> We have to this point been unable to produce such a measure
We have not been able to produce a precise measure. But we’re able to e.g. identify quartiles. Compressing the measure to “likely to graduate” tends to take care of a lot of the lifting.
That’s why a random selection component makes sense. It is honest about our limitations. And it removes the stigma from those who met the fuzzy threshold but didn’t get admitted.
We have to this point been unable to produce such a measure, despite more than a century of trying.