Sure, people with (near) perfect scores on the SAT have aptitude for college work, but they aren't the only ones. Plus, given the way the test is scored, it isn't clear to me that anyone can even honestly discriminate between "aptitude" in students with scores between ~650 and 800.

Executive Summary
There is a common misperception that test scores do not predict above a minimum threshold
(Sackett, Borneman, & Connelly, 2008). That is, test scores may be useful for identifying
students with very low levels of ability; however, higher scores are considered unrelated to
higher performance for those above a certain threshold. This study aims to examine whether
this is true for the Preliminary SAT/National Merit Scholarship Qualifying Test (PSAT/NMSQT®),
which is used for that very purpose — to differentiate among very high performing students.
The linearity of the relationship between PSAT/NMSQT scores and first-year college GPA
(FYGPA) was explored in this paper, using a regression approach. This relationship was
explored over the entire range of the PSAT/NMSQT score scale, known as the Selection
Index, ranging from 60 to 240 as well as the upper end of the score scale (≥ 200), where
initial screening decisions are made for scholarship programs conducted by National Merit
Scholarship Corporation (NMSC). For the full PSAT/NMSQT scale, the addition of a quadratic
term improved model fit; however, the effect size was small as indexed by the change in the
squared multiple correlation coefficient (R2
) of 0.001. That is, including PSAT/NMSQT Selection
Index2
in the model accounted for an additional 0.1% of variance in FYGPA. As for the subset
of students who had a PSAT/NMSQT score of 200 or higher, the results indicated a strong
linear relationship, which suggests that even among very high-scoring students, the PSAT/
NMSQT score scale differentiates between students in terms of academic success measured
by grades earned in the first year of college. In sum, the results of this study support the use
of the PSAT/NMSQT as a screening tool for selecting Merit Scholarship winners.

A response.
http://www.newrepublic.com/article/11932...ndardized-tests
The Trouble With Harvard; The Ivy League is broken and only standardized tests can fix it. By Steven Pinker. Sept 4, 2014.

Stephen Hsu comments on Steven Pinker's article.
http://infoproc.blogspot.com/2014/09/what-is-best-for-harvard.html
I'd really like to know how many are admitted on academic merit alone, especially in math.
I assume MIT and CalTech are more meritocratic, though they are more expensive that Harvard for those with modest incomes.

This is something I thought we'd been pretty clear about. Mollie blogged about it back in 2006. Our institutional research website says, quite specifically, that "alumni relations" are "not considered." And I can tell you, from having sat on countless committees, that we simply don't care if your parents (or aunt, or grandfather, or third cousin) went to MIT. In fact, one of the things most likely to elicit a gigantic facepalm is when a student namedrops some incredibly attenuated connection because they think it is going to help them get into MIT.

...

I want to reiterate that I agree wholeheartedly with everything Mr. Kahlenberg said about why legacy admissions are bad. I personally would not work for a college which had legacy admission because I am not interested in simply reproducing a multigenerational lineage of educated elite. And if anyone in our office ever advocated for a mediocre applicant on the basis of their "excellent pedigree" they would be kicked out of the committee room.

I don't believe MIT and CalTech consider legacies.