I read on the CAT exam scoring page that you have modified your scoring system as of June 11. It seems to me that you have made the math score a lot more lenient than it was before as the number of questions I got wrong didn't change that much but my score went up. Can you please explain if this is closer to the real GRE scoring method than the previously used system?
Yes, the scores should be more accurate now. Score on the GRE is a function of both % right and the difficulty of the questions. We are now better able to gauge question difficulty--we have accumulated data on how people respond to each question, and are better able to correlate performance on individual questions and ultimate performance on the official GRE.
Also, on average, how many questions maximum can one get wrong in the quant section / verbal section to still maintain a 90% overall score in each? I understand that each question is of different weight and the algorithm takes this in account, but would getting say 10 questions wrong out of 40 still keep you in the 90% percentile?
My best estimate is that missing 10 of 40 in the math section would place you in the mid-80s percentile. This is a VERY rough guess, bit I think 90 percentile would require no more than 7 or 8 wrong answers. But as you mention, the difficulty mix is relevant.
It also matters when the errors occur. You get two math sections and two verbal sections. How you do on the first math section determines whether you get the easy, medium, or hard second math section (same for the two verbal sections). If you get the easy second section, your ultimate score will fall in a certain range depending on your performance on that second section, but is effectively limited by the fact that you got the easy section. An 800 (or 170 on the new scale) is probably impossible at that point--you would have needed to earn the hard second section to reach that maximum. So, even if you could miss 7 or 8 questions and hit 90th percentile, that may not be true if all of the errors to occur in the first section.
The main message: Your score depends on both how many right AND the difficulty mix AND on which second section you get. So comparing two different tests, you could miss the same number of questions and get slightly different scores.