Forester #: Not a lot of deals. And you can't see their distribution over time. Maybe it's like this: That Python doesn't have a graph builder in the timeline? It's not that complicated. I draw my own graphs in my browser.
Why are you attached to this time :) algorithms for finding patterns are being discussed.
What to do with them - everyone does what he wants :)
Different algorithms find the same patterns, but in different bad ways, you see.
Aleksey Nikolayev #: Quite logical, as it corresponds to the classics (Dow theory).
It seems like you're discussing refining a modeling approach to improve results by incorporating specific filters and a more focused training set. If you'd like, I can help clarify or reframe this concept to fit the ongoing work in your document, or provide additional suggestions for implementing it effectively.
I think different patterns exist and can be found. In my example above there are 3 white swans and 1 black swan. I would like something more stable.
It seems to me that TP/SL regularities are such in a weak sense. For example, if a symbol does not have long trends, it is easy to get overstayers, where TP is much smaller than SL.
But overturned TPs (not necessarily round-the-clock) are strong patterns.
Waqas Khan #: It sounds like you are discussing refining the modelling approach to improve results by incorporating specific filters and a more focused training data set. If you'd like, I can help clarify or reformulate this concept to fit the current work in your paper, or provide additional suggestions on how to implement it effectively.
The specific topic discussed was the use of machine learning for trend-following trading systems in which rare large profits are interspersed with frequent small losses. Such systems do not lend themselves well to formalisation within a machine learning framework.
It seems to me that TP/SL laws are such in a weak sense. For example, if a symbol does not have long-lasting trends, then it is easy to get over-situators with TP noticeably less than SL.
But overturned TPs (not necessarily round-the-clock) are strong patterns.
I don't remember what I got the screen with swans on (TP/SL, probably). I'm studying the reversal one on ZZ now. Just filtering deals. More stable. Approximately like Maxim's. But something interesting is on very strong movements from 1000pts and one deal can exist for 2 weeks. From the minuses: you can wait a long time for the first deal to appear, you wait a long time to close. Deals are few out of 800 deals in the markup, there are 200-400 remaining for 8 years. That is not very representative. I will hardly trade such trades. On movements of 200 pts and below nothing interesting has been found so far. I wanted to search for ZZ with threshold 10-20-50 pts in ticks, as you did, but I was drained at the speed of markup.
I see my results in the timeline, and there's nothing I can do about yours to see them the same way. Unfortunately.
Aren't you curious what it looks like in time yourself?
I get constant cognitive dissonance from this forum for some reason. When writing one thing, in response translation to other topics. All the time :)))
I am interested in many things, I can't describe everything....
Forester #: Just filtering the deals. More stable. Approximately like Maxim's. But something interesting is on very strong movements from 1000pts and one deal can exist for 2 weeks. From the minuses: you can wait a long time for the first deal to appear, you wait a long time to close. Deals are few out of 800 deals in the markup, there are 200-400 remaining for 8 years. Which is not very representative.
Some in the classics (not MO) put in the optimisation criterion such conditions that long positions are considered either always unprofitable or their profit is penalised as a function of duration.
Also for variants with a small number of trades OnTester = 0. And other logical tricks to prevent swans from flying in, etc. I have a lot of input-parameters in the optimisation criteria. Due to their settings, I can search for different types of regularities. Including them in Optimisation is a bad solution, because a "better" pattern will be found.
Allow the use of cookies to log in to the MQL5.com website.
Please enable the necessary setting in your browser, otherwise you will not be able to log in.
You are signed in. Two-factor authentication has been disabled. Set it up again and scan a new QR code, or continue with two-factor authentication disabled.
Not a lot of deals. And you can't see their distribution over time.
Maybe it's like this: That Python doesn't have a graph builder in the timeline? It's not that complicated. I draw my own graphs in my browser.
I see my results in a timeline, and there's nothing I can do with yours to see them the same way. Unfortunately.
Don't you yourself wonder what it looks like in the timeline?
I think different patterns exist and can be found. My example above has 3 white swans and 1 black swan. And I'd like something more stable.
Quite logical, as it corresponds to the classics (Dow theory).
I think different patterns exist and can be found. In my example above there are 3 white swans and 1 black swan. I would like something more stable.
It seems to me that TP/SL regularities are such in a weak sense. For example, if a symbol does not have long trends, it is easy to get overstayers, where TP is much smaller than SL.
But overturned TPs (not necessarily round-the-clock) are strong patterns.
It sounds like you are discussing refining the modelling approach to improve results by incorporating specific filters and a more focused training data set. If you'd like, I can help clarify or reformulate this concept to fit the current work in your paper, or provide additional suggestions on how to implement it effectively.
The specific topic discussed was the use of machine learning for trend-following trading systems in which rare large profits are interspersed with frequent small losses. Such systems do not lend themselves well to formalisation within a machine learning framework.
It seems to me that TP/SL laws are such in a weak sense. For example, if a symbol does not have long-lasting trends, then it is easy to get over-situators with TP noticeably less than SL.
But overturned TPs (not necessarily round-the-clock) are strong patterns.
I'm studying the reversal one on ZZ now. Just filtering deals. More stable. Approximately like Maxim's.
But something interesting is on very strong movements from 1000pts and one deal can exist for 2 weeks. From the minuses: you can wait a long time for the first deal to appear, you wait a long time to close. Deals are few out of 800 deals in the markup, there are 200-400 remaining for 8 years. That is not very representative. I will hardly trade such trades.
On movements of 200 pts and below nothing interesting has been found so far.
I wanted to search for ZZ with threshold 10-20-50 pts in ticks, as you did, but I was drained at the speed of markup.
I see my results in the timeline, and there's nothing I can do about yours to see them the same way. Unfortunately.
Aren't you curious what it looks like in time yourself?
Just filtering the deals. More stable. Approximately like Maxim's.
But something interesting is on very strong movements from 1000pts and one deal can exist for 2 weeks. From the minuses: you can wait a long time for the first deal to appear, you wait a long time to close. Deals are few out of 800 deals in the markup, there are 200-400 remaining for 8 years. Which is not very representative.
Some in the classics (not MO) put in the optimisation criterion such conditions that long positions are considered either always unprofitable or their profit is penalised as a function of duration.
Also for variants with a small number of trades OnTester = 0. And other logical tricks to prevent swans from flying in, etc. I have a lot of input-parameters in the optimisation criteria. Due to their settings, I can search for different types of regularities. Including them in Optimisation is a bad solution, because a "better" pattern will be found.
I think different patterns exist and can be found. In my example above there are 3 white swans and 1 black swan. I would like something more stable.