Hearst index - page 10
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
...We are interested in the correlation coefficient between neighbouring samples in the first difference series of the original BP. It is the correlation coefficient that shows the dependence of the expected increment on the previous increments. It is this coefficient that is identical to Hearst shifted by 1/2.
Now I don't get it. Why do we need the first difference ? By doing this transformation over the original series, we kill the trace - something we can capitalize on.
Drawing again.
For red, QC equals 1, both by the commonly known formula and by yours. ACF is nice as well, it shows that the series is correlated with itself, all is fine, we may work.
Now we have the blue line, no trend, no KK, and according to your formula KK is negative (not growing, but falling). So, the conclusion is that it is better not to trade.
If only Hurst showed the trend so nicely when the first difference is at the entrance, it would be good. If you don't mind, try using Matcadet, you've already written what you've got.
Before using any kind of matrix machine. I always try to check it with known functions (models), then it becomes clear how and what is there. How to interpret the results, where the pitfalls, etc..
There seems to be some truthhere.
Tried to implement, but I got very close to 1 for the courses.
______________
I reread the article - I seem to have made a mistake.
For currency pairs the Hearst index must be calculated for a derivative, I calculated it for rates.
OK, thank you. That's the best there is. I spent all Sunday surfing the web and could not find it.
By the way, the article says that the exchange rates are close to 1, so you have the right result. But here's the conclusion (in the article) to use the difference in time series (stump). I do not understand why? The article contains test signals, and everything is OK, it's clear, it should be that way, at least it's logical. They have already started to make various transformations before analyzing BP.
Did Hurst measure water in the Nile the same way? Apparently not. Why should we kill the trend and then analyze it? I don't understand it.
Okay, thanks. That's the best there is, I think. I spent all Sunday surfing the net and couldn't find it.
By the way, the article says that it is close to 1 for currency rates, i.e. your result is correct. But here's the conclusion (in the article) to use the difference in time series (stump). I do not understand why? The article contains test signals, and everything is OK, it's clear, it should be that way, at least it's logical. They have already started to make various transformations before analyzing BP.
Did Hurst measure water in the Nile the same way? Apparently not. Why should we kill the trend and then analyze it? I don't understand it.
Here's a screenshot from Peters' book.
Rosh thank you. There's a word maybe. It's not a nice word. There's no justification, I don't see why we should take the difference, because we kill the trend. So we can take the difference if we want, but we don't have to. Arbitrary. What is right? Here everything is beautiful, good analysis. All tested on models. And then bam, we take the difference because it's close to 1.
Even this underlined phrase may be understood differently. That chart that we see is the daily change in price (every day changes), it doesn't say that this is the change relative to the previous day.
Why do we have to kill the trend and then do the analysis? I do not understand it.
Sergey, the trend is not killed by the first difference!
Take the first derivative of the linear function and get a constant - this is essentially the first difference and it's not equal to zero. How else can I tell you? I don't see what's bothering you with this first difference.
Sergey, the trend is not killed by the first difference!
Take the first derivative of a linear function and you get a constant - it's essentially the first difference and it's not equal to zero. How else can I tell you? I don't understand what's bothering you with this first difference.
That's how I showed it's a constant. There's a blue horizontal line in the picture.
And it doesn't bother you that before the conversion the AC was = 1 and became = 0. That the shape of ACF has changed. That according to your formula KK was = 1 and became -0.5 ?
From my point of view it is a mistake to assume that after conversion the characteristics of BP have not changed. It has changed and very much so.
Can you explain why we have to take the difference?
Z.I. And the trend gets killed, whatever it may be. It does. It was up. And it wasn't. It would have been down and it's gone too. Two different trends, one up, one down. And after the transformation is a constant. Draw a zigzag 500 bars up and then also 500 bars down. So what ? all the information is lost.
Z.I. And the trend gets killed, whatever it is. It does.
OK, it is killed, but its presence remains -- expressed as H > 0.5 .
_________________________________
I didn't fully understand the algorithm, the posted version didn't work correctly.
Two different trends, one up and one down. And after the transformation a constant. Draw a zigzag 500 bars up and then also 500 bars down. So what? All information is lost.
The autocorrelation coefficient in the row of the first difference, it is a tool that has a certain window width - through it the tool looks at the kotier and sees only "gradations gray" - warmer-colder. It is clear that if there is a change of trend direction within the window, the instrument will integrate it and will not notice it (like an average hospital temperature). But this does not mean that the method is not functional, you just need to select the appropriate window.
As for the transformation of the series, which is inevitable in finding its first difference, there is nothing in it from the point of view of TC. Indeed, by opening and closing a position we, in fact, try to catch the increments of series. Namely they determine our profit or loss in the market. Thus, the increments but not the absolute values of quotes are of primary importance for us traders and when passing to them we focus on the main thing and reject secondary ones.
This I have tried very artistically to find understanding with you, Prival.
a script for those who are not familiar with matcad
generates CSV (n, R/S), then easily calculates Hearst coefficient and V-statistics in Excel
a script for those who are not familiar with matcad
forms CSV (n, R/S), then it is easy to calculate Hearst coefficient and V-statistics in Excel
Is there a built-in Hurst calculation in Excel? If yes, please name it. >>Thank you.