A possible way to search for constraints that improve optimization.
The usual way of thinking about portfolio optimization is to first consider the utility and then restrict to where the constraints are satisfied. A perfectly reasonable view.
We use random portfolios to get a different point of view: first ensure that the constraints are satisfied and then look at utility.
Going down this less travelled road suggests that we might be able to modify the constraints in order to achieve better optimization results.
Annotated slides for “Portfolio Optimisation Inside Out” (from my presentation at the Computational and Financial Econometrics conference on Monday) show why the idea might work, and also outline the technique that Portfolio Probe uses to do trade optimization and to generate random portfolios.