correlated CURRENCIES - page 2

 
Stanislav Korotky:
Some currencines ARE correlated (positive or negative). You can calculate correlation for currencies strengths - for example using your indicator.

Yes the indicator shows currency strength. It does not print correlation between currencies. This was the pol here to make.

But there have to be correlations because just ask yourself why there are pairs which are knows as mostly flat or consolidated like EURGBP or EURCHF. And on the other hand pairs which have good moves and are often trending like GBPJPY. The reason lies in the correlation of currency strength. We can call it like on a higher level. Don't know if exist already a name for it.

 
Bernhard Schweigert:

Yes the indicator shows currency strength. It does not print correlation between currencies. This was the pol here to make.

But there have to be correlations because just ask yourself why there are pairs which are knows as mostly flat or consolidated like EURGBP or EURCHF. And on the other hand pairs which have good moves and are often trending like GBPJPY. The reason lies in the correlation of currency strength. We can call it like on a higher level. Don't know if exist already a name for it.

Of course it does exist. And it's name exactly as "correlation of currencies". Having the "strengths" this is just a matter of one more step of calculations using these buffers as input data. Here is what I have right now for 3000 D1 bars of a cluster indicator (name it currencies strengths if you like):

:        USD    EUR    GBP    CHF    JPY    AUD    CAD    NZD
: USD  28.56   0.36  -5.37   4.10  17.75 -19.81  -8.08 -17.28
: EUR   0.36  21.60   0.48  11.15   0.42 -15.16  -9.55  -9.74
: GBP  -5.37   0.48  26.80  -6.86 -23.41   1.03   1.56   4.79
: CHF   4.10  11.15  -6.86  33.05  15.92 -19.49 -20.23 -17.69
: JPY  17.75   0.42 -23.41  15.92  73.23 -25.32 -23.51 -34.17
: AUD -19.81 -15.16   1.03 -19.49 -25.32  39.95  15.54  24.43
: CAD  -8.08  -9.55   1.56 -20.23 -23.51  15.54  31.20  12.93
: NZD -17.28  -9.74   4.79 -17.69 -34.17  24.43  12.93  36.93

This is not a conventional correlation though (otherwise you should see 1s on the diagonal), but a measure of volatility (for example, EUR vs CHF correlation is positive but mutual volatility is small, and EURCHF range is indeed relatively small). And JPY is most volatile in itself.

 
Stanislav Korotky:

Of course it does exist. And it's name exactly as "correlation of currencies". Having the "strengths" this is just a matter of one more step of calculations using these buffers as input data. Here is what I have right now for 3000 D1 bars of a cluster indicator (name it currencies strengths if you like):

This is not a conventional correlation though (otherwise you should see 1s on the diagonal), but a measure of volatility (for example, EUR vs CHF correlation is positive but mutual volatility is small, and EURCHF range is indeed relatively small). And JPY is most volatile in itself.

this table is interesting. So it means the more negative number we have the more volatile will be the pair.

Like GBP to JPY -23.41

But something doesn't seem right because USD to USD should be always 100. You must have a different calculation.

 
Bernhard Schweigert:

this table is interesting. So it means the more negative number we have the more volatile will be the pair.

Like GBP to JPY -23.41

But something doesn't seem right because USD to USD should be always 100. You must have a different calculation.

As I said, this is not a correlation (conventional correlation, if it was - indeed 100 or 1.0 should be on the diagonal), but something like a potential profit (in internal cluster units) you can make out of every combination.

When you compare USD/USD (USD on itself) with say JPY/JPY (JPY on itself) you can deduce that JPY is more volatile (JPY currency is more responsible for movements than any other currency combined with it in a pair, on average).

The sign of values defines if the "correlation" between currencies is positive or negative. Volatility itself is the modulo of the values. So +10 and -10 are equally volatile, but +10 means you should expect they move in same direction, and -10 means opposite direction.

 
Stanislav Korotky:

As I said, this is not a correlation (conventional correlation, if it was - indeed 100 or 1.0 should be on the diagonal), but something like a potential profit (in internal cluster units) you can make out of every combination.

When you compare USD/USD (USD on itself) with say JPY/JPY (JPY on itself) you can deduce that JPY is more volatile (JPY currency is more responsible for movements than any other currency combined with it in a pair, on average).

The sign of values defines if the "correlation" between currencies is positive or negative. Volatility itself is the modulo of the values. So +10 and -10 are equally volatile, but +10 means you should expect they move in same direction, and -10 means opposite direction.

It depends how you do measure the strength in the first place. But I agree JPY is more volatile than USD, still it surprices me EUR is lesser.
 
Bernhard Schweigert:
It depends how you do measure the strength in the first place. But I agree JPY is more volatile than USD, still it surprices me EUR is lesser.
This is just a sum of products of changes of corresponding cluster indicator lines (lines with currencies strengths) on a given bar interval. Nothing fancy. You can do it with your currency strength indicator as well. Mine is based on publicly available CCFp (but thouroughly updated).
 

The very definition of correlation is strength and direction. Volatility over time has little to do with it.

GBP is responsible for most volatility, not JPY, simply because it has the highest value. 1 pip move equates to 1.x pip move in any opposing currency.

So to infer that, because GBPNZD is the most volatile pair it is the least correlated, is a logical fallacy. A series/set of xy values is needed to determine strength and direction, preferably on the same scale (for instance with an oscillator).

Because currency pair comparison is a zero-sum equation, not all 28 combinations are needed.

 

We do not want to make it too much complicated. Ignore the volatility and get only the direction.

If currency 1 moves up and currency 2 moves up too over one bar its positive correlated and we can get the statistics.

I did not look into a code to get results jet. I just asked for traders opinion. That would be first hint. After can be a statistic but statistics always rely on the conditions been ask and can have so very different results.

 
Ernst Van Der Merwe:

The very definition of correlation is strength and direction. Volatility over time has little to do with it.

I'm afraid you're wrong. Correlation is formulated via covariance which is a sum of products of changes, observed on a value in time. And the changes in time (specifically their absolute values) is nothing else than volatility.
 
 Look at GBP/JPY AND EUR/JPY on the 1m. They are support correlated, but in no way shape or form do you have any sort of "edge" by looking at one instrument in deciding what the other will do. Why is that? Well because since they are correlated, they move exactly the same, almost tick for tick. Correlation only proves that this market is algo based and that volume can not be trusted once or ever.