Discussing the article: "Detecting and Classifying Fractal Patterns Using Machine Learning"
To avoid the need to calculate on different time scales, it probably makes sense to convert quotes into Renko beforehand.
In an academic sense, the approach is interesting (I would call it more specifically "mirroring", because as far as I understand, we had to move away from fractals and take a constant window length of 100, on which the more beautiful forward is), but imho the dependencies between the past and the future in the series of quotes are constantly and unexpectedly broken by external forces.
ZЫ. typo: "set in the range from 15 to 30", it should be up to 300.
To avoid having to calculate on different time scales, it probably makes sense to convert quotes to renko beforehand.
In an academic sense, the approach is interesting (I would call it more specifically "mirroring", because as far as I understood, as a result we had to move away from fractals and take a constant window length of 100, on which the more beautiful forward is), but imho the dependencies between the past and the future in quotes series are constantly and unexpectedly broken by external forces.
ZЫ. typo: "set in the range from 15 to 30", it should be up to 300.
I haven't thought out yet how to take into account the variability of fractals in time. Because MO starts to memorise them, and on new data exactly the same ones are not found. With a fixed window, the variability is apparently less - less to memorise :)
In time something will work out, because it is not known in advance what exactly should be done.you don't know in advance exactly what you're supposed to do
There was a James Simons, he knew exactly what to do.
Machine learning, AI, that's all great.
It's the human being who sets the tasks for the machine. Human consciousness, especially the unconscious, is very poorly understood. I'm not talking about the collective unconscious. But in my opinion, that is where the clue to the functioning of markets lies.
Did you happen to read Yamvlich and Rybakov's "The Bronze Bird"?
There was a James Simons, he knew what to do.
Machine learning, AI, it's all great.
It's the human being who sets the tasks for the machine. Human consciousness, especially the unconscious, is very poorly understood. I'm not talking about the collective unconscious. In my opinion, this is where the clue to the functioning of markets lies.
Did you happen to read Yamvlich and Rybakov's "The Bronze Bird"?
This article is about machine learning and fractals. Articles shape the collective unconscious.
I'm not just talking about this article. The article is not bad, within the mainstream. It's about something else.
"I have not yet figured out how to take into account the variability of fractals in time" - meanwhile, this is a key parameter that determines the effectiveness of any forecast.
And it is not only your problem, it is a global problem - the change of all coefficients under variables with time.
In order to understand the essence of the problem, you need to go higher, to rethink the initial concepts. For example, most of fractals are not self-similar, 1 dollar in 2000 is not equal to 1 dollar in 2025 (i.e. 1 is not equal to 1).
Many more examples can be given, in society (economy) Pareto distribution prevails, not Gauss distribution, so most statistical methods are not applicable to market analysis, etc.
Simons' success suggests that there is a solution to the problem, but you have to look elsewhere.
I'm not just talking about this article. It's not a bad article, within the mainstream. It's about something else.
"I have not yet figured out how to take into account the variability of fractals over time" - meanwhile, this is a key parameter that determines the effectiveness of any forecast.
And it is not only your problem, it is a global problem - change of all coefficients with variables with time.
In order to understand the essence of the problem, you need to go higher, to rethink the initial concepts. For example, most fractals are not self-similar, 1 dollar in 2000 is not equal to 1 dollar in 2025 (i.e. 1 is not equal to 1).
Many more examples can be given, in society (economy) Pareto distribution prevails, not Gauss distribution, so most statistical methods are not applicable to market analysis, etc.
Simons' success suggests that there is a solution to the problem, but you have to look elsewhere.
He seems to have a point about arbitrage. Many arbitrage strategies also stop working over time.
Hilbert's?
In general, there is almost no detailed information about Simons' working methods, and this is understandable. But it is known that he doubled his capital every year, and by the end of his life his fortune was estimated at over 20 billion.
But it's not about him, it's about the very possibility of finding a formula. Multidimensional spaces is today's terminology for Pythagorean ideas. It's a very deep topic. Multifractality can also be seen as some primitive analogue of multidimensional space, where vertices and graphs are projections onto a graph of hidden motions. If the topic is interesting for you, I can share my considerations - developments, but it is better in individual correspondence.
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
Check out the new article: Detecting and Classifying Fractal Patterns Using Machine Learning.
In the first article, we have examined in detail the fundamental aspects of the multifractal market theory. We noted that price charts are capable of forming certain repeating structures under the influence of external information that organizes them. Market participants create a complex dynamic system that has memory elements that take the form of certain market symmetries (patterns). These patterns may evolve over time or may repeat. Due to the self-similarity of fractal market structures, patterns can be expressed across different time scales.
This article presents an original approach to identifying and classifying fractal patterns. The analysis will be conducted in Python, with the ability to export final models to the MetaTrader 5 terminal in ONNX format.
Author: dmitrievsky