Feb

28

Given the following parameters in macro (FX and rates primarily):
Portfolio return target 10%
Vol average 8%
Max drawdown 8% - portfolio cut and out
Drawdown 4% - capital cut in half

How best to consider various ways to trade within these parameters:
What do you do at -2% drawdown in terms of adjust position sizes, risk, etc.?
How adjust position sizes, risk etc. at +4%, etc.?
How consider position sizing at point 0 relative to percent of portfolio at risk, trade return goal, etc.?
How best overall utilize a strategy to reach those objectives?

Generic questions and these and others can be expanded of course. Any thoughts or reference materials I might read?

Zubin Al Genubi comments:

Parameters need to be dynamic and adjust with volatility.

John Floyd responds:

Yes, agreed. I am thinking more along the lines of specific quantitative measures and rules such as fractional Kelly criterion and assessing probabilities outside a fixed system and how to size in drawdowns and upturns both when trading tactically and on longer-term basis.

Kim Zussman offers:

The authority on this is Ralph Vince, who is here or elsewhere. Search optimal F.

Zubin Al Genubi agrees:

Ralph Vince's books are quite good and improve on Kelly's risk measures.

Theodosis Athanasiadis writes:

it depends on the type of approach and systems. those parameters are for a high sharpe approach (>=2) which is hard in rates and fx. they also make you become ostensibly a trend follower and short-term trader in your approach and pnl.

there are many methods for sizing that at the end of the day are similar; expected return divided by a measure of risk. Kelly is just a subset of the whole spectrum. Vince's work is a must read. The question is the inputs and as you know better they are very hard to estimate. fractional sizing, sizing inverse to vol etc will help but it is not going to change your return profile immensely. Most likely you want an approach similar to constant portfolio insurance. the best approach would be to use simulations along with extremes/penalties (eg haircut your returns by half etc) to create a distribution of pnl and then decide the optimal way. it is tedious but you avoid making more assumptions and you incorporate worse than historical scenarios.

from my experience, those don't work well in practice and you end up reducing allocation when expected returns are higher but if you have those restrictions you need to have something in place.


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search