Machine learning in trading: theory, models, practice and algo-trading - page 841
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
Yes, thanks. I am working on it.
We take a tick stream, but read it not every tick, but at exponential time intervals (more precisely, at time intervals obeying a discrete geometrical distribution with p=0.5). We have the simplest flow of events. Then we "sift" this initial VR. We obtain Erlang flows of different orders. It is necessary to input returns of these streams.
This experiment will unambiguously answer the question of whether or not we need BP thinning.
Everything is as clear as a day.
We take a tick stream, but read it not every tick, but at exponential time intervals (more precisely, at time intervals obeying a discrete geometrical distribution with p=0.5). We have the simplest flow of events. Then we "sift" this initial VR. We obtain Erlang flows of different orders. It is necessary to input returns of these streams.
This experiment will unambiguously answer the question whether it is necessary to thin BP or not.
Alexander, may I clarify once again, i.e. we take a tick flow, thin it exponentially (the algorithm from Habrahabr), write the quote at the set point in time, if there is no quote, write the previous value and obtain Erlang of the 1st order. From the resulting series, remove every second quote, we get 2nd order Erlang, 3rd order-remove every 3rd quote, etc. Thanks for the data.
Alexander, may I clarify once again, i.e. we take a tick stream, thin it exponentially, (an algorithm from hubrahabr), write a quote at the set point in time, if there is no quote, write the previous value, we obtain a 1st-order Erlang. From the resulting series, remove every second quote, we get 2nd order Erlang, 3rd order-remove every 3rd quote, etc. Thanks for the data.
That's exactly right.
Recommended reading:
https://www.mql5.com/ru/forum/221552/page317#comment_7108837
Recommended reading:
https://www.mql5.com/ru/forum/221552/page317#comment_7108837
Source data?
Baseline data?
Algorithm for collecting and sifting data see https://www.mql5.com/ru/forum/86386/page841#comment_7107322. Only we don't delete, but Leave it at Every K-quote, we discard the rest.
See https://www.mql5.com/ru/forum/86386/page841#comme nt_7107322 for the algorithm of data collection and sifting. Only we don't delete, but leave But we do not remove every K-quote and discard the rest.
Then you can easily reject the whole study. The results on different sources of tick data will be different.
Moreover, the misunderstanding of what is a tick is evident. On such a foundation it is nothing to talk about returns.
Knowing the basics of pricing, you can create any number of ticks.
Then it is easy to reject the whole study. The results on different sources of tick data will be different.
Moreover, there is a clear misunderstanding of what a tick is. On such a foundation it is nothing to talk about returns.
Knowing the basics of pricing, you can create any number of ticks.
I will not argue. But I recommend to try this method of preparation of predictors. The convergence to the normal distribution is obtained for all currency pairs. It can't be that I was just lucky with my broker and tick-flow.
I will not argue. But I recommend to try this method of preparing predictors. The normal distribution convergence is obtained for all currency pairs' returnees.
What is the reason for the distribution shown? Do you want to say that if we add any imaginary tick history with such a distribution, there is a TS (which one?), showing the graphicality of this history?
It can't be that I got stupidly lucky with my broker and tick flow.
This statement fits well with current geopolitical realities, but not with the scientific approach.