Machine learning in trading: theory, models, practice and algo-trading - page 1252
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
And here is the result - specially on the basis of real ticks made
Obviously, I proposed a different concept of model creation, perhaps something similar is used in catbust, when the trained sample searches for rules, and the test sample approves these rules, so I had training on part of the sample, then checked the obtained rules (leaves) on the whole available history and the same selection, but in fact it is more efficient in terms of the number of leaves in the model. Yes, there is no test on a completely independent (not involved in creating leaves(training) and model selection ) sample. The main idea is that the fewer leaves are used to make decisions in the model, the more stable the model will be if the number of events described is large enough.
The trading system works on bar opening, therefore ticks do not play such an important role here.
I remembered that I only had history on file through October 2018, now I will test on November and December - that will be an independent sample.If by the opening prices it's ok, it is desirable to reduce the number of signs, leaves or whatever... less fitting.
Happy old dog year which is passing by, and to all the strange pig year which is approaching :D
Doesn't that problem seem to be solvedhere ?
Thanks, but isn't it about the bridge between python and R? However, I don't believe in perfect model, but I want to take useful information from received models - leaves or pieces of trees, and kjtbust is interesting first of all because it works much faster than the R script I use.
if opening prices are normal, preferably yes, reduce the number of signs, leaves or whatever... less fitting
Happy to all with the passing, some strange year of the dog, and with the coming of not less strange pig year :D
I join congratulations, may the new year bring lots of working ideas and may it bring us closer to our cherished dream!
And here is the result - specifically on the real ticks made
Not bad!
What brokerage company? There are 20 million ticks on 34 thousand of minute bars. 600 ticks per minute on average (daytime ticks should be 3 times more frequent than night ones). I have never seen anything like that
Not bad!
What is your brokerage company? You have 20 million ticks on 34k minute bars. At 600 ticks per minute on average (daytime ticks should be 3 times more frequent than night ones). I have never seen anything like that
I do not have a DC, but a broker, I trade on the exchange Moex - tool - bonding Si.
And here's the result - I made it using real ticks on purpose
I don't think so, but your screenshot doesn't look like the use of NS, MO and other heresies ;)
Do you have a trailing stop? It looks like a TS with a well-matched trailing stop, I think I've seen such tests with trailing stops on ATP
A little tip, just the other day I got interested. For those who understand mathematics and stoch. analysis. Streams (date-streams) and their integration into Rough paths (iterated integrals). Used for BP transformation, including machine learning.
Kind of sounds like Erlang.
https://en.wikipedia.org/wiki/Rough_pathIt's strange, but your test screenshot doesn't look like the use of NS, MO and other heresies ;)
Is there trailing? it looks very much like TC with well-matched trailing, I think I've seen such tests with trailing on ATR
Yes, trailing by Donchian channel is used (as in the preparation of a sample for training). Closing by SL only here, the result may be improved by using smart TP of course, but I still cannot finish this idea.
A little tip, just the other day I got interested. For those who understand mathematics and stoch. analysis. Streams (date-streams) and their integration into Rough paths (iterated integrals). Used for BP transformation, including machine learning.
Sounds a bit like Erlang.
https://en.wikipedia.org/wiki/Rough_pathAnd then how do you transform it in real time? Or is it just a function output with a window that you can apply or what - tell me in your own words, for those who don't understand higher mathematics.
You get these things in the output
Why do I see squiggles at the ends backwards?