From the article:
> Winkler’s research told him that between 5 and 40 percent of psoriasis patients feel their symptoms are alleviated when taking placebo pills.
Between 5 and 40 is a lot of variation!
The placebo effect has many causes, you don't know witch percent of the general population will react to the no-treatment, you don't know how the subgroup you selected will react to the no-treatment, you don't know if just going to the doctor once a week as part of the study will change how the people fell, how they sleep, eat, whatever. There are too many variables, and in a serious study it's impossible to eliminate all of them. The best second option is a double blind control group, so you hope that whatever affect the subject affect in approximately the same weight to the control group.
[1] Make 100 double blind studies. In half of them use the screening method to avoid placebo susceptible people and in the other half use the usual mix of people. But make these selection double blind, so the doctors don't know if they are running the normal or the filtered study. Then analyze the 100 studies an see if the 50 filtered studies have more success than the 50 unfiltered.
P.S. What about if the drug is better to the people that has no the genes they are blaming for the placebo effect, but it will kill half of the people with these genes?