Dependency statistics in quotes (information theory, correlation and other feature selection methods) - page 48
![MQL5 - Language of trade strategies built-in the MetaTrader 5 client terminal](https://c.mql5.com/i/registerlandings/logo-2.png)
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
Let's talk about the first post of the topicstarter. Did you find any incurable errors there?
Firstly, it is the price returns, not the price itself, that are being investigated. That's the object of the study. Looking for further arguments to show that price cannot be investigated in the same way.
Second, it is already done: divide the distribution of returns into quantiles (the topikstarter did it differently). Each quantile is a letter of the alphabet. There can be from 2 to infinity, but more than 40 is not reasonable, perhaps.
I read both the article and the first post of the topicstarter carefully. Only in the article I didn't see how information theory applies to the question under investigation. That's something, but just not the methodology of applying TI.
Of course, I could be wrong, I don't claim to be the final truth. Yes, TI uses statistical methods, but - in the sense of pattern recognition, it does not make predictions. It uses special coding methods to catch and correct recognition errors. Signal recognition probability, channel congestion probability, scales, fractalities are evaluated. Coding and encryption methods are investigated.
But the work of multipoints very clearly estimate the probability of formation of patterns of a certain type, the prediction in their model is a method of interscale interaction. True, there is a drawback - these methods are geometric and therefore difficult to formalize.
I am not against the study of increments, after all we get the profit from the increments. The question is how to correctly study the positive and negative increments and how to perform the cutting correctly. After all, the price movement is the sum of the increments. Letters are put into words, words into sentences and then into essays or novels. And the length of words, phrases, sentences varies. You have to determine how you will read - word by word, phrase by phrase or page by page diagonal reading. And only then determine the degree of influence of letters on pages in terms of syllables, words, phrases, etc. That would be the interscale interaction.
Quantiles you have taken from the ceiling. Answer, by what criterion of usefulness are they selected, why are there 40 and not 476, what is the expediency of this choice? Probably in the word "Perhaps".
It's not up to you to set the breakdown into ranges, it's up to the market. It doesn't care what you think it is. That is why the ranges will be different, both in price and in time. You have sliced the price chart into timeframes. The market doesn't care about this chart - it just draws its own one. That is why the opening and closing prices are meaningful only in the timeframes above the days; all that is below are just values in the quotation flow that have no statistical advantage over others.
Thank you, Nikolai!
Just let's refrain from making a hundred and fifty percent statements. We all have things to work on and strive for;)
P.S. Vadim says hello.
Ready to abstain, which is why I said I might be wrong.
Vadim is invisibly present in this thread.
The point is not clear... Does this mean that by isolating "only one price" from the candle, you can't make calculations? After all, "with a partial loss" is "waving". Or is that not what you mean? Moreover, the straight lines through the "centre of the candlestick's body" are very strong factors for price testing. They are as significant as the extrema. And in the case of a line built on the "doj-dodge" where "three in one" but "one price" is concentrated, it has even "the highest priority". This is to say that highlighting "one price on a candle" is ignoring (losing) the data.
You yourself said that it is an averaging, and by definition it is a loss of information. Of course it is possible to make calculations, but one has to wonder about their prognostic value.
Of course you can. Otherwise why the bar and other representation at all. I was talking about predictions like "*ussian" - you can easily replace the asterisk. It's tempting to invent a formal language and search for patterns in that way.
It's not a temptation, it's a methodology for applying TI
Yes, if only we had a reference signal!
Getting it is no less difficult than deciding on the intended type of signal itself.
A very sensible idea. A benchmark signal is a pattern, a pattern that should describe the market everywhere and always in the same way. I know such a pattern, and not just one. One of them is the Adversarial Tactics. By the way, the Elliott Wave Model is also one of those models.
How does this relate to the market?
How do neural networks relate to the market? Or wavelets? Or Fourier transform?
It is a mathematical method of describing a phenomenon. We try it on the market. That's all.
But not directly applicable to prediction, as described in the Russian example above.
A controversial statement
And how do neural networks relate to the market? Or wavelets? Or Fourier transform?
It is a mathematical method of describing a phenomenon. We try it on the market. That's all.
It was already a specific application of the model to the market - looking for repeating sequences, patterns or whatever you want to call it. A mathematical model of this, for example, "formal grammars". But that doesn't mean that any mathematical model can be applied to any practical problem.
A very sensible point. A reference signal is a pattern, a model that should describe the market everywhere and always in the same way. I know such a model. One of them is the Adversarial Tactics. By the way, the Elliott Wave Model is one of those models as well, but there is one problem - its repeatability is not 100%.
The most repeatable elements (not in the sense of specific values) on the chart are HCLO.
Everything else is a figment of the imagination with the hope of being adequate to the market.
I should read about Advers's tactics.
This was already a specific application of the model to the market - searching for repeating sequences, patterns or whatever you want to call it. A mathematical model of this, for example, is "formal grammars". But that doesn't mean that you can apply any mathematical model to any practical problem.
So, just not any, but a well-defined one. TA (Tactics of Adversity) is one of them.
Apply any mathematical model to any problem, and then evaluate the effectiveness of that application.
Click the button and you get the result.