Dr.Fx corner
....
Fs is a kind of digital filter, the properties of which are of interest.
P.S. I will open a EURUSD sell trade when the market opens.
The properties of the digital filter are not determined by the opening/closing of the trade. DF has other well-known characteristics, you have to start with them, they are AFC, FFC, etc. + Filters can be optimal (theoretically the best, let's say no better, cannot be), which is often not achievable in practice, and there is a consistent ... And before using TF in practice it is necessary to define at least what AFC and FXF of used filter is matched with...
In hope that it can turn out good (informative branch) I attach to the message lectures on TF. I hope they will be useful for somebody.
The properties of a digital filter are not determined by the opening/closing of a transaction. DFs have other well-known characteristics, you have to start with them, these are AFR, FFC, etc.
Prival, thank you. However, Prival is in his repertoire. I will correct him: Linear filters, or ADAPTIVE filters (that is = linear with Slowly (compared to something) changing parameters) have AFR, IFR, etc. But for arbitrary nonlinear algorithms the concepts of AFR/FFR are meaningless, so the concepts of lag (and lag measures), smoothing (and smoothing measures) - are not defined in any way.But, I emphasize, we don't need to limit ourselves at all to "linear filters with slowly changing parameters", any physically realizable (exactly - without peeking into the future) algorithms are available to us. Also - and this is fundamental - we should distinguish the nature of the following things: lag and transient.
+ Filters can be optimal (theoretically the best, let's just say it can't get any better, it can't be)
in its - very narrow - class - linear with slowly changing and no more. it should not be imagined that there are any fundamental prohibitions on the synthesis of filters better than Kalman family filters.
Don't limit, you can apply any filters and they have many characteristics, besides those mentioned above there are phase and group delay (lecture #2, lag is jargon) and these characteristics are in no way related to trade opening and closing.
Suppose it is non-linear, i.e. its characteristics change with time, for example the input signal spectrum changes ... filter's frequency response has been changed, etc. But this very characteristic of AFC and VSF is not going anywhere, it exists, it just depends on something. That's something you need to deal with first of all, and that's the answer to the question.
What are your filter's AFC and IFR coordinated with? Are these characteristics constant or do they change over time?
Regarding Kalman, it is good for a certain range of tasks and rather difficult to implement in practice. Outside of this range of tasks, there are many filters superior to it, moreover there are tasks to which the Kalman filter is not applicable at all.
But, I emphasise, we do not need to limit ourselves at all to "linear filters with slowly changing parameters", any physically realisable (exactly - without peeking into the future) algorithms are available to us.
Let's say it is non-linear, i.e. its characteristics change over time, for example the input signal spectrum has changed ... ... the filter's AFC has changed, etc. But this very characteristic of AFC and VSF does not go anywhere, it is
Privat, with all my extramural respect, get this crucial thing straight: for an arbitrary non-linear algorithm, the AFC/FF does not exist in principle. Not in the sense that it is impermanent and varies, but from the word "no".
The very concept of "physically unrealizable filter" seems to me farfetched and far-fetched.
I was writing about physically realisable filters. "Not" you added it yourself. The term is clearer, it's not a stretch, it follows from the principle of causality.
and in real time the future can be extrapolated and then looked into.
Hmm. If it were that simple, everyone would be a billionaire :-) there are naive people... I'll tell you a secret: if you have somehow "extrapolated the future", then don't waste your time building filters, run straight away on the "interpolation of the future" to trade.
I want to make it clear to everyone: the ultimate realization is NON-FUTURE. Here's a non-linear REAL filter that is absolutely physical. There are no bans (in the class of linear filters there is a ban, it cannot be implemented).
I wrote about physically realisable filters. "Not" you added it yourself. The term is clearer, it is not pulled by the ear, it follows from the principle of causality.
If there is a physically realisable filter, and there is a corresponding notion, then there must be a physically unrealisable one, isn't there?
And if someone has physically realised a physically unrealisable filter, then either the physics has changed, or the term "physically unrealisable filter" is incorrect. I know what the point is, but the terms should have been changed a long time ago, there are lies embedded in them. I hope it's accidental and not maliciously intended to hide the most interesting features of DSP.
Dr.Fx:
Let me tell you a secret: if you have somehow "extrapolated the future", then don't waste your time building filters, run to trade immediately on the "interpolation of the future".
if someone has physically implemented a physically unrealisable filter
There is no interest in discussing your posts. What you have written is impossible by definition.
Privat, with all my extramural respect, get this crucial thing straight: for an arbitrary non-linear algorithm the AFC/frequency response does not exist in principle. Not in the sense that it's impermanent, and varies, but from the word go.
....
For an arbitrary non-linear algorithm, maybe there isn't ....., but for of a non-linear digital filter this characteristic exists, just has to exist. It is very simple, there is a filter, so it has characteristics. Just like a person, as long as it (a person) exists, it has characteristics, weight, height, etc.
Don't confuse a digital filter with an algorithm, not every algorithm is a digital filter. The key word here is filter, and there is always the question of what is it filtering (what is that filter matched to) ?
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
I sincerely ask the moderators not to touch the thread. I am sitting quietly in the corner, not bothering anyone. I'm not selling anything, I'm not expressing any crazy ideas. I only intend to post some pictures. And the occasional comment on them. Who is not interested, will pass by, who is interested - read. Using tractor as an example, I see that this kind of threads, generally speaking, has a right to exist. Due to interest of most traders in EURUSD, I will start with it. The reason of starting this branch: discussion with some forum participants in private messages about nature and properties of somedigital filters.
So. The chart below shows 2*288 bars of the M5 timeframe, i.e. 2 days in time. EURUSD. Actual at the moment (as of Friday evening 1.5.2015). "A" is the original price (close), Fs is some kind of digital filter whose properties are of interest.
P.S. I will open a EURUSD sell trade when the market opens.