From theory to practice - page 528

[Violetta Novak]

Smokchi Struck:

For me, a channel is noise superimposed on a certain trajectory.

The trajectory of a trade channel can be:

Mash can only cope with a horizontal trading channel.
It is important to find a function that can handle all these types of trajectories.

You know what came to mind, maybe it's stupid, but with all these trajectories you mentioned, the best way to cope with the price itself, I'm in no way teasing you, you look, and I want to help, by the way there are many intelligent and very smart people reading this forum, we are no match for them, maybe someone will be brave enough to express their point of view.

[[Deleted]]

Novaja:

You know, what came to my mind, maybe it's stupid, but with all these trajectories you mentioned, the best is the price itself, I am in no way teasing you, you are looking, and I want to help, by the way there are many smart and very intelligent people reading this forum, we are no match for them, maybe someone will have the courage to express their point of view.

There is no doubt that the best price indicator is the price itself.

At least it never lags!

[Yuriy Asaulenko]

Novaja:

Yuri as always))

Mm-hmm.) The rest is a trade secret.

Generally, there are a lot of them in the kodobase. Once, back in 2008, I started "fixing" a MAshka with this -https://www.mql5.com/ru/code/8538 . I don't use it for a long time now, but it may help someone to think).

[Violetta Novak]

Yuriy Asaulenko:

Uh-huh.) The rest is a trade secret.

In general, there are a lot of them in the kodobase. Once, back in 2008, I started "fixing" a MAHKA with this -https://www.mql5.com/ru/code/8538 . I don't use it for a long time now, but it can suggest some thoughts).

The eternal problem of which is better, Butterworth or Kalman filter, Prival was a plus for Kalman.

[Yuriy Asaulenko]

Novaja:

The perennial problem of which is better, Butterworth or the Kalman filter, Privall plumped for Kalman.

I like Butterworth better, and for good reason.

But that train, at the link, is no longer relevant - long since passed.

There are a lot of "improved" MAs in the kodobase, but I can't say anything definite about them - I don't know, I haven't tried them.

[secret]

Smokchi Struck:
When using the waving, can't the channel trajectory change quickly? The point is to find an indicator that will work better than the waving.

Mashka is a non-parametric indicator, it does not build any assumptions about the "function" of price movement, it simply works (as it can) with what it has. If the trajectory changes, it will follow it.

And let's say a parabola assumes that price always follows a parabola. And at the point of an abrupt change in direction, the parabola will be blatantly wrong.

Yes, the wristwatch does not work well with trends, there is no concept of a trend in its price model. Well, we can take a Holt-Winters noodle, it takes trends into account. But it will not change much, it will start lying again when the trend changes. You can introduce trend reversal into the dummy model, but it will start making noise where there is no trend. And so on.

It is necessary to filter the market - to take only flat areas, on which the ordinary waving works well. A model of the whole market cannot be built anyway.

[[Deleted]]

secret:
This is a non-parametric indicator, it doesn't make any suggestions about the "function" of price movement, it simply works (as it can) with what it has. The trajectory changes - the wristwheel follows it.

And let's say a parabola assumes that price always follows a parabola. And at the point of an abrupt change in direction, the parabola will be blatantly wrong.

Yes, the wristwatch does not work well with trends, there is no concept of a trend in its price model. Well, we can take a Holt-Winters noodle, it takes trends into account. But it will not change much, it will start lying again when the trend changes. You can introduce trend reversal into the dummy model, but it will start making noise where there is no trend. And so on.

It is necessary to filter the market - to take only flat areas, on which the ordinary waving works well. A model of the whole market cannot be built anyway.

Well said! The only thing that remains is to clearly separate the flat sections from the trend ones.

I propose to do it using a wiffle ball with a large period...

no?

[Yuriy Asaulenko]

Cat Libre Black:

Good words! All that's left to do is to clearly separate the flat sections from the trend sections.

I suggest to do it with the help of a long period...

Nah?

No! You'll detect a flat when it's already over.) At the expense of the MA delay.

I was beginning to think in the right direction, but got away from it.

[[Deleted]]

Yuriy Asaulenko:

Nah! You'll find a flat when it's already over). At the expense of the MA delay.

Wrong answer! We will detect a flat when it is unclear whether it is over or not.

So I suggest we suggest a sensible way of detecting a flat so that your suggestion of "take only the flat sections" can work.

[Yuriy Asaulenko]

Cat Libre Black:

The answer is wrong! We will detect a flat when it is unclear whether it is over or not.

So I propose to suggest a sensible way of detecting a flat so that your suggestion to "take only flat areas" can work.

That's not my suggestion).