Page 1 of 1

What happens when your porfolio correlates to 1?

Posted: Thu Oct 06, 2011 1:17 pm
by LeviF
I have not found it beneficial to limit total risk in my systems. I think this is primarily due to having uncorrelated instruments in the portfolio. Adding more and more uncorrelated instruments while increasing total risk, should decrease "effective" risk.

However, what happens if the instruments become highly correlated? (nothing says they cant) We could have a big problem. One idea I have been tinkering with is reducing position size across the portfolio if volatility of the equity curve exceeds some predetermined figure.

This could be measured as % change over sometime frame, average daily change, standard deviation of returns, etc. I like to normalize this figure by adjusting it by the % or $ at risk over the same time period.

One concern is the robustness of this rule. If we want to reduce position size if vol > x% and this rule only is triggered 10 or so times over the life of a test, is it something we should hang our hats on? Although perhaps we are erring on the side of caution if we do...

Has anyone else pondered this issue?

Posted: Thu Oct 06, 2011 3:11 pm
by rgd
I have pondered this issue at length. I limit total risk. Yes it does put a cap on upside to a certain extent. But, backtesting will always lead you to take too much risk. If you do not employ a logical override, like limiting total risk or portfolio level volatility, then nothing will separate you from the guys who have 40-50% drawdowns. While the risk limiter I employ only triggers twice per year on average, in the absence of hundreds of observations, I have to answer the question: Does it make sense?

I will take the CTA who understands risk and uses common sense, over the one who employs only statistical optimization anyday!

Posted: Thu Oct 06, 2011 4:25 pm
by drm7
I recall that Tom Basso has a similar system for limiting risk (at least at the time of his interview in Market Wizards.) He limits "total open volatility" (i.e., the total $ value of ATR in his open positions). If volatility goes up, he trims back positions until his portfolio is back at the target. I don't know how he prioritizes the instruments, though.

Edit: I found a simplified version of this system here:

http://www.cmegroup.com/education/files ... 1-3-08.pdf

Page 63. It is written by Tom Basso.

Re: What happens when your porfolio correlates to 1?

Posted: Thu Oct 06, 2011 10:00 pm
by Aaron01
LeviF wrote: One concern is the robustness of this rule. If we want to reduce position size if vol > x% and this rule only is triggered 10 or so times over the life of a test, is it something we should hang our hats on? Although perhaps we are erring on the side of caution if we do...
I believe you've hit an important issue here. If a signal is generated with such infrequency then it would be hard to validate whether it is of any statistical merit or if it's simply a product of chance.

Posted: Fri Oct 07, 2011 2:17 am
by Chris67
RGD

RE:

"then nothing will separate you from the guys who have 40-50% drawdowns"

If you dont have at least one or two of these drawdowns in your time then I'm afraid there will probably be a lot that seperates you from teh guys who make 20% CAGR over 20 years

i.e you cannot have one without the other

Best
C