FR H-Volatility - page 38
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
Alas, it is not so simple. It's not even the width of distributions. We can build a confidence interval in which, say, 90% of reversals occur. And I have done such constructions.) It appears that the price may stay within this interval for a long time, which means that the problem is the prediction of the moment when the price leaves this confidence interval.
I agree with you, and I think the problem is, to a greater extent than you indicated, the width of the distribution. The uncertainty that arises because of the width of the distribution, does not allow you to aim at the enemy (in Prival's terms), and mistakes are costly.
P.S.
My colleagues, I think everybody got acquainted with this article, a fragment of which is displayed below (the archive was too small to fit in). I have a question about the algorithm of calculation of optimum number of inputs for NA based on Box-counting method. I attached a description of the method at the end of the article, after the list of references.
2 Neutron. Good afternoon . And on what time interval did you get the result?
Good!
I was building on TF=1 min.
2 Neutron. Good afternoon . And on what time interval did you get the result?
Kind!
I performed the builds on TF=1 min.
I see. What period are the statistics for, a week, a month, a year?
2 Neutron. Good afternoon . And on what time interval did you get the result?
Kind!
I performed the builds on TF=1 min.
I see. What period are the statistics for, a week, a month, a year?
How interesting!
In the figure from the previous post, I gave the dependence of the time of formation of the top of the zone on its amplitude for a step of construction H=10 points. I did not see any dependence. But, look at the result obtained for H=20:
The dependence is clearly seen! Yura, I take back what I said about the absence of the latter, you are absolutely right - the time of formation depends on the amplitude of the SZ.
to FION.
minutes from the beginning of 2004 up to now.
I think the distribution should have two value areas - one for the flat, the other for the trend, you can choose seasonally appropriate parts of the history.
There are two ways to do this: the first is universal... one by eye, the other by the standard deviation from the regression line over the calculation period.
Here are the preliminary results of converting the chart to bars with roughly equal volumes. This is so far in MS Excel:
Above is the traditional representation (based on astronomical time, H4, from 01.10.07 to 17.12.07), below are the equivolume bars for the same period. Start and end times of the periods are almost the same, so the stories are similar here too.
In general, there are no particular differences: same disasters in profile. Differences in the number of bars on the long history are aligned rather precisely.
But still something can be noticed. If one prints both charts so that the sizes are the same (it is not convenient to compare them on the screen), one can see slight differences in the ratio of trend and flat sections. Quite often (not always) the trend sections are a bit longer compared to the conventional representation, while the flat sections, on the contrary, are a bit shorter.
In my calculations I did not take into account the change of average volume per common bar in the long term. Probably, it should be done. Well, in short, there is still some work to be done...
P.S. Yura (Reshetov), Hi! Where did you disappear for so long?