For those who have (are) seriously engaged in co-movement analysis of financial instruments (> 2) - page 29

New comment
 
hrenfx:

Here is the correlation indicator shown above. It instantly calculates hundreds of thousands of CCs for any size sliding window. The readings coincide with the matrix packets 100%.

What counts with you is a mystery.


I've seen it.

WHAT DOES SMA HAVE TO DO WITH IT!!!?

How can you use sma's to calculate correlation?

How can you average. It's perverse.

It's been proven that if you use sma's to calculate correlation and to average two series - you get nothing more than a beauty. The correlation between 1.3333 and 1.51111 etc. almost imperceptibly already falls under 100% coincidence.

 

Where in the calculations did you see the SMA?!

If this is what you mean:

ExpKoef - коэффициент для построения экспоненциального сглаживания КК. Используется для оценки мат. ожидания КК

It has nothing to do with calculating QC. It is written that it is used to estimate the QC expectation (already calculated QC series).

 
hrenfx:

Where in the calculations did you see the SMA?!

If this is what you mean:

It has nothing to do with calculating QC. It is written that it is used to estimate the QC expectation (already calculated QC series).


I rejected the use of the CMA at once, as it does not work.
 

It is always possible to check - compare with the result of some matrix package. Pearson's QC is such an elementary thing, it is implemented everywhere.

P.S. The connection between EURUSD, GBPUSD and EURGBP was added for some reason. Such a connection is everywhere. For example, XAGUSD, XAUUSD and XAGXAU.

 
new-rena:


I had that... in the beginning. The point is that you have to take and apply Pearson's formula for two series at the same time and get one graph and don't forget to bring the pairs to the same denominator.

--- At what values does your correlation indicator open-close?

You can see if it's more or less than zero. Close and then... almost open.


I think what Renat means is a trivial correlation calculation, like the one I did back in the day in Excel (Pearson formula there), old screenshots are still there:

 
hrenfx:
It is always possible to check - compare with the result of some matrix package. Pearson QC is such an elementary thing that is implemented everywhere.

All right. Let's suppose. Then what does the correlation indicator on the chart you presented tell us, i.e. how do we trade and what do we get out of it?
 
Cmu4:


I think what Renat means is a trivial correlation calculation, like the one I did back in the day in Excel (Pearson formula), old screenshots are still there:


In general, let's put it this way. The formula is the same, only I take the correlation residual, i.e. for example we have two row values 1.3 and 1.32 plus 1.5 and 1.52. We find the minimum and add up what and to what: 1.3-1.3=0 and 1.32-1.3=0.02 plus 1.5-1.3=0.2 and 1.52-1.3=0.22. And so "N" times at the moment. We move to another point and again. Here after that we look at the correlation result. Otherwise - we consider correlation of almost identical values, and it is already not clear and often misleads.

.

 
  1. Building QC on a sliding window of 8 values - no representativeness. If you want 8 hours, take at least 480 M1 values. Your indicator may not be able to handle such calculations, because calculation of thousands of Pearson QC values for large sliding windows is a very resource intensive task.
  2. The subject of Pearson QC is completely covered on this resource. Moreover, there is even something that can hardly be found anywhere else - QC for more than two BPs.
  3. Trading based on the Pearson QC is paired trading (spread trading). The topic has been ridden over and over. Right down to automatic calculation of lots and position directions.
  4. Moreover, this resource even has statistical arbitrage as a generalized case of paired trading (not for two, but for more financial instruments).
 
hrenfx:
  1. Building QC on a sliding window of 8 values - no representativeness. If you want 8 hours, take at least 480 M1 values. Your indicator may not be able to handle such calculations, because calculation of thousands of Pearson QC values for large sliding windows is a very resource intensive task.
  2. The subject of Pearson QC is completely covered on this resource. Moreover, there is even something that can hardly be found anywhere else - QC for more than two BPs.
  3. Trading based on the Pearson QC is paired trading (spread trading). The topic has been ridden over and over. Right down to automatic calculation of lots and position directions.
  4. Moreover, this resource even contains statistical arbitrage as a generalized case of the pair trading, but not for two, but for a larger number of financial instruments.


Oh, come on. We have been there.

Number 8 is better in terms of equity. 480 min is also OK. My calculations of 5000 are not too shaky either. Nobody forbids working with data arrays. As soon as I set the best parameters for equity and specify the sampling size, I will not start counting everything since 1999 in real trading. For example, I will take 480 minutes and that's it.

 

Throw in the source. What is there to be afraid of?

1...222324252627282930313233343536...38
New comment
Reason: