If the test is based on the standard middle/high school curriculum (if there is one), then it can be a fairly good predictor.
Otherwise, even if it is an arbitrary test, it takes study skills and discipline to be able to master it. I imagine that both of those traits correlate with performance at school.
I doubt that, humans are capable of overfitting standardized tests extremely easily. Just doing a single practice run will usually give you significant advantage. You would have to "regularize" by changing the test structure/types of questions/whatever every year, which is not really feasible.
> Otherwise, even if it is an arbitrary test, it takes study skills and discipline to be able to master it. I imagine that both of those traits correlate with performance at school.
Sure they do - except now you have got students training for a significant part of their middle school life for an arbitrary test instead of some "real" education.
Second, that it may be contrary to their social engineering goals to place high schoolers into schools based on test taking ability in the first place. They don’t want popular schools to be loaded with the students who achieved the highest test scores. Probably this is exactly not what they want.