This is a very good question for which I have a poor answer.
My selection criteria are based around Graham and my own personal observations without real quantitative back testing.
I really should set some time aside for this.
For what it's worth, much of what drove me to more direct...