Thoughts on some of the absurdity of multi-currency analysis. - page 29

 
AlexeyFX:

Exactly, apart from minor variations due to spreads and digital noise.

Then why not take IND_AUD = AUDUSD and IND_USD = {1} - constant.
 
hrenfx:
Then why not take IND_AUD = AUDUSD, and IND_USD = {1} - constant.


Because then it will become unclear why we should invent indices at all, to equate them to currency pairs afterwards.

And there will be no orthogonality here.

Are there such indexes that IND_AUD/IND_USD is not equal to AUDUSD? Then what to do with such indices, how to use them?

 
AlexeyFX:

Are there such indexes that IND_AUD/IND_USD is not equal to AUDUSD? Then what to do with such indices, how to use them?

For example, EURX / USDX!= EUR / USD.

Tell me about the orthogonality condition.

 
hrenfx:

For example, EURX / USDX!= EUR / USD.

Tell us about the orthogonality condition.


What a good idea... I didn't know about such a euro index.

It is difficult to tell about the orthogonality condition without a formula. When the indices are orthogonal, it is possible to say whether EURUSD grew at the expense of EUR, USD or both. When the indices are not orthogonal, we cannot say anything of the kind, it just creates more confusion.

 
I would still like to have a conversation on a technical level. In your screenshot above, do the indices have zero correlation with each other?
 

I can't make an absolute zero. As close to zero as I can get from the available data.

 

Since Ind_XXX / Ind_YYY = XXX / YYY, it all comes down to the task of finding only one VR ZZZ/USD: Ind_EUR = EUR/ZZZZ, Ind_AUD = AUD/ZZZ, Ind_GBP = GBP/ZZZ, etc. Then the equality Ind_XXX / Ind_YYY = XXX / YYY will always be fulfilled.

Obviously, the BP of ZZZ/USD can be anything, either SB or constant (non-zero).

It turns out that the indices(ZZZZ-majors) obtained this way have the minimum possible correlation.

The mathematical interpretation of this looks like this:

There is such a New BP that if it is added to the original BPs, the resulting BPs will have a minimum-possible correlation.

 

It is possible to compare the mutual correlation for:

  1. USD-majors: EURUSD, GBPUSD, USDCHF, ...., USDUSD ( = {1}).
  2. JPY-majors: EURJPY, GBPJPY, CHFJPY, ...., JPYJPY ( = {1}).
  3. etc.
  4. ZZZZ-majors: EURZZZZ, GBPZZZZ, ZZZCHF, ...., ZZZUSD.
 
AlexeyFX:


Exactly, except for minor deviations due to spreads and digital noise.

Not exactly a grail, but something close to the limits of technical analysis. I'm going to make a robot out of this, digital filters and some other stuff.


Tick multicurrency analysis allows for a more accurate entry and increased frequency of trades. (but you have to program a lot and handle ticks correctly....-hemorrhoea though useful).

And what is the frequency of deals by one pair per day (two, three or more) on average in your case (in your method of analysis) if it uses not ticks but minutes?

 
hrenfx:

The mathematical interpretation of this is as follows:

There is such a New BP that if it is added to the original BPs, the resulting BPs will have the lowest possible correlation.



I did not introduce such a concept as VP ZZZZ, all calculations are done "on the fly" for each bar. But in general, everything is correct.