Taken from http://processdriventrading.wordpress.com/page/2/
Sometimes, it’s so easy. I was one of only two risk management majors in my graduating class, compared to about 500 Finance majors on the other side of the Field. I still don’t know if it was wise dropping Finance for Risk Management; I only know that it was honest. Risk Management folks are the wet blankets who take the punchbowl away just when the party gets going. Risk Management people avoid the loss of money, instead of looking for ways to gain money. Above all, Risk Management people can’t really be assessed, not on a consistent basis; you cannot objectively manage someone whose job it is to navigate something inherently random, if not unpredictable; risky, if not uncertain. (Aaron Brown has a lot to say about this distinction) Worse still, really trully honest Risk Management folks can’t even be sure of their own value. So when an opportunity comes along to make money purely from Risk Management alone, we jump at it.
Here we take a leaf from CSS-Analytics’ playbook, DVaR position sizing. I’ll make that required reading before you proceed on with this post.
Now see how that looks in SG & HK:
and the “conservative” version:
You will note that although the strategies do better on the two measures of risk, Stdev and Worst Day, they still lose out in CAGR terms. It seems the Risk Management guys save the money but don’t bring home the bacon – but that simply isn’t smart use of Risk Management guys. Since the VaR method frequently makes the portfolio less than fully invested at all times, here are the same charts goalseeked to have 100% weights on average:
It’s a resounding WIN for Risk Management! However note that Sharpe falls whenever the Risk Level parameter is adjusted upwards in order to go for more CAGR. This is an example of the temptation to take on more risk in order to get more return, even when the increase in return doesn’t really justify it.
Notes on this strategy:
In case you were wondering like I was, this strategy has not lost effectiveness yet:
Having played out the trading simulations, I also wanted to think about the source of these profits. Since VaR was the only decision rule, and weights varied according to Risk Level/VaR, then 1/VaR must be predictive of next day returns. Turns out, it isnt:
But that’s only if you think the market has the same reaction to risk downside as well as upside. Lots of academic literature beg to differ. Here is the same thing filtered for only negative days:
q.e.d.
Sometimes, it’s so easy. I was one of only two risk management majors in my graduating class, compared to about 500 Finance majors on the other side of the Field. I still don’t know if it was wise dropping Finance for Risk Management; I only know that it was honest. Risk Management folks are the wet blankets who take the punchbowl away just when the party gets going. Risk Management people avoid the loss of money, instead of looking for ways to gain money. Above all, Risk Management people can’t really be assessed, not on a consistent basis; you cannot objectively manage someone whose job it is to navigate something inherently random, if not unpredictable; risky, if not uncertain. (Aaron Brown has a lot to say about this distinction) Worse still, really trully honest Risk Management folks can’t even be sure of their own value. So when an opportunity comes along to make money purely from Risk Management alone, we jump at it.
Here we take a leaf from CSS-Analytics’ playbook, DVaR position sizing. I’ll make that required reading before you proceed on with this post.
Now see how that looks in SG & HK:
and the “conservative” version:
You will note that although the strategies do better on the two measures of risk, Stdev and Worst Day, they still lose out in CAGR terms. It seems the Risk Management guys save the money but don’t bring home the bacon – but that simply isn’t smart use of Risk Management guys. Since the VaR method frequently makes the portfolio less than fully invested at all times, here are the same charts goalseeked to have 100% weights on average:
It’s a resounding WIN for Risk Management! However note that Sharpe falls whenever the Risk Level parameter is adjusted upwards in order to go for more CAGR. This is an example of the temptation to take on more risk in order to get more return, even when the increase in return doesn’t really justify it.
Notes on this strategy:
- I actually didn’t need to make so many charts; just the simple fact that Sharpe improves in the strats is sufficient to make these conclusions. It’s just fun to do.
- Note that the “Actual VaR” measures out of sample 5th percentile returns, and is in all cases in excess of but very close to the “preset” VaR or “Risk Level” parameter at the top of the chart. This means it actually is not a fool’s errand to talk about “risk budgeting”.
- An excess CAGR of 1-2% a year over more than 30 years can be hard to visualize. This might help:
In case you were wondering like I was, this strategy has not lost effectiveness yet:
Having played out the trading simulations, I also wanted to think about the source of these profits. Since VaR was the only decision rule, and weights varied according to Risk Level/VaR, then 1/VaR must be predictive of next day returns. Turns out, it isnt:
But that’s only if you think the market has the same reaction to risk downside as well as upside. Lots of academic literature beg to differ. Here is the same thing filtered for only negative days:
q.e.d.
No comments:
Post a Comment