Interestingly, there is a growing body of research that suggests not only that how brokers pick stocks is actually no better than picking at random but that the perceived differences between good and bad brokers are in fact simply a result of clustering.
A couple of years ago, a magazine in Belgium did an experiment. They let some stock brokers, random people and... a couple of bonobos pick stocks. They followed things up, and after a couple of months the conclusion was that the bonobos had performed best! That was pretty funny I have to admit, lol.