Correlation trading / calculation

Dear traders,

I was wondering how others look at correlation trading and calculation.

If you use correlation in either your algorithms or your trading, what period do you use, and what method do you use to calculate correlation?

Personally I use a same-timeframe period of 60 bars to calculate correlation percentages. This simply to avoid taking correlated entries in different pairs at the same time. For example, I don't want to go long on EURUSD and GBPUSD at the same time while their correlation is at 99%, because then I'm effectively doubling my exposure/risk.

However, if I see an entry for a BUY on GBPUSD while I have a long open on EURUSD, and my correlation is negative, then I'd definitely enter as the negative correlation indicates they act as opposites.

Interested in your thoughts and implementations,

Jos.

https://www.forexticket.co.uk/en/tools/01-01-correlation

That formula I believe is this one:

https://www.forexticket.co.uk/en/tools/01-01-correlation

hmm, I don't see a formula at the url you specified...could you be more specific PipGnostic? Thanks for the reply btw! :)

Sorry liquidity,

Posted the same link twice by the looks of it.

https://en.wikipedia.org/wiki/Correlation_and_dependence

I saw an interview with the owner of that website and he said he got the algo from wikipedia.

Quite important to know what's correlated to what. Can end up with all kinds of accidental exposure of one doesn't watch it.

  [i] liquidityFX posted:
   

However, if I see an entry for a BUY on GBPUSD while I have a long open on EURUSD, and my correlation is negative, then I'd definitely enter as the negative correlation indicates they act as opposites.

Interested in your thoughts and implementations,

Jos.[/i]

I'm wondering if you really understand what trading correlated pairs involves? To have a true correlated setup involving the eur and usd you would include the eur/usd, gbpusd and eur/gbp... however to round it out nicely I'd include the jpy/usd, eur/jpy, gbp/jpy ... 6 pairs in which all 4 currency's are included a total of 3 times.

Some others...
aud/usd, usd/cad, aud/cad, cad/jpy, aud/jpy, usd/jpy

Another... gbp/usd, aud/usd, gbp/aud, chf/usd, gbp/chf, aud/chf

hi pt49,

yes, I know one can also actually trade by correlation, that's also a very interesting and effective method. However, I currently only use it to avoid the additonal risk exposure...perhaps I can code a few correlation rules in at some point. Only issue with that is the inability to backtest it :(