Great discussion, guys, and as David mentioned I fully agree with him and with Adam about the idea that GMAC will tend to hold back its hardest questions. It's probably not all that relevant to the way that you study, but since the discussion is out there I figure I'll chime in with a few reasons for that.
1) Like David says, these questions cost $2500 on average to create, and the really-tough ones have to cost more; having created some questions myself, I know that the harder you make a question the more eyes you need to get on it to make sure that the trap answer is "devilishly clever but 100% incorrect" and not merely "kind of defensible and probably a second correct answer". The hardest questions are the toughest to write, so GMAC really needs to make sure it gets a high ROI on those.
2) There are exponentially more "average" questions out there than there are incredibly hard questions. The GMAT scores fall along a normal-distribution bell curve, and accordingly more people score in the middle bands (400-600) than score on the extremes. Add in the adaptive scoring factor and GMAC has to create a lot more 600-level questions than 700-level questions, since most examinees will require scores in that 400-600 band. So there just aren't as many hard questions to retire.
3) The difficulty on the GMAT has to "move" and evolve - particularly with the advent of sites like Beat The GMAT (tip of the hat to Eric, David, and Bea), what is "hard" becomes easier fairly quickly nowadays...people get a sense of what they need to study for. 6 years ago, Permutations/Combinations was the elite level of GMAT quant...people couldn't figure it out when they saw it for the first time. But within a few months, internet forums started discussing it, companies built lessons around it, and the difficulty was eroded a pretty good amount. Once a hard question type gets out in the public domain it loses a percentage of its difficulty, so GMAC really needs to protect its hard questions and also to stay ahead of the curve.
4) GMAC itself will admit that the Official Guide series is much more a "collection of retired GMAT problems from which to study" than a "blueprint for what the test looks like today". Its stated mission is not to be proportionally representative of the concepts you'll see or of difficulty levels overall, so GMAC isn't putting much effort into doing so. If anything, the official mba.com practice tests, because they have to include difficulty-level based questions for their scoring algorithm, are the study tools to follow to find out where the difficulty may come from.
So, like David says, the OG is a great study tool at an attractive price point, so I'd certainly endorse it as a powerful study tool, but I'm also quite leery of the arguments that "I can do all the questions in the OG so I will score 700+" or "The OG doesn't have a question like that so you won't see it on the test." Combine the OG with some imagination - train yourself to Think Like the Testmaker and find other ways they can test those concepts or add twists to those questions along the same lines - and then you ought to be in great shape.
Brian Galvin
GMAT Instructor
Chief Academic Officer
Veritas Prep
Looking for GMAT practice questions? Try out the Veritas Prep Question Bank.
Learn More.