Random selection is important to know if your system actually improves anything, or if the results are only good because the input was good to begin with.

For all we know, those universities that only pick the best and brightest may actually produce inferior results that are nevertheless still very good.

Thomas Sowell, when challenged on the success of charter schools in New York as being biased due to a "better selection" of students, was able to point out that this bias was limited by the fact that admission was based on a lottery.