How can the scale be made in advance? Surely the way it works it is figured out based on how difficult, those taking the test, found it to be? If the absolute best score on the test is 95/100, then surely a 180 should be 95 non? Or will that sort of issue merely be reflected in percentile? I didn't think it was arbitrarily made based on how difficult they thought the test would be, but rather, how the difficult the test actually was. I re-read the little thing in the intro that talks about this but all it talks about is score bands and SEM and accuracy coefficients or something like that. If it's pre-determined, that may kind of suck.
Well the LSAC has a reputation to uphold of consistency. That means a 170 in June 2008 means the same level of competence as a 170 in February 1995. So if they based it on how people did, that means that if everyone was dumb one year (hypothetically) and the curve was ridiculous, then a 170 that year would not equal a 170 in a prior year.
That's why I'm sure they determine the curve beforehand, based on experimental sections and how they feel the level of difficulty is. They spend millions doing research into this kind of stuff to develop a bell-curve for each test that very accurately correlates the difficulty to the curve.
So I'm hoping that the trend is a 89-91 for 170. I think last June's 92 for a 170 curve was ridiculous because they were experimenting with the comparative passages and therefore made the LG and LR easier to compensate. I've got a good feeling that this curve will be -10 or even -11 for a 170 because the RC was tougher than usual and the LGs were not as straightforward (vertical linear games haven't been thrown into a test for a long time) Not to mention that the LR was not as easy (there were at least 4 tricky ones where the credited answer is the least wrong choice)