Filter without delay - page 16
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
I looked it up, of course. Nothing new. So what's it for? How to apply it?
It's good to hear there's nothing new for at least one forum member.
Alas, if you say so, it only means that you have stopped developing. Progress is being made and very noticeable. The DOS is :) well, there are no words.
Read the current state of the art https://ru.wikipedia.org/wiki/ERP
It's not about me. In the current terminology above I wrote about ERP. The difficulties are also indicated there: only for large companies, high cost, difficulties with implementation, etc. Before 85 there was socialism and similar systems, maybe worse or better had industries and prdp enterprises in the country! You, as a young specialist, got into an organization that was developing such products and became a specialist of another level. It took about 10 years. That was all over the country. And now Wikipedia.
Дело не во мне. В нынешней терминологии выше я писал о ERP. Там же указаны и трудности: только для крупных компаний, дороговизна, трудности с внедрением и т.д. До 85 был социализм и подобные системы, может быть и хуже или лучше имели отрасли и прдприятия страны! Вы, как молодой специалист, попадали в организацию разрабатывающую подобные продукты и стновились специалистом другого уровня. Уходило на это около 10 лет. Это было во всей стране. А теперь Википедия.
I see, talking about nothing.
Thanks again for the first link, didn't know.
You're welcome, I didn't know either. I found it through a search :)
Had a look, maybe not carefully. Yes applied wavelets. But why? what rinket characteristic are they trying to identify and what is it for? The main advantage is the analysis of non-stationarity. Is it available there? A lot of questions arise, especially if we take into account the unusual character of the tool.
First of all, one should formulate the problem. Or initial assumptions that can be checked.
It is stupid to take a wavelet, stretch it onto BP and try to extrapolate.
One should not take any wavelet, but a specific one that corresponds to the development of a real market process. Like with Elliott - there is a basis - several types of waves. And there is an attempt to decompose the entire series on this basis. But I would not try to be so universal. I would try to locally recognize one process to be able to fit into the rest of it. But we do not really need wavelets for that either.
How do you set your task? What is the market idea from which wavelets are logically derived?
Please, I didn't know either. I searched for it :)
First of all, you need to formulate a problem. Or initial assumptions that can be verified.
It's silly to take a wavelet, stretch it over BP and try to extrapolate.
One should not take any wavelet, but a specific one that corresponds to the development of a real market process. Like with Elliott - there is a basis - several types of waves. And there is an attempt to decompose the entire series on this basis. But I would not try to be so universal. I would try to locally recognize one process to be able to fit into the rest of it. But we do not really need wavelets for that either.
How do you set your task? What is the market idea from which the use of wavelets follows logically?
Let's start by identifying the BP:
I believe that there are several groups of investors in the market, variable in their financial power and having different interests at different points in time.
Depending on the temporal composition of the groups of investors by the number in the group and the time we get:
bullish and bearish trends of different duration and slope
sideways where forces are balanced
small noise generating
This is a descriptive market model, which we take as a starting point. It can be adjusted, but should not be transformed into other, more familiar forms.
Weillet analysis tries to obtain:
filter out the trends that are currently in place;
evaluate by their correlation to other trends the prospects of entering the market
Separate dead trends (which is impossible with Fourier analysis) from live trends and try to estimate their lifetime
Cleanse the trends from noise.
try to identify overbought/oversold
do reverse synthesis to get indicators
make forecasts of price direction
Actually, there are dozens of wavelet analysis functions of not very clear purpose, but we are sitting in our own boat - non-stationary VR
Actually, there aren't very many wavelets - less than a dozen. There are no proposals, and it's early days.
We need to learn how to transfer the quotes and parameters to matlab and then return the results to TS. At the moment the plan is as follows.
We start with a descriptive BP model and master the connection between MQL and Matlab.
faa1947 писал(а) >>
(blurb deleted)...
At the moment, the plan is roughly as follows.
We start with a descriptive BP model and learn how to interface MQL with Matlab.
Take it up at your leisure. Report your findings in writing.
Take it up at your leisure. Report your findings in writing.
>> I got it.
We need to learn how to transfer quotes and parameters to matlab and then return the results to the TS. At the moment the plan is as follows.
We start with a descriptive BP model and learn how to interface MQL with Matlab.
IMHO, that's going the wrong way round. Time and effort wasted.
The quotes saved in csv-file or in any text format that Matlab understands - take 30 min. to figure out and do the work.
After that, it will take another week to experiment with Matlab's built-in wavelet tools on the data. You will get some understanding of the tool (wavelets) and get rid of your illusions.
Next - solving simple problems as training exercises. For instance, smoothing out a price series to get something like a mask, but without a lag. The result - understanding of the method, understanding of the edge problem, deprivation of the last illusions - just one month.
After that, on the morning of the seventh day, you wake up feeling extraordinary clarity in your head - it means you can switch to something else with a light heart and clear conscience. Or, just the opposite outcome, you feel the arrival of specific ideas about the use of wavelets. Then, and only then, you start to study the language, write programs, transfer parameters, link to MQL, etc. But not before.
PS
By the way, why have you decided that continuous wavelets are good for you? What's wrong with discrete ones? After all, time in the market is discretised.
Let's start by identifying BPs:
I believe that there are several groups of investors operating in the market, variable in their financial power and having different interests at different points in time.
Depending on the temporal composition of investor groups in terms of the number in the group and the time they are in, we get:
bullish and bearish trends of different duration and slope
sideways where forces are balanced
small noise generating
This is a descriptive market model, which we take as a starting point. It can be adjusted, but should not be transformed into other, more familiar forms.
Weillet analysis tries to obtain:
filter out the trends that are currently in place;
evaluate by their correlation to other trends the prospects of entering the market
Separate dead trends (which is impossible with Fourier analysis) from live trends and try to estimate their lifetime
Cleanse the trends from noise.
try to identify overbought/oversold
do reverse synthesis to get indictors
make forecasts of price direction
Actually, there are dozens of wavelet analysis functions of not very clear purpose, but we are sitting in our own boat - non-stationary VR
Actually, there aren't very many wavelets - less than a dozen. There are no proposals, and it's early days.
We need to learn how to transfer the quotes and parameters to matlab and then return the results to TS. At the moment the plan is as follows.
We start with a descriptive BP model and master the interface between MQL and Matlab.
We may not even pass anything to Matlab. The wavelet decomposition is implemented as a dll that is launched from the indicator. The exchange with the dll through a csv file is easier and more reliable. But keep in mind that wavelets redraw, nothing good can be done without compression and averaging.
IMHO, coming in from the wrong side. Time and effort wasted.
Save the quotes in a csv-file or any text format that Matlab understands, figure it out and do it, it only takes 30 minutes of work.
After that, it will take another week to experiment with Matlab's built-in wavelet tools on the data. You will get some understanding of the tool (wavelets) and get rid of your illusions.
Next - solving simple problems as training exercises. For example, smoothing out a price series to get something like a mask, but without a lag. The result - understanding of the method, understanding of the edge problem, deprivation of the last illusions - just one month.
After that, on the morning of the seventh day, you wake up feeling extraordinary clarity in your head - it means you can switch to something else with a light heart and clear conscience. Or, just the opposite outcome, you feel the arrival of specific ideas about the use of wavelets. Then, and only then, you start to study the language, write programs, transfer parameters, link with MQL, etc. But not before.
PS
By the way, why have you decided that continuous wavelets are good for you? What's wrong with discrete ones? After all, time in the market is discretised.
I'd like to recall the fate of the Fourier package - all unfinished, closed to outsiders. So far no one has proven that you can't build a TC using Fourier (Burg). This is the result of the indulgence of one, two people. They tried it. It didn't work. Maybe someone else did. We need to make normal access to Matlab. There's a lot more out there, including a huge Fourier-related complex. One should simply master Matlab as a mathematical package for MQL - it is the only decent language that does not have a developed mathematical package of its own.