|
PureBytes Links
Trading Reference Links
|
Trey,
Measure the correlations of your percentage change in equity for the two
markets. So if you have periods where one market is trading and the
other is not, you input 0 for the percentage change in equity in the
market that is not trading. Then you do not have to worry about
different sample sizes or margin differences as long as both markets
trade over the same time period.
Gabriel
-----Original Message-----
From: Trey Johnson [mailto:dickjohnson3@xxxxxxxxxxxxxx]
Sent: Wednesday, October 27, 2004 3:56 PM
To: omega-list@xxxxxxxxxx
Subject: RE: Correlation
Gabriel,
Thanks for the reply. Price wasn't the issue, just the dollar
value of the historical trades. Pardon my ignorance, but could you
elaborate on, 'just calculate correlation based upon your returns'? Do
you mean something like taking the dollar value of each trade and
dividing it by the margin for that market, thereby converting to a
percentage return on margin? This seems to address the issue of
comparing trade results from different markets because a $500 dollar
trade per contract in natural gas isn't the same as $500 dollar trade
per contract in treasury bonds. However, this doesn't address the issue
of different sample sizes. The reason I'm worried about this is because
anytime I've seen correlations done the sample sizes are of equal value.
Thanks, Trey
-----Original Message-----
From: Gray, Gabriel [mailto:gabriel@xxxxxxxxxxxxxxx]
Sent: Wednesday, October 27, 2004 1:28 PM
To: Trey Johnson; omega-list@xxxxxxxxxx
Subject: RE: Correlation
Trey,
Just calculate correlation based upon your returns and ignore the
underlying price.
Gabriel
-----Original Message-----
From: Trey Johnson [mailto:dickjohnson3@xxxxxxxxxxxxxx]
Sent: Wednesday, October 27, 2004 11:34 AM
To: omega-list@xxxxxxxxxx
Subject: Correlation
Hello List,
I want to measure the correlation coefficient between the
historical trades of two different markets with the same system. I'm
doing this in order to determine whether I should add a market or not to
a portfolio, but I have a couple of questions. I've gotten the idea from
the book "Money Management for Futures Traders" by J. Nauzer. In his
example, he compares a series of historical trades between gold and
silver and it appears that the trades results are for single contracts.
If one is comparing the trades from different markets, shouldn't the
dollar volatility be held constant, by adjusting the number of
contracts, so that 'apples are being compared to apples' so to speak?
Why does he seem to ignore the time aspect of the trades, dates entered
and exited? Let's say over a 5 year period, I have 100 trades in one
market and 150 in the other, both are from the same system which is
either long or short, never flat. The trades obviously don't line up
with each other. Does this matter as it relates to the correlation
coefficient? If so, how does one work around this problem? Thanks, Trey
|