We can't really resolve this with the current data, because the current system encourages people to get as high a score as possible at the cost of maximum stress. How do we know whether changing the system would lead to a less stressed system where people just do enough to leap the bar, perhaps like the diving test?

The selection process has a cost. If it doesn't produce value it should be replaced with a lottery.

It's pretty clear however that the selection process at elite universities, however imperfect and to whatever degree it factors in things (like legacy admissions) that you may think shouldn't be factored in, still produces better results than just randomly admitting some percentage of whoever applies. (Especially if applicants knew it would be a completely random process.)

It's not clear that the existing selection processes at most universities actually adds value in proportion to its cost. What value, and to who? What are the shortcomings you think they're trying to fix?

Even if everyone agreed on a standard and its impact on admissions (+3 for participation in the Model UN, for example) it's not clear how we'd judge "better results" as we evaluate our admissions criteria. Are we optimizing for an even racial/religious mix of students? Higher output grades? Higher wages? More children raised successfully by the age of 40?

Then even with an agreed upon goal and some objective measures to used to approach it - is it worth doing? If we could get 1% better results by doubling the difficulty of the application process for the students then it probably wouldn't be worth doing it when the plan was considered holistically. What's an acceptable trade-off? What's the cost on transparency and perception of bias when we start considering subjective criteria?

What university has done this experiment?

For the SAT/ACT specifically, it's optional at a fair number of schools--elimination mostly started during COVID has been extended. (And not used at all at a few.) Of course, they still look at high school grades and class rank. I guess they'll have a better understanding in a few years how things turn out.

Of course, the process is still a far way from random.