty knowledge.
About custom models in leaves, "extrapolation by forest", already there was a conversation about it in this branch, search, on leaves you do not average targetets (for regression), and build linear regression, then in ensemble you average not constants but outputs of linear reg.
I don't use alglib, as well as mql, I don't show my working code, and I'm too lazy to write a simplified example for you personally, sorry.
What are support and resistance levels
04:45 min.
*** and you won't understand it
In general, it is very instructive to watch all the video of the author, otherwise it will be very difficult to understand the essence
Do not look dumbasses already do not understand
In general, it is very instructive to see all the videos of the author, otherwise it will be very difficult to understand the point
Thank you for the warning. I won't watch it.
Thank you for the warning. I won't be watching.
)))) I don't mean you.
What are support and resistance levels
04:45 minute
Don't watch it for morons, you won't understand it.
In general, it is very instructive to watch all the video of the author, otherwise it will be very difficult to understand the gist
Reminds me of
Maxim DmitrievskyYuriy Asaulenko
You know perfectly well what the audience is here and who I mean so calm down it does not concern you ...
And to watch or not to watch, everyone's choice, just these videos give a lot of useful information if you know how to see it
for example the fact that to present the price in a classical time series or even as a BP may not be the right idea
Reminds me of
in appearance but not in content