--- In [email protected], "janhausd" <janh...@...> wrote: > > Hi Angelo, > > I have in fact just recently contacted Herman regarding this subject, > since running that code on my data resulted in mismatching > correlations, where Correl(A,B) was not equal to Correl(B,A) due to > the Foreign() function and data holes between A and B. The result is > actually somewhat similar to the the picture herman posted in that > link, where you see that the correlation between AAPL,ALTR is -0.624, > whereas ALTR,AAPL is 0.421, except when I ran the code, all of the > pairs mismatched and not just the ones on the outer edge of the > correlation square. Anyway, if you have data without mismatching > times, this shouldn't be a problem :) >
Thanks jan, this problem is clear to me now. I'm on my way to build a pair system on amibroker, using as a template the code posted by TJ on this list some time ago. My code is pretty simple and open a position for symbol1, and - at the same time - an opposite position for symbol2. Of course ,in the report this trade is counted as "two" and this makes unuseful of all the statistics based on the total nauber of trades (e.g. percent winners... profit per trade... ). As anyone thought of a way of telling Amibroker that - in this particular case - the 2 trades representing both leg of the pair should be counted as one?
