Optimizing the weights of futures commodities in a portfolio can
reduce risk and increase return. The biggest risk is from serially
correlation between commodities. For instance, we would expect there
to be some correlation between the Dollar Index and other
currencies. Systems are likely to be short the Dollar Index at times
the other currencies are long. This can cause the risk to increase
dramatically because they are both likely to be in drawdowns
simultaneously. By optimizing weights, we can detect these problems
and alter the portfolio accordingly.
You can choose to optimize in two different ways. The first is
optimizing Return/Risk using the Markowitz solution where risk is
the standard deviation of residuals of the equity curve around the
linear regression line. This attempts to modify the weights so that
the equity curve is as straight as possible, however drawdown is
likely to increase. The second method is preferred and was suggested
by Perry Kaufman, and defines risk as the standard deviation of
drawdowns. As you can see below, the portfolio has not been
optimized. Weights are all set to 1 except bonds which are set to 2.
