Dependency statistics in quotes (information theory, correlation and other feature selection methods) - page 39
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
"Nevertheless, the problem of choosing a timeframe where there is less noise of any kind remains and needs to be researched."
Once the noise is identified, yes. Perhaps we should look for a timeframe or something. Before the "noise" is identified - it's a hypothesis. And to build a proof of one on the unprovenness of the other is probably not possible.
DDFedor says the same thing : The presence of "noise" depends on the research methods. You have to decide on the "method", and then probably talk about the "noise".
And how will you look for noise? You have to look through the timeframes anyway, if only to measure the "noisiness" of the predictive models, and count the noise as the residual modulo / modulo average of the forecast. For example, on a clock, the average modulo forecast would be 20 points and the average modulo error would be 23.
A deviation from the forecast is an error.
Passing off an error as noise is sly;) If the point of noise is to explain the error, then we go back - how can one prove one thing through the unproven other?!
Here is an example.
On 12/28/11 price breaks the trending pattern at 1.3060. My prediction is that the price will reach the 5th target (1.2629).
EURUSD Day
Days later, the price reaches the target, but goes slightly lower at 1.2623. Is the 6 pips difference between the forecast and the real price a result of the noise? Or is it a calculation error, a lack of model, methodology or a problem with the data?
From the breakdown of the trendline to the target is more than 4 figures. Exactly - 431 points. And 13 days. If it is noise after all, where is it hiding?
So what's your own proof ? Still no response to the post here https://www.mql5.com/ru/forum/135430/page34
So what's your own proof ? Still no response to the post here https://www.mql5.com/ru/forum/135430/page34
In this case, the "noise" is "in time". "Breaks the trend pattern", "starting point", "ending point"... And the "range"? What about the "frame" of the calculation? Has the "calculated" day of "arrival" been specified? Assuming, indeed, that 13 days is the settlement. "Before lunch" or "after lunch"? This is to say that in this case time is a "stray" parameter. I would like to ask the question - what are the "additional" parameters of the calculation? Isn't the calculation based on more than one factor? How are the criteria for cancelling a calculation for "reaching a price" determined?
I will be happy to answer some of your questions. But let's talk about them in another thread. Here we are discussing a specific method.
I want to explain - I use my own mathematical tool for convenience and clarity. Not more. If my charts lead away and distract me, then I will try to do with words.
In this post your thoughts and not one question ;) I have read it, your thoughts are heard.
OK, I'll ask more specifically.I certainly don't have a strictly mathematical proof, but nevertheless this proof I assume you don't have either, that TA is described strictly by mathematical methods and models.
You have no proof in formulas of TA consistency, you have only results that say it is consistent, only by experience.
Creation had at one time only empirical proof, he had no mate proof and no causality. He didn't need them, he didn't need them, he didn't need to understand WHY the stat advantage pops up, in order to take advantage of it and pull it by the ears, strengthening it, but making it rare at the same time.
You are demanding proof of randomness from the opponents.
Attention question. Do you yourself have proof of non-randomness, mathematically? After all, there is no limit to the number of experiments on which you base such questions, when you can count a certain number of test results as a proof barrier.