How To Creatively MVO Your Ranked Signals

Introduction

Getting ranked inputs as a result of your signal generation process is really common.

There’s just a tiny problem.

Your portfolio optimiser wants you to pass it expected returns but your signals outputs ranks, what do you do? MVO needs all your inputs to be on compatible scales. However, your rank signals are ordinal, they tell you that stock A is better than stock B, not by how much, and this does not play nicely with your covariance matrices.

The question is how to translate that ordinal information into something the optimizer can use. There are a few ways to bridge the gap. Today’s article talks about all the ways in which you can optimally apply MVO to ranked inputs.

We have a very special/novel technique discussed here as well that is in fact, NOT MVO.