The 50-100 (really 140-450 points) is from Princeton study on Asian SAT stats for admission in private colleges vs other races. Apologies for wall of text.

Alternate explanation is that particularly sneaky elite admissions boards have learned to coordinate to devalue SAT scores for Asians, they'd rather engineer their mean Asian SAT score around something not embarrassingly high as cover for other cohorts, especially legacy which AFAIK doesn't need AA skew, since they're... legacy. AA primarily brings down average scores of non-legacy, and in this case make legacy look better, like average at Harvard for non asian & non white is 1450 according to your link, which is low considering 1500+ scores is 9th percentile for 223k Asian takers [0&1]. This is slightly conspiratorial, but also essentially what started the anti-AA case was when an Asian applicant with 1590 got rejected from multiple elite colleges, and the case essentially revealed high scores make Asians seem like bland test takers with bad personality even though personality assessment showed otherwise. Asian scores are high enough to break Harvard's argument not just for AA but IMO legacy as well.

Some other considerations, with respect to Harvard, according to latest admission data, @28% asian enrollment out of 2000 for class of 26, that's about 550 asians (previously this was ~20%). The real question the Harvard admissions stats don't show is how many Asians with 1550+ scores are being rejected due to "personality". If you look SAT score distribution [0], Asians are the only group where perfect math score is NOT 99+ percentile rank but "only" 95, and on ERW Asians 750+ score is 95 percentile vs other groups where it's 99+ except whites / multi racial at 98. Meaning Asians have disproportionate candidates with most perfect math and close to perfect ERW scores of all groups. Asians also total score outlier in that 1500+ is "only'' 91 percentile vs 98 to 99+ for everyone else. Collegeboard also has info that about 223k Asians (out of 2.2M takers) took SATs (2020 data [1]), so there's ~20k Asians with 1500+ scores (9th percentile). If extrapolated from 98th percentile of over 1500+ scores from 2.2M takers, that's ~44k total pool over 1500+ candidates of which 45% are Asians. 45% Asian is also in line with UCs where prop 209 banned AA and led to 40+% asian campuses.

Question is how many of 1550+ ones are Asians, keeping in mind ~7000 Asians have perfect math scores. My gut feeling is there are many 1550+ Asians being rejected by elites for skewing scores too high to keep legacy admissions charade going, especially at Harvard, because any Asian who scores that high, which rough napkin math is multiple size of their annual enrollment, would have an application to Harvard and eager to accept. Hence the narrative harvard stats paints via omission even by their student journalists, is that legacy does better non-legacy whose down by affirmative action, and legacy slightly worse than Asian Americans, when reality is Asian Americans have more than enough candidates to saturate scores above legacy and affirmative actions so their aggregate score has to be kept down through admission shenanigans.

[0] https://satsuite.collegeboard.org/media/pdf/sat-percentile-r... [1] https://i.redd.it/l6hb0eq4k5271.jpg

I am not sure why that argument would not also apply to white people then where there are ~909,000 who took the test. That leaves a large number of white students who got perfect scores or were 1550+. For example 1% of the total is about ~9000 people. So a strong case could be made that Harvard actually discriminates even more against white students who are not on parity and actually the only group underrepresented by total population size at Ivy League and other elite schools. The Asians who made this argument never mentioned this but actually made the opposite claim that whites were favored over Asians.

Of course, I don't think SAT score alone should determine entrance and no school has said that was all that mattered for admissions and I think that could be valid because the test score is really just one metric and doesn't give a complete picture. For example is the person a good test taker but a lazy student? That would reflect poorly on the student.

Also for the record and to be clear the legacy score of 1523 is still a high score and less than 1 std dev away from a perfect 1600 so it's not like we're talking about legacies who are totally out of range of admissions.

It does also apply to whites, hence anti AA actions was coordinated between white conservatives and some Asian Americans. One common talking point was AA + legacy was a double whammy on poor whites. Also it's not 1% it's 99+ percentile, i.e. any value less than 1%, which can be substantially less than 9000. States that have ban AA have demonstrated increased white and asian enrollemnt, so AA is indeed discriminatory towards whites. But if we use stats for institutions that banned AA in California, Asian enrollment uplift was highest, suggesting Asians are more represented on that end of SAT bell curve.

I agree at some point a 1500 and 1600 score are probably high enough even for elite universities. My broader point is that in composite admissions system like harvard, the score it self can be gamed to generate outcomes like racial composition. And institutes will figure out other ways to circumvent. But if they ever get slapped to move onto a score only admissions, that 1STD will make all the difference.