Discussion of article "Creating Non-Lagging Digital Filters" - page 3
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
You have written a new article with ideas. Some of the pictures reminded me of my own research. I'll have a closer look later.
Actually, it is not quite correct to speak about "non-lagging" filters without specifying what this "non-lagging" is applied to.
That is, it is assumed that there is a signal and interference. But it is necessary to be able to divide their mixture into components and determine their characteristics.
The test with a step to identify the so-called "non-delay" is, pardon me, "there is an elderberry in the garden, and an uncle in Kiev".
The step test is used to determine such important characteristics as inertia, oscillation, transient time, and link order.
There is such a concept of "transport lag", which can be visualised in the form of a conveyor belt. Apparently, by some misunderstanding, this concept has been transformed by econometricians into "non-lag".
By the way, the output of the reaction above the step, followed by its convergence, indicates that the filter is an oscillating link. ( and what "non-delay" has to do with it is not clear at all).
Actually, it is not quite correct to speak about "non-lagging" filters without specifying what this "non-lagging" is applied to.
That is, it is assumed that there is a signal and interference. But it is necessary to be able to divide their mixture into components and determine their characteristics.
The test with a step to identify the so-called "non-delay" is, pardon me, "there is an elderberry in the garden, and an uncle in Kiev".
The step test is used to determine such important characteristics as inertia, oscillation, transient time, and link order.
There is such a concept of "transport lag", which can be visualised in the form of a conveyor belt. Apparently, by some misunderstanding this concept has been transformed by econometricians into "non-lag".
By the way, the output of the reaction above the step, followed by its convergence, indicates that the filter is an oscillating link. ( and what "non-lag" has to do with it is not clear at all).
For a trader, as I understand it, non-lag is a timely reaction to a trend change. All other "non-delays" are irrelevant, even if they are strict from the mathematical point of view.
For cluster filters, traditional tests make no sense at all. A cluster can contain dozens, hundreds, or thousands of filters. Only the inventor of the filter can know which of them will work, for example, for a step. Not even the inventor of the filter, but the inventor of the approximation algorithm. For example, in the above test with a step, at least three filters worked: the first one drew horizontal sections, the second one reacted to the step itself, the third one reacted to the fall of the momentum line. In general, it's a mess.
Thanks to the author for an interesting and informative article, but I thought there was some inconsistency.
If the purpose of the article is to describe methods of combining known indicators(digital filters) into clusters, then why the statement about non-lagging?
After all, no combination of delayed signals is impossible to get ahead of time, it is like making one fresh cutlet from several yesterday's stale patties.
If the article is about the possibility of creating full-fledged non-lagging filters, which are mentioned in the conclusions, then where is the description, at least of theoretical foundations?
Probably the secret is in the words "potentially possible", which characterise not the objective possibility of creating filters, but exactly the subjective potential of the author, about which this article is intended to tell the readers).
I would like to formulate the following important idea.
According to the results of my long games of trying to create a non-lagging filter, I came to the following conclusion: only such a filter can be non-lagging if its value at the moment Ti is determined by the price values ONLY at that moment Ti. For example, the goal is to filter EURUSD, so the value of a non-lagging filter at the i-th moment can be oriented to anything, even to NZDCHF, but ONLY to the values at the moment i.
If it is not so for the topikstarter, then the filter can be non-lagging with a probability of 0.00001% in my opinion.
The step test for detecting the so-called "non-delay" is, pardon me, "the elder tree is in the garden, but the uncle is in Kiev".
The step test is used to determine such important characteristics as inertia, oscillation, transient time, and link order
After all, no combination of delayed signals can get ahead of time, it's like making one fresh cutlet out of several yesterday's stale patties.
Exactly. Moreover, I would like to clarify. For example, instead of SMA(100) - a moving average of 100 points, I took the construction 0.5*(SMA(99)+SMA(101)). The resulting chart may be even smoother than the SMA(100). But the lag has increased. Let's continue. Let's introduce SMA(98) and SMA(102), 97 and 103, and so on. Visually, the curve will remain almost the same as SMA(100), only smoother. For trading purposes, in fact, we can say that the lag is like SMA(100)/ But in fact (and it will be easy to see it on the step), the lag (transient time, if you will) corresponds to SMA(103).
To denkir - Matcad, of course.
(1) If the purpose of the article is to describe methods of combining known indicators(digital filters) into clusters, then why the statement about non-lagging? After all, it is impossible to get ahead of time by any combination of delayed signals, it is like making one fresh cutlet out of several yesterday's stale patties.
(2) If the article is about the possibility of creating full-fledged non-lagging filters, which are mentioned in the conclusions, where is the description, at least of theoretical bases?
(3) Probably the secret is in the words "potentially possible", which characterise not the objective possibility of creating filters, but the subjective potential of the author, about which this article is intended to tell the readers).
1. The purpose of the article is to demonstrate one of the approaches to working with streaming data. In a cluster, filters are combined not to determine the signal, but to track, if I may say so, changes in the characteristics of non-stationary series and the formation of probable values of the signal. Not only lagging indicators can be used as filters, but also all those that are not lagging, but are redrawn. Or various transformations, for example, wavelet transform. The signal itself is determined at the very last moment (see the scheme).
2. The very little theoretical background is given at the very beginning of the paper.
3. There is no question of the author's subjectivism. Reread the article from the section "The Effect of Advance Smoothing". Watch the video. Full objectivism. You will be able to try it yourself after the publication of GMomentum_test.
Transition time is the lag.
Transition time and delay are NOT the same thing. They are different entities.
The established terms are:
1) Transition time
2) Lag
have universally accepted, well-established and well-defined meanings.
To equate them is to misunderstand them.
Confusion in terminology leads to mutual misunderstanding.