Software products from Piligrimm - page 3
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
Why are the tests so short? At least show from the beginning of 2004. Better from 2001. And on all ticks, with 90% quality.
And now you're selling something that doesn't make any sense.
This is the first time I've heard of deep neural network analysis. What is that?
Why are the tests so short? At least show from the beginning of 2004. Better from 2001. And on all ticks, with 90% quality.
And now you're selling something you don't understand.
I'm selling indicators, not a trading system.
If you mean the test ofTRNSK trading system , which I attached to the subject, then this trading system was created specifically to demonstrate the possibility of using the offered indicators in TS, for the presentation of this subject. That these indicators can be used quite effectively in the creation of TS this example shows, and for this enough tester for the period shown, because I present at the moment not a trading system with any particular strategy, but only the indicators. When in the future I will put the TS for sale, of course there will be tests on both demo and real, confirming their work for a sufficient period to be able to make objective conclusions.
This is the first time I'm hearing about deep analysis with . What is it?
I would rather say deep analysis with expert systems based on neural networks, since neural networks are just the material for expert systems, while their core is the strategy underlying their operation, and the strategy effectiveness determines the accuracy of forecasts.
This is the first time I've heard of deep neural network analysis. What the hell is that?
:-)
This is the first time I've heard of deep neural network analysis. What the hell is that?
:-)
:))
Here is another picture, at this point in time. The yellow and pink lines have merged, which means there will be a short flat period.
Piligrimm
You are trying to sell a black box (several boxes) without revealing how they work. Indicators are a dime a dozen, and the ones you offer are no better than any other. If you use any indicator and its application to build TS, you have to know the principles of its work and you don't get it here. Maybe someone will buy beautiful pictures, but I would recommend you to show trading strategies that bring profit and use your indicators. They have been tested on the Forex market for years.
You use wavelet transforms in the basis of your indicators, but they cannot give price behavior prognosis (that is very important for profitable TS) because all wavelets are localized in the time domain. I want to explain, let's say you defined that at a given time a time series can be represented as a set of "Mexican hats" or something similar, but the prediction of this function in the future makes no sense, since it is equal to zero there. Second example, to clarify, the Fourier decomposition (basis functions of a sine wave) is localised to the frequency domain only, i.e. by defining parameters of a sine wave (set of sine waves) you can predict the time series into the future.
Piligrimm
You are trying to sell a black box (several boxes) without revealing how they work. Indicators are a dime a dozen, and the ones you offer are no better than any other. If you use any indicator and its application to build TS, you have to know the principles of its work and you don't get it here. Maybe someone will buy beautiful pictures, but I would recommend you to show trading strategies that bring profit and use your indicators. They have been tested on the Forex market for years.
You use wavelet transforms in the basis of your indicators, but they cannot give price behavior prognosis (that is very important for profitable TS) because all wavelets are localized in the time domain. I want to explain, let's say you defined that at a given time a time series can be represented as a set of "Mexican hats" or something similar, but the prediction of this function in the future makes no sense, since it is equal to zero there. Second example, to clarify, Fourier decomposition (basis functions of a sine wave) is localized to the frequency domain only, i.e. by defining parameters of a sine wave (set of sine waves) you can predict time series into the future.
When I started learning trading I tried all the indicators that were available, most of them just showed vague images that reflected the market events in a very abstract way. Many of them focus on manual trading and it's not always convenient to combine them and form a single input data for further processing by neural networks. Therefore, I decided to develop my own ones, which, on the one hand, would be visually appealing, on the other hand form data for further processing, and most importantly, would increase the information value of the input signal and create a presentable selection of signals derived from the main signal to be used, overlapping the maximum possible range of the main signal change . As you know, the market contains everything, but it is not so easy to select the necessary signals and remove the interfering ones. Long practical experience with time series to some extent helps me in this. If we compare the standard ZigZag indicator and my Trend indicator, which is close to ZigZag by its working principle, but differs from it considerably showing the trend dynamics besides pivot points, then the forecast using neural networks of the next ray is much more accurate using Trend indicator. All 4 latest indicators were developed in addition to each other and are maximally effective when used together.
If you have any questions about the working principle of the indicators, I am ready to answer them.
On the basis of these indicators, there are many strategies, and each developer has his own approaches and his own idea about how to trade. Besides, it is much harder to create an effective and profitable strategy than to write any indicator or Expert Advisor, and it costs incommensurably more, practically it is a trading system algorithm, and to program it is a matter of technique.
Wavelet transforms are used only in the first indicator and it is mainly oriented for hand trading. The second indicator is based on polynomials. I have already shown you examples of polynomials in various threads. But discretization uses not standard time intervals, but dynamically reconstructed in proportion to price dynamics, it has nothing to do with wavelet transforms or Fourier decomposition. This approach allows me to efficiently compress the input signal without loss of information content and create a representative sample for the neural network. As a result, I do not input a history of, say, 2000 bars, but only 100 samples; this proportionally increases the learning speed. Moreover, such sampling is in itself an efficient filter that removes random noise and insignificant fluctuations, because only signal extrema get to the sampling points. As a result of such compression I am able to work in real time with sufficiently deep history and make predictions on M1 at the arrival of each new bar, dynamically retraining neural networks in the interval from 3 to 15 minutes depending on the specified accuracy of learning and strategy used.
This is the essence of difference between indicators I am offering and standard ones that are abundant, but they are not oriented on cooperative work with neural networks.
New real time slide, just off the market, still warm. As I predicted from the indicator readings, it's going flat.