Author's dialogue. Alexander Smirnov. - page 43

 
LeoV писал (а): There are differences, but they are not significant, I think. And the period is not long =14. So the algorithm is of high quality.
Leonid, I have a question about JMA, which is in Code Base. About the meaning of the phase parameter. From the code we can see that it really makes sense to change the phase from -100 to +100. Outside of these values the indicator does not change, and a certain parameter f10 depending on the phase changes as a discriminative function between 0.5 and 2.5. As far as I understood visually, phase changes the "swing" of the indicator, and at the JMA extremums the difference between JMA(phase=100) and JMA(phase=-100) is maximal in modulo. It seems that somehow the derivative of the indikator itself is taken into account here. Maybe you know what role this parameter actually plays?
 
Mathemat писал (а): Леонид I have a question about JMA, which is in Code Base. About the meaning of the phase parameter. From the code you can see that it really makes sense to change the phase from -100 to +100. Outside of these values the indicator does not change, and a certain parameter f10 depending on the phase changes as a discriminative function between 0.5 and 2.5. As far as I understood visually, phase changes the "swing" of the indicator, and at the JMA extremums the difference between JMA(phase=100) and JMA(phase=-100) is maximal in modulo. It seems that somehow the derivative of the indikator itself is taken into account here. Maybe you know what role this parameter actually plays?

Well, it's true - the phase changes the oscillation amplitude and at extremes the difference between JMA(phase=100) and JMA(phase=-100) is maximal in modulo. I am attaching a screenshot - below is this difference.

 

This is the result of digging through the JMA source.

Both have a 'phase' of zero. Blue is standard JMA, green is modified. Somehow the green resembles QRMA - because of the differentiation spikes. The price to which all this is attached is PRICE_MEDIAN.

 
Mathemat:

This is the result of digging around in the JMA source.


You can consider me a fellow sufferer. I've never been into mash-ups, but after reading a lot of smart people, I got in. :-) It's the second time it's happened to me and the second time I've got evidence that even in this tired field one can find flowers.

I haven't had the task to look into Djurica's algorithm or JMA source code from codebase. I just had my own ideas. Having implemented one of them, I was convinced that the Golden Rule of Mechanics works. If we win in smoothness, we lose in phase delay. And vice versa. Nevertheless, using not too primitive algorithms (i.e. reducing friction losses :-) one can reach a tolerable compromise and find more or less suitable mask (if one really wants to use it).

For example. There are such advanced variants of EMA - DEMA and TEMA. As described by Bulashev, they take into account smoothing error and therefore have less phase delay. In particular, DEMA has less FS than EMA and TEMA less than DEMA. I have written an implementation of this algorithm for any arbitrary order. Increasing the order it is possible to decrease FP, but the indicator line tends to the price line and loses smoothness as a result. After some experiments I have found the ratio that is not too far from JMA in your picture and from Djuric in LeoV' s picture.

In both cases you can see that this one has slightly less FZ, but also slightly inferior smoothness. Probably by applying some external smoothing method with a sufficiently short period, you can achieve even more similarity. My point is that Djuric and JMA are good algorithms, but they are not unique and unrepeatable.

 
Having implemented one of them, I was convinced that the Golden Rule of mechanics works. If you win in smoothness, you lose in phase delay. And vice versa. <br / translate="no">
Very nice looking HMA in igorad's version - with complete linearity and the simplest idea (difference of two LWMAs with coefficients and different periods). Figure below. In blue - HMA(13, PRICE_CLOSE, MODE_LWMA), in green - slightly modified Djuric(9,phase=100). I don't see much difference, and I'm extremely surprised by it. Smoothness of both curves is very close, responsiveness to sharp movements is about the same, delays are also about equal.
 

And here is my "work of art", the aforementioned product of an idea. An adaptive, line-weighted WAMA. Not as smooth as Djuric, but as good as it in terms of FZ.

 
Hello all!
Mashka is a contagious thing. Every now and then you come back to them from time to time. You know in your mind that there is a limit and the golden rule (as Yurixx mentioned) will never be cancelled, but you never know if you have reached the bottom. I think the adaptive mouving is just as good as Jurika's:
 
Good pictures. It would probably be even better to show the mash-ups compared to the same Jurik with comparable visual parameters. But that's not even the point right now. We are intuitively accustomed to the following requirements for the mashups as such: smoothness, low FP, and quick response to gaps. And, let's not lie, practically perfect mash-ups have in fact already been created (HMA, JMA, regressions, DEMA, etc.).

Nevertheless, the "two wipers" system does not work satisfactorily with any wipers, not even with the best Juriks. And it is unlikely to work - until we understand what we need from the SYSTEM (the simplest system - i.e. just two wipers). An idiot's dream, of course, but it seems I haven't taken everything from the mash-ups yet to give them up for good...
 
I think the two-mask crossing has become standard because of its primitiveness. From my point of view MACD and MA-based oscillators are much more interesting. They allow, at least theoretically, to catch the price extremum. Provided that it is sufficiently smooth, it is not too difficult. Moreover, in this condition the extremum can also be caught directly, using a single MA. In this sense, the combination of low FZ and smoothness is actually quite interesting. And I think it is not worth counting on more, even in theory.
 
Mathemat писал(а) >>
...- until we understand what it is that we need to demand from a SYSTEM (simple system - i.e. just two mash-ups). An idiot's dream, of course, but I don't seem to have taken everything from the mash-ups yet to give them up for good...

the system is presented in the simple and intuitive form of one or two "simple" (system) mash-ups - why dream.... that's not a good way to put it...