Updated the formula (see original post), got rid of some of the rounding inaccuracy. It should add about .7 to predicted LSAT scores. Interestingly -- and this is probably just coincidence -- a 1600 corresponds almost exactly with a 178. Also interesting to note is how the median SAT of 1020 equates to a 150, only slightly below the LSAT median of 151. Apparently, people taking the LSAT really aren't any smarter as a group than those taking the SAT.

Someone above said that the SAT verbal should be weighted more heavily than the math. I'm no so sure about this. Clearly, RC should have an very high correlation with SAT verbal. But I'm willing to bet that games have a much higher correlation with math. As for the other half of the test, LR, I really have no idea whether this correlates more strongly with verbal or math. It doesn't make much sense just to assume it correlates better with verbal simply because it's in word form, not number form. Without anything more to go off of, we might as well give equal weight to both in this section, and as the correlations of RC and games cancel each other out, we're left with math and verbal receiving equal weighting in the final equation.