> These were called "Achievement Tests" in the early 90s, and man, they were much harder than the SAT. If the current "Subject Tests" approximate their level of difficulty, then it's a clear mistake for MIT to disregard these datapoints
Difficulty as perceived by test takers is irrelevant. What MIT cares about is how well test performance predicts undergrad performance. As TFA notes, MIT is retaining the standard SAT because their data indicates that performance on that test is predictive.