HOW TO TRADE FOR NOTHING ? - page 5
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
that's for sure... but the branch author's example is in fact almost a sine wave... and it doesn't take much intelligence to predict it... simple grids can easily do it....
http://s58.radikal.ru/i159/1007/66/036b70022eca.jpg
forecast (yellow line) for 24 bars (if you take a point from raw data as a bar) the grid is 2 layer...1 layer = 4 neurons...2 = 2...
http://s59.radikal.ru/i166/1007/63/ef9141d985cf.jpg
for 55 bars... 2 layers...1=6 neurons...2=2...
there is nothing to see in the scatter plot - it's all 1 to 1....
Anyway, the author's example is not a good one... just to play with it...
imho all ....
Can you do the second layer (output layer, right?) with one neuron?
Curious to see.
Can I ask you to make a second layer (it's an output layer, right?) with one neuron?
I'm curious to see.
https://www.mql4.com/go?http://i078.radikal.ru/1007/56/885cb9bceaa5.jpg
at 24 bars...first layer 4 neurons...second layer 1
https://www.mql4.com/go?http://s005.radikal.ru/i210/1007/47/e0b69b015bcb.jpg
at 55 the first 6 the second 1
I guess it would be possible to make just more neurons on layer 1 as well (where the first layer is 6 - no change in the prediction)... but in my opinion it's not that important... the data is inherently repetitive and easy to predict... even if you add more noise....
Vizard:
I guess we can just make more neurons on layer 1 as well (where the first layer is 6 - no change in the prediction)... but in my opinion it's not that important... The data is inherently repetitive and it's very easy to predict... even if you add more noise....
Well, actually, yes, that's exactly what I wanted to see.
Thank you.
Well, as a matter of fact, yes, that's exactly what I wanted to see.
Thank you. (chuckles)
you're welcome... It's a pity the market cannot be predicted like this ))) in fact - the grid remembers the training period and shows only what it has seen.... and this rarely coincides with the future behaviour of the market....
that's for sure... but the author's example is in fact almost a sine wave... and it doesn't take much intelligence to predict it... simple grids can easily do it....
https://www.mql4.com/go?http://s58.radikal.ru/i159/1007/66/036b70022eca.jpg
forecast (yellow line) for 24 bars (if the point from the raw data is taken as a bar) the network is 2-layer...1 layer = 4 neurons...2 = 2...
https://www.mql4.com/go?http://s59.radikal.ru/i166/1007/63/ef9141d985cf.jpg
for 55 bars... 2 layers...1=6 neurons...2=2...
nothing to see on the scatter plot - it's all 1 to 1....
So the author's example is not a good one... just to play with it...
imho everything....
Now that's interesting. Can I ask you to do the same for a case like this?
I took the original signal and added rnd(20) noise to it. As you can see from the drawing, the signal no longer looks so periodic. But still I can get periodic components out of it with Fourier. Hence, I can retrieve original function (with some precision, more noise less precision) and use it to make predictions.
How NS deal with this signal ? I would also like to see a picture if it is not difficult. Thanks in advance.
I'm a bit confused about the finished solution, the solution to what? I've attached the Matcad file, I didn't use Photoshop to draw it all. It takes 3 minutes to program it.
Here's the file you can check yourself, it shows how to use a spectrum to determine what's in the input... (I showed an example above), it's a bit different, but hopefully someone who's interested will understand it.
The so called optimum filter is programmed.
The question of the topicstarter was - what are the signs that determine the entry and exit of the trade. How to buy? How do I sell?
You have already said yourself that you need to find the kinks for it. And in Matkadec you have only calculated the spectrum, i.e. only half the job. That is why I asked about the final solution. There must be ready-made tools for analyzing kinks in functions in Matcadet.
The question of the top-starter was - what signs can be used to determine entry and exit from a trade. How to buy? How do you sell?
You have already said yourself that you need to find the kinks. And in Matkadec you have calculated only the spectrum, i.e. only half of it. That is why I asked about the final solution. Perhaps there are ready-made tools for analyzing inflections of functions in Matkadec.
If I reconstructed the original function without knowing it, then the inflection of the function is very easy to determine, no built-in matcadian tools are needed for this.
If the previous value of the function is less than the current value - then we go up,
If the previous value is higher - then we fall, i.e. there is a maximum or minimum between the sign change points...
If I reconstructed the original function without knowing it, the inflection of the function is determined very simply, no built-in matcadian tools are needed for this.
If the previous value of the function is less than the current value, we grow,
If the previous value is greater - then we fall, i.e. there is a maximum or minimum between the sign change points...
this is interesting. may i ask you to do the same for this case?
I took the original signal and added rnd(20) noise to it, as you can see in the figure, the signal no longer looks so periodic. But I can still pull out the periodic components by using Fourier, the blue arrows in the figure. Hence, I can retrieve original function (with some precision, more noise less precision) and use it to make predictions.
How NS deal with this signal ? I would also like to see a picture if it is not difficult. Thanks in advance.
Yes... I agree that there is less of a frequency pattern here... I can try it... but I need the signal itself (line) in csf or txt file... format = 1 column time, second column signal... + row a couple of times longer.... because it's too short....
The question of the top-starter was - what signs can be used to determine entry and exit from a trade. How to buy? How do you sell?
You have already said yourself that you need to find the kinks to do this. And in Matkadec you have calculated only the spectrum, i.e. only half of the deal. That is why I asked about the final solution. Perhaps there are ready-made tools in Matkadec for the analysis of function inflections.
https://www.mql4.com/go?http://i054.radikal.ru/1007/a0/08aaba6e7c05.jpg
the first thing that comes to mind, and the author probably got it.... blue= signal...black lag1 of this signal...on the crossover sell or buy.... with such a signal we will be in the + ... not in the market ))))
well Prival already wrote about it...