"To refute this, we either have to come up with a way of measuring whether streaming students based on their scores is a benefit, or we have to come clean and s

"To refute this, we either have to come up with a way of measuring whether streaming students based on their scores is a benefit, or we have to come clean and say it's arbitrary." This is a false dichotomy. You could also come up with a good reason/argument/explanation of why the test would be effective, which no sees a refuting criticism of. It's not just evidential measurements or arbitrariness. Ideas matter too! (And in fact you need ideas to interpret measurements or evidence. Is your attempt to measure itself correct? You could try to measure that, too, but you'll immediately face the same problem again. At some point you'll have to do something other than measure.)