Correlation

[Chris67]

arrhhh the age old topic rears its ugly head
I have been looking at CSI 's multi market analyser - they recommend when you look at correlation of futures markets you look at weekly bars, perpetual contracts and de trended !!!
Does anyone have any thoughts on the best way to look at correlation - if we trade back adjusted futures contracts then should we analyse back adjusted for correlation purposes

Any help warmy appreciated

#Chris

[painless]

I assume by correlation that you are comparing a cash market to a futures market.
If you test futures markets using a period based indicator then the data must be adjusted for the difference that occurs at the time when the new contract is concatenated to the previous one. If not then your chart and therefore your indicator will show a large gap that will upset your results.
Alternatively you can backfill the each new contract with its own data or adjusted data. Such a chart would show multiple contracts and thus multiple indicators.
There is an intrinsic problem with adjusting data and that is that if it is done perpetually then these differences can accumulate. Such data is not indicative of the real price although the change in price that it shows is correct. Many tools and trading methods will show erroneous results eg because your cash market will show new highs but your adjusted futures will not reflect this. You must understand the reason for the premium/discount between the cash and futures and factor this into your analysis.
I treat each market on a case by case basis and adjust my methods accordingly. For example I have done tests on Russell 2000 futures using the cash as the signal and therefore I could ignore the affects of the adjusted futures price.

[Chris67]

thanks for your reply painless

I was actually only talking about a basket of futures markets compared against eachother - for example a basket of 30 futures markets - i dont trade cash markets
working out correlation on futures markets can prove a minefield as i said there are so many different ways to do it

thanks

chris

[painless]

Oh ok!
If I wanted to do that with say coffee vs cocoa then I would use the cash market data. No doubt there are correlations there but trying to use the futures would give me a headache. I think if you can unravel all that then you deserve a medal.

[AFJ Garner]

Somewhere or other Sluggo made his usual thoughtful kind of post tabling correlations not of closing prices (or whatever) of the futures concerned but of the equity curves of different systems trading such futures. Which can be different and illuminating.