[R-SIG-Finance] Trading system correlation?

[Mark Knecht]

Hi,
   Sorry that I don't have any R code just yet but basically I'm
searching around a bit for some initial ideas. Please keep in mind
that I'm not a statistician or mathematician so it will be easy to go
over my head. Thanks in advance.

   Assume that I have 1 trading system and that I trade this system on
multiple markets, or the same market using different time frames. As a
futures oriented example:

@ES.D - 1 minute bars
@ES.D - 3 minute bars
@TF.D - 8 minute bars
@NQ.D - 5 minute bars

   My work flow optimizes the system for each market, then executes
Walk Forward processing to determine a reasonable set of weekly setups
for each system/market pair. I trade the systems for a week or two,
depending on the WFP results, and then update. In the end I generate a
set of trades for each pair over many years and save the trades to
files.

   I know the markets are correlated to a high degree most of the time
so I might logically expect the results to be correlated also. How can
I measure this?

Thoughts:

1) Create minute-by-minute equity curves and do correlation on those.
Downside is really too much data and therefore somewhat slow. Also it
doesn't really focus on entries and exits as much as both systems just
being in or out.

2) Somehow evaluate entry and exit times setting some sort of time
window. If the two enter & exit within a few minutes of each other
then call them correlated. (Is there some name for this sort of test?)

   My goal here is to find interesting sets of market/system pairs
that actually have lower correlation under the assumption that if both
make money but have low correlation then it would result in less
drawdown in the sum of equity curves. (I.e. - one system losing when
another is winning...) I'd like to test that assumption.

   Note that there are special time cases for entry and exit, such as
all systems exiting at the end-of-day market that I'd likely want/need
to filter.

   Thanks in advance for any ideas.

Cheers,
Mark