a trading strategy based on Elliott Wave Theory - page 41

 
Thank you solandr. And VG thank you very much.
 
Rosh!
У тебя такой красивый индикатор Мюррея, с надписями. Если не трудно, дай ссылочку откуда он.
Или что еще проще, кинь мне на мыло ANG3110@latchess.com


This is actually the Vladislava indicator, which is taken from www.mql4.com
only the caption has been inserted. You can get it here.



Yep, your picture is even prettier, some captions "Pivot and other". I specially looked at the properties of VG indicator - there are no parameters responsible for introducing additional labels. Although, maybe I'm wrong, I just did not look at the code.
 
<br / translate="no"> Three channels instead of one are also obtained according to the principles of the strategy. You most often have several areas in which you can build channels that meet the criteria. These are the areas where I choose the channel that has the lowest RMS. I also introduced a cutoff so that each next channel selected for plotting on the chart was at least 2 times as long as the previous one. If not cut off, there may be up to 7 closely spaced channels that would grossly blur the chart with lines. But with this truncation, we usually obtain 2-3 channels, which shows the picture quite clearly without clouding the picture with lines.


The hole is getting deeper, with more and more forks where the options diverge. Visually three zones can be seen, but it is unlikely to be so easily identified by the algorithm.

 
This raises the question of whether to sacrifice a smaller deviation (better approximation) or a longer channel (greater certainty)

 
This begs the question of whether we should sacrifice a smaller deviation (better approximation) or a longer channel (greater certainty)

Why should we sacrifice anything? For each of the zones, the channel with the lower RMS is chosen. You can't really say in the literal sense that a smaller deviation is a better approximation! More often than not the shorter channels have the lowest RMS. As they say, you need channels that are the best in their class (by number of bars). As a result the crossing of borders of different channels gives a pivot zone.
You can see three zones visually, but it is unlikely to be so easily detected by the algorithm.

I do not understand what the problems are in this respect. You have calculated the best channel for one zone and then for the other. Plot the best channels for your 3 zones on the chart. Made conclusions on the current market situation.
 
Hi Rosh,

What are StDev and StDev23 and how do they differ?
Something I've stopped understanding the situation. Especially the last 2 charts. Would it be possible to clarify ?

Thanks.
 
StdDev - RMS of the linear regression approximation error on the channel sample, respectively, StdDev23 - RMS on two thirds of that sample. The graphs show these values when calculating channels for a particular instrument, a particular timeframe and a particular location. This is an algorithm for selecting the right channel by the type of these RMS.
 
StdDev is the RMS of the linear regression approximation error on the channel sample, respectively, StdDev23 is the RMS on two thirds of that sample. The graphs show these values when calculating channels for a particular instrument, a particular timeframe and a particular location. I am talking about the algorithm for selecting the right channel on the basis of these RMS values.


Thanks, Rosh, I got it. As far as I understand, you are experimenting with an array of 1000 bars and the sampling for calculating the channel is fixed and naturally has a shorter length. Or do you build the channel on the entire 1000 ?
 
Yep, your picture is even prettier, some "U-turn and other" inscriptions. I've specially looked at the properties of VG indicator - there are no parameters that are responsible for the appearance of additional inscriptions. Although, I may be wrong, I simply wasn't looking at the code.


If you want you may make it a parameter - it's a matter of technique. With time you will get used to it and captions will be unnecessary.

2 Solandr&Rosh&ANG3110&Yurixx
I see that there are not so many inquisitive people here (alas, there are few of them everywhere), although more than I initially thought - that's good. I think there is no doubt about the system's performance ;). I propose to leave everything in the public domain at a level still sufficient to screen freeloaders. That is, do not publish ready-made solutions - at most a methodology and fragments of codes, and that all is not worth it ;). Methodologically, the information is enough to obtain a profitable strategy with a little effort. Otherwise, it will end up as with Murray - he is being sold for 80 quid a copy on the Murkansk continent without any remorse. I've received so many e-mails with questions :) and everyone was wondering how ?????? is in the public domain. :). I do not think (or rather I hope :) ) that a person who independently obtained a version of the system (which finalized, and not just stole the code) will sell it for 20 quid per kg. All details (and there will be more) can be sent via e-mail. It's much more efficient to create an MTSina (which I'm currently finishing, or rather hope I'm finishing :) and I hope you're not too far from this stage either. By the way, code exchange is not obligatory at all, as I see you get almost the same thing, except for a few small things, but that can be easily solved in practical use - the main thing is that the markings correspond to each other. As I said similar solutions within given errors and probabilities may be obtained by a few - the main thing is that the same forecasts are obtained by them) and then, if you want, translate the project into a professional commercial environment - better to manage the assets, even their own, even attracted the costs will not be so great and the profit is much higher (I hope everyone comes to the market to earn money), and some kind of monopoly on the method will remain :).

However, everyone chooses for themselves.
In any case, I reserve the authorship of the method for myself :).

Good luck and helpful trends.
 
StdDev - СКО ошибки аппроксимации линейной реггресси на выборке канала, соответственно, StdDev23 - СКО на двух третях этой выборки. Графики отражают эти величины при расчете каналов на конкретном инструменте , конкретном тайм-фрейме и в конкретном месте. Речь идет об алгоритме выбора нужного канал по виду этих СКО.


Thanks, Rosh, I understand. As I understand you are experimenting on an array of 1000 bars, while the sample for calculating the channel is fixed and naturally shorter in length. Or do you want to build a channel on the entire 1000 bars?


The quality of the sample will be essential here. If you choose a "fannish" length, you risk that you may miss a part of the already missed trend or a part of the current one. This will have all consequences.

Good luck and good luck with passing trends.