In your experience as a system trader... 80/40 or 40/40 ?

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2040
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner » Fri Mar 23, 2007 3:03 pm

Yep all of my 30+ positions went against me for days. So much for diversification and non-correlation. But thanks to not getting stopped out, I'm now back in positive territory for the year. It could have all ended differently of course.

What a roller coaster.

tovetski
Senior Member
Senior Member
Posts: 46
Joined: Fri Jan 13, 2006 10:49 am
Location: Montréal, Québec, Canada

Post by tovetski » Fri Mar 23, 2007 3:23 pm

Good point, your example couldn't be any clearer and fresh to mind.

Have you done such research on historical JOINT distribution of daily returns?

I'm not sure how one would go about creating synthetic price series that could reflect both each market's daily return distribution and the correlation of each of those daily returns between each market.

Since the correlation between markets varies on a daily basis, do you:

1 - Compute correlation for each market using yesterday's data.
2 - Then sample daily returns for each markets by filtering on current correlation level. (If yesterday's correlation for a market is positive, only consider the historical daily returns of that market where the correlation of the previous day was also positive.)
3 - Project price series forward with sample obtained.

One could probably put historical daily correlations into norrower bins then "Positive vs. Negative" to narrow down on daily returns with correlation levels as similar as possible.

Thanks for replying

nickmar
Roundtable Knight
Roundtable Knight
Posts: 191
Joined: Tue Oct 26, 2004 12:38 pm
Contact:

Post by nickmar » Fri Mar 23, 2007 3:31 pm

Lessons learned over the years:

1 - For Trend-followers, high negative correlation is as dangerous as high positive correlation

2 - Correlations during stress events are much more important than long-run correlations

BARLI
Roundtable Knight
Roundtable Knight
Posts: 650
Joined: Sat Jan 17, 2004 6:01 pm
Location: USA

Post by BARLI » Fri Mar 23, 2007 3:43 pm

I've actually been playing around the Turtle System and found it is good at cutting losses short using the time stop

RedRock
Roundtable Knight
Roundtable Knight
Posts: 939
Joined: Fri Jan 30, 2004 3:54 pm
Location: Chicago

Post by RedRock » Fri Mar 23, 2007 3:50 pm

nickmar wrote:Lessons learned over the years:

2 - Correlations during stress events are much more important than long-run correlations
Hmm. I wonder if there is any order or predictability as to which markets will gang up when the 'stress' time comes. Or if its pretty much just disaster du jour ...

nickmar
Roundtable Knight
Roundtable Knight
Posts: 191
Joined: Tue Oct 26, 2004 12:38 pm
Contact:

Post by nickmar » Fri Mar 23, 2007 5:20 pm

The folks over at the RiskMetrics Group have thought long and hard about this subject: http://www.riskmetrics.com/stressovv.html

The full article can be found starting on page 61 of the following document: http://www.gloriamundi.org/picsresources/jkcf.pdf

Interesting approach but not trivial to implement. Even if the approach is implemented, how do we incorporate the results into our everyday trading? Perhaps one way is to modestly limit exposure to groups of markets that tend to have high absolute correlations during "hectic" periods. Ah, the trade off between risk and return...

Chelonia
Roundtable Knight
Roundtable Knight
Posts: 497
Joined: Mon Apr 30, 2007 3:37 pm

Post by Chelonia » Sun Sep 30, 2007 9:12 am

Ever tryed to exit 2000 Orange Juice futures when your stop gets triggered, or initiating a new position with 1500 Coffee futures? If you do so you probably appear on National Television later that day!!!
Did you take liquidity into account. I see many markets in your test that won't allow for relatively large positions without being killed.

Rg,
C

Post Reply