Ferret1 wrote:
1. I have heard from a number of sources that the level of difficulty of the questions in the OG is not as high as those one would encounter in order to achieve 700+ level score in the actual test.
I'd agree that's true of OG11. The 12th edition of the Official Guide does seem much more representative of the current test to me, despite what you might read in some company reviews of the book. Just to take one example (from the middle of the book, no less), a question like no. 110 in the PS section of OG12 tests prime factorization in a much more advanced way than any question in OG11. And the DS section of the book in particular is much more up to date.
Still, the questions in the book span all difficulty levels, and to see a lot of realistic high level questions, you'd do well to exhaust the supply available in the GMATPrep software, which certainly contains many very challenging questions and which will give you a very good idea of what to expect on the real test. For additional math questions, you may want to consult GMATFocus, though many GMATFocus questions appear in OG12 as well, and I don't know yet if these questions have been removed from the GMATFocus database. If there's a lot of overlap, GMATFocus may be less valuable now than it used to be - something I'll look into soon.
Note also that the diagnostic test at the beginning of the OG (which is unchanged in the 12th edition) contains some of the hardest problems in the book, so is useful if you need some challenging material.
Ferret1 wrote:
2. To my mind, the logic of the GMAT scoring algorithm would mean that one should be getting a 50% hit rate on questions that are at the level of the score you "should" obtain. I know this is an oversimplification, but in general does this mean that if I have been scoring say 90% in the hardest questions of each section of the OG that I should score above a 650-670 in the actual test (assuming this difficulty level is true of the OG)?
That is an oversimplification, though it's repeated in a lot of test prep materials. The 50% threshold is only used in 'free response' tests - that is, tests which don't offer answer choices. On the GMAT, with five available answer choices, the algorithm takes into account the possibility of guessing correctly.
The technical details, which you don't really need to know: on many questions, where ruling out wrong answers is likely going to be difficult for the weakest test takers, the probability of guessing correctly is going to be 20% (the probability of correctly answering if guessing randomly). I'd think question 110 in the PS section of OG12 is one such question; a low level test taker won't have any criteria to use if guessing on that question. You need to get more than 60% of such questions right, not 50%, to convince the test you're above the level of such questions.
That said, some questions will have a 'guessing probability' which is higher than 20%. Looking, for example, at question 12 in the DS section of the new OG, the question asks about i+j, and Statement 1 doesn't give any information about j. I can't imagine anyone guessing A or D for such a question, and C also seems unlikely, so the 'guessing value' for this question is likely to be somewhere between 33% and 50%. On these kinds of questions, to persuade the test that you can handle their difficulty level, you'd need to answer with an accuracy above 67-75% (depending on the 'guessing value').
When the GMAT administers experimental questions, one of the statistics they compute for each question is this guessing value (called the 'pseudo-guessing parameter'), which represents how well the weakest test takers answer the question.
Long story short, you need a hit rate quite a bit higher than 50% on questions of a certain level to persuade the test that you're above that level, and the exact percentage depends on the question. If you do have a 90% hit rate on the questions at the end of the OG, you're still certainly well above the required threshold, however - that's an excellent performance. Still, to get a realistic score estimate, take a GMATPrep test; it's hard to judge from hit rates on the OG.
Ferret1 wrote:
3. Lastly, I have found that the question difficulty in the OG does not seem to vary greatly between the "easiest" questions and the "hardest" questions especially in the verbal sections. For example for both the SC and RC sections I averaged approximately 90% for the first half of the question bank and about 95% for the second. I realise part of the reason for this may be that I had picked up on a few extra tricks during the process but I can still say that I did not notice a substantial difference in the level of difficulty. Would anone know if this would also be true of the difference between a "650" level question and a "750" level question, or beyond 700 would qustions get substantially more difficult? Can any one who has achieved a 700+ and worked through the official guide indicate approximately what your "hit rate" was in each section of the OG and how many questions you would estimate you missed in the each section of the actual test?
I'd be very surprised if the difficulty level of real GMAT Verbal questions was as varied as of Quant questions, though of course the only people who would know that for certain are the test developers. This does make the Verbal scoring less forgiving of mistakes; you can make several mistakes on the Quant and still get a scaled score in the 47-50 range, but one mistake on the Verbal will lower your score to 49 or 50, and two mistakes will lower your score to 47 or 48, at least judging by GMATPrep results.
I do think GMATPrep gives a very good indication of the style of question you'll encounter on the real test, though on my most recent GMAT late last year, I did encounter a small number of questions which did seem substantially more difficult than any I had encountered in GMAT prep materials - one heavily academic RC passage full of definitions of new terms, which went on to draw conclusions about these newly defined phrases; a couple of SC questions with two answers which were grammatically correct, and where the correct answer would certainly have been that which changed the original meaning least; and one CR question I found ambiguous, since it wasn't clear from the stem whether certain information presented was an assumption of the author, or a fact. Still, if you do well on GMATPrep, you can be confident of a strong performance on the real thing.
Good luck!