FR H-Volatility - page 36
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
Well, that's impressive. As a first result - a quick sketch and a go - it's very good. It's not for nothing that I like ZZ. You can't give it to the input so directly...
What NS (if it's not a secret) did you build for it?
Well, I didn't feed it "so directly". I had to make it. It should be noted that these transformations allow me to restore the initial WP appearance without losing information, so they are of cosmetic nature, which, however, increases the accuracy of the prediction. I took NS as described in the article 'Predicting prices using neural networks', screwed around a bit with the number of inputs and neurons in the second layer, but it didn't change anything radically.
Has anyone measured such things for their ZZs?
I finally plotted the duration of the zigzag segments for my ZZ. The x-axis is the segment numbers in chronological order, the y-axis is the duration of the segments in minutes.
I.e. if you take the OIM for Y this will be the most likely lifetime of the zigzag. (The time from the start of its detection). By eye it's about 300 minutes, which is good
I finally plotted the duration of the zigzag segments formy ZZ. The x-axis shows segment numbers in chronological order, the y-axis shows segment duration in minutes. The unexpected (and sad) for me was the presence of a significant trend. The picture shows 2 approximations: linear and polynomial of 4th degree, the second one to demonstrate, that the linear one is good enough. The situation is similar for all the majors. The raw data is from 1999.
Has anyone measured such things for their ZZs?
Very interesting idea, just very interesting!
I finally plotted the duration of the zigzag segments formy ZZ. The x-axis shows segment numbers in chronological order, the y-axis shows segment duration in minutes. The unexpected (and sad) for me was the presence of a significant trend. The picture shows 2 approximations: linear and polynomial of 4th degree, the second one to demonstrate, that the linear one is good enough. The situation is similar for all the majors. The raw data is from 1999.
Has anyone measured such things for their ZZs?
A very interesting thought, just very much so!
The first question is to what extent this effect is due to a particular algorithm (which is why the words about their ZZs were highlighted).
I don't think it's worth stressing about it. IMHO, if you plot the same over a longer period (but with the same parameters), it turns out that mo just oscillates. The market does have different phases and they do change. And this change is not too noticeable to the eye, i.e. oscillation frequency is not too high.
I don't know, there are almost 8 years in this picture. So for real game horizons this is a real trend, and the fact that it could be some 100 year oscillation I would just disregard.
I don't think it's worth stressing over. IMHO, if you plot the same over a longer period (but with the same parameters), it turns out that mo just oscillates. The market does have different phases and they do change. And this change is not too noticeable to the eye, i.e. oscillation frequency is not too high.
I don't know, there are almost 8 years in this picture. So for real game horizons this is a real trend, and the fact that it could be some 100 year oscillation I would just disregard.
To be sure of this, it is sufficient to plot a moving average of the duration of the segments with a period of, say, N=100. The above chart corresponds to N=1, and the regression line takes into account all 3600 values. To see the local trend you need to take something in between. Then it will become clear that the stretching of the right edge is a result of a certain market behaviour in the vicinity of 3000 counts. If you don't mind, post it with such a dummy instead of polynomial regression.
Then it will become clear that the rising right edge is a result of a certain market character around 3000 oscillations. If you don't mind, post with such a dummy instead of polynomial regression.
It's not entirely clear why the 4th degree polynomial essentially ignored the vicinity of the 3000 reference. In principle, if I want to look for some correct(i.e. predictable) oscillations, I'll apply fft to the data to start with. But for now the trend issue is much more important to me, or rather whether the trend is a property of my algorithm or a property of the market. If you don't mind, post similar data for your favourite zigzag algorithm.
P.S. I would like to point out that short oscillations cannot cancel the trend in any way, only oscillations with a period significantly longer than the interval covered by the chart can cancel it.