Optimal moving average - page 15

New comment
 
FAQ:
No, Victor makes good turkeys :) I often use them myself. Have a look at his branch, there is a lot of interesting stuff there.
I've added it to my bookmarks, I'll read it tomorrow.
 

DmitriyN:


Then I'll share the maschka as well :) The most common Ma, with a period of 10, is based on a Ma with a period of 10, which in turn is also based on a Ma with a period of 10 and so on.
In short, all this is done without leaving MT, with the usual mask using the "Previous indikator's Data". In other words, it is a mask from a mask. I have not yet found an application for this smoothed MA.
In this case the screen shows a 4th order mask. The period is 4*10=40


Read AlexeyFX's posts, about shifting filters by delay bandwidth, it's either a cascade of cic filters, or adaptive filter shift value.

The same scale can look better if you shift it not exactly by the number of periods, but +/-, for example, not by 1 period, but from 0 to 1, you need to look for optimal position of the scale in this interval, which will give more information. So it will turn out, that the expression "a wrecker from a wrecker from a wrecker", will take a different turn, each shift of a wrecker is nonstationary, the shift, will influence the next averaging.

The point is to build an ideal smoothed line that does not distort the phase, and will be as close as possible or in other words equidistant from the spread of the price (if you shift it back)

PS: By the way, from what considerations can you pick up a chain from such ma ma, say ma50 from ma 123 from ma 220 from ma 263, i.e. unequal periods of ma.

 
But that is half of the problem, the wizard still needs to be stretched vertically to avoid vertical narrowing during averaging, at multicurrency the usual averaging introduces a wrong interpretation, everything collapses, indices from such simple wizards distort the true speed characteristics of the instruments, because instruments have different volatility, and averaging distorts / narrows this volatility not proportionally for different instruments. And the indexes should be based on the ratio of increments, that is, as if they were based on the speeds, only the speeds of increments contain information, and the cluster values of pairs are unnecessary information in the multicurrency analysis, in clusters and indexes, when they are constructed.
 
DmitriyN:
Didn't know you were doing this. Any luck?
Better than just crossing moving averages anyway. But I haven't done any research yet to be able to say for sure. I am experimenting with auto-optimization, it is conducted on several timeframes, and on several timeframes mabyma is taken.
 
Nikitoss:


I don't know if this is the one or the other, but this one is interesting too, I couldn't find any implementation in the base.

https://www.mql5.com/ru/forum/100105/page2

 

I'll put in my five bits and pieces. JMA.

 

No redraw? I don't know, sometimes I notice that they can post a mashup with its backward shifted, without showing the right edge))))

 

Does A (((x1+x2)/2)+x3)/2)+x4)/2,,,,,,, have the advantage of this averaging

than B (x1+x2+x3+x4)/2

????? in terms of reducing the impact on the dynamic properties of the final curve, of each individual reference on the whole sum. or in terms of where the reference stands in the series.

Say for series 1 2 3 4

А ((((1+2)/2)+3)/2)+4)/2 =2,875

but for row 4132 it will already be (((4+1)/2+3)/2)+2)/2=2,375, not only what is the sum of the row, but also what is the arrangement of the values in the row.

whereas at B (4+1+3+2)/2=(1+2+3+4)/2

 
Vasilisa:

Does A (((x1+x2)/2)+x3)/2)+x4)/2,,,,,,, have the advantage of this averaging

than B (x1+x2+x3+x4)/2

????? in terms of reducing the impact on the dynamic properties of the final curve, of each individual reference on the whole sum. or in terms of where the reference stands in the series.

Say for series 1 2 3 4

А ((((1+2)/2)+3)/2)+4)/2 =2,875

but for row 4132 it will be already (((4+1)/2+3)/2)+2)/2=2,375, not only what is the sum of the row, but also what is the arrangement of values in the row.

whereas at B (4+1+3+2)/2=(1+2+3+4)/2


It turns out like LWMA. And count the brackets in the examples! Useful! :)
 

I'm thinking that this kind of mashka(A) is probably closer to the word LF filter than the standard one. And more useful, as it doesn't hide the dynamics (probably).

Of course there are more computational costs, but it's not about that, it can be optimized later.

I would like to ask about advantages.

1...891011121314151617
New comment