I have been playing with several transformations of mean variance based portfolios to adjust for concentrated portfolios that go outside of margin requirements. This one does not solve the problem but does allow some control over it, and it's about as simple as they get.
The gist:
- solve for the efficient frontier weights
- Take the unit vector in the same direction
- Multiply by some scalar for the portfolio weights
The scalar could vary based on some other function but here its set constant at 1.25. It's the context.margin variable, not the correct name but I used an existing variable. Weights get multiplied by (1 + context.margin).
Any thoughts on the implications of this type of model.