Gathering a team to develop an IO (decision tree/forest) in relation to trend strategies - page 9
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
Here on the site in posts, though the group itself can only be created from the terminal from the phone for some reason :) There's no sense to listen to automatons and other old guard, I have never heard anything useful from them
It's a pity this service only works via the phone for now, I'll have to think about an alternative site. Maybe someone knows a similar one? I think we need a kind of board where you can share pictures and edit them somehow, a separate chat room, and something like a reservoir of clever ideas.
Regarding the "old guard", if people want to do something, why not give them the opportunity? The purpose of the team is not so much to discuss ideas as to test those ideas in order to find rational seeds.
If I've never heard anything useful from them, I've never heard anything useful from Maxim Dmitrievsky. Parameters are picked up much faster via reinforcement, though with overfit also in seconds. There are no constant patterns in the time series in the market, they should be taken from adjacent correlated BPs, and then it will be more stable. In fact, everything is very simple, but everyone is doing the wrong things. But to understand it and to feel the futility and despair it is necessary to try. In any case, the more people the better, but not floodriders :)
I'm talking about the genetic method of building a tree. The principle of tree decomposition based on voracity (where the predictors for branching that cover the largest range are selected first), this approach may be generally correct, but there are often alternatives with small deviations that need to be tested. And, I'm not convinced in general that the maximum decomposition tree approach gives the best solution, fast and close to sampling yes, but correct... I'm not quite sure why different sequences would affect overtraining?
What do you mean by adjacent correlated series, can you explain by example, and why this should give more stability?
I would like to check my decomposition method, I think it has a rationale, I just manage to catch patterns that work for a long time in small numbers, but it's not clear how often they appear.
And good riddance to you, I have never understood people who at every opportunity try to turn to discussing personalities - Wiki to help "Ad hominem"
Why on earth would I want to discuss your personality? On the contrary, I don't know you well enough to discuss it. However, I can see the principle of your activity in this thread, which seemed to me to fit the slogan "They won't succeed, and I'll tell them that now". That's why I say that instead of having some kind of constructive conversation, you are generating principles that will not work, but what is your goal? Do you want to save time for everyone here, because you know that the world is perishable and unrepeatable, and this truth makes it impossible to describe it with mathematical formulas/methods? Thank you, if so, but lost souls must find solace and no one will do this work for them but themselves, understand?
Do you have low self-esteem? ...
I have low self-esteem. Does that help the discussion?
I don't want to be rude to people I don't know... These are your messages where I see your fantasies and?
oh yes, you are trying to change the situation, so to speak, you gather a team of professionals, you know how it looks from the outside: you have recruited smart and talented and let's do it.... You go over there, you do this, .... and I'll go around in circles.
)))))
You see my fantasies-- great! So all is not lost! A person deprived of imagination and unable to see other people's imagination is incapable of creating anything new. Such people suffer, waiting for doomsday.
Oh, yes, you are trying to change the situation, you gather a team of professionals, you know how it looks from the outside: you've recruited smart and talented and let's do it.... you go over there, you do this, .... and I'll go around in circles.
)))))
You should have said right away that you do not like working in teams, it is not nice when you have to do work that you think is stupid, when someone tells you what to do. Did I get your message right?
You should have said straight out that you don't like working in a team, you don't like having to do work that you think is stupid, when someone tells you what to do. Am I getting your message right?
Not correctly.
I have already written that to find the audience you need in the public domain, you have to try to give
You have even begun to draw block diagrams and share some code snippets - well, this is positive, but alas, there is no interest, no systematic approach - no interest in the presentation
Only your first messages on the first page of this thread were systematically formulated, but alas - they are not meaningful
ok, once again ... do not take me seriously, but alas, you come to me, and I can not always pass and do not respond, and this forum is what I need now in terms of finding the necessary information
The man is doing something similar https://smart-lab.ru/blog/353092.php
Yes, interesting blog, thanks.
His idea and the article "finding meaningful predictors through random tree analysis" is also worthy of attention and needs to be tested and explicitly automated.
No systematic approach - no interest in such a presentation
I'm not writing an article to spell out something in detail (and in my opinion, it can be reproduced point by point, can't it?). I'm not agitating for machine learning. Who is interested, asked what it is, I gave links where to look and read. Who is interested in my method, what I'm doing now, I'm ready to explain the essence of my thoughts in more detail.
However, I'm not interested in my method, but in the complex discussion and implementation of various ideas of applying MO (decision trees) to trading.
I didn't arouse your interest, sorry.
Great!
In what capacity are you ready to join?
I know about your experience in studying the question of MI, but I do not know about the programming - in what language do you work.As an observer for now. I program mostly in plush, java and sisharp, sometimes in python. I advise you to create a githab account for your project and post crafts there. Honestly admit that in "balablabla" as in branch about MO, to participate no time nor hunting, but with pleasure look at the ideas expressed in the code, if possible.
Although, of course, taking into account the specifics of our activities it would be surprising to see something grail in the public, but nevertheless different kinds of spare parts, fragments of ideas and approaches can be valuable.It's a pity this service only works via phone at the moment, I'll have to think about an alternative site. Maybe someone knows a similar one? I think we need a kind of board where you can share pictures and edit them somehow, a separate chat room, and something like a reservoir of clever ideas.
Regarding the "old guard", if people want to do something, why not give them the opportunity? The purpose of the team is not so much to discuss ideas as to test those ideas in order to find rational seeds.
I'm talking about the genetic method of building a tree. The principle of tree decomposition based on voracity (where the predictors for branching that cover the largest range are selected first), this approach may be generally correct, but there are often alternatives with small deviations that need to be tested. And, I'm not convinced in general that the maximum decomposition tree approach gives the best solution, fast and close to sampling yes, but correct... I'm not quite sure why different sequences would affect overtraining?
What do you mean by adjacent correlated series, can you explain by example, and why this should give more stability?
I would like to check my decomposition method, I think it has a rational point, I just manage to catch small number of regularities that work for a long time, but it's not clear how often they appear.
As an observer for now. I program mostly in plush, java and sisharp, sometimes in python. I advise you to create a githab account for your project and post crafts there. Frankly speaking I must admit that I have no time or inclination to participate in "balablablabla" as in branch about MO, but I will gladly look at ideas expressed in code, if it's possible.
Although, given the specifics of our activities, it would be surprising to see something grail in the public, but all sorts of spares, pieces of ideas and approaches can be valuable nonetheless.The point is that it is meant to be a closed (in terms of not public) circle of people involved in the work.
I like this approach:
1. Everyone throws their ideas into a common pot.
2. ideas are discussed and if something is not clear, questions are asked and clarifications made.
3. The more interesting ideas are chosen by voting, with the host of the idea attached
4. The carrier of the idea makes a presentation, where the essence of the idea is explained in more detail and an approximate TOR is formed
5. The participants of the team decide themselves whether they are ready to take up the idea or not, potentially who is ready to implement the idea completes the ToR together with all participants
6. There is the implementation of the idea in code with a group of people who have decided to work on it.
7. Intermediate results are distributed for testing and searching for mistakes in realization, code optimization
8. Other ideas which received fewer votes are implemented in parallel in the same way
9. Finished algorithms/modules are put into a common code piggy bank and may be used by anyone in the group.
So purely passive observation is not the best thing to get from a group member.
Perhaps more active participants will get extra points which will allow them to have more weight in voting.
There will also be ongoing referrals:
1. Searching for information on topics on various resources
2. Creating ready bases with chips and targets (on different tools)
3. Collecting results of algorithms - decision trees/forests for specific tools with a certain set of features and targets. Which will allow everyone to work with the results of calculations to test their ideas
You can create one on your phone and then correspond from the website. They also wanted to give them an opportunity to monetize, like selling memberships.
If so, you can, but a little later, while there are too few of us. Where did it say that? I can't find it.
Correlating instruments like sber and sberpref, why not make one a predictor for the other. The relationship is clear and the correlation of the chip with the target is good.
It's interesting. And what is the real time lag there, ie how long it takes raskorrelyatsiya before convergence as a rule?
If so, we can, but a little later, while there are too few of us. Where did it say that? I can't find it.
It's interesting. And what is the real time lag there, i.e. how long does the rascorrelation last before convergence as a rule?