Re: No. of easy vs. no. of tough questions
[#permalink]
16 Jun 2021, 06:37
It's different from test to test. An adaptive algorithm can adapt quickly or slowly, and the integrity of the scoring won't be affected. An adaptive test can also start with an easy, medium, or hard question, and the integrity of the scoring won't be affected. It's impossible to be completely certain about this type of thing since there's no way to get reliable evidence, but I have the strong impression from test taker reports and other data that the GMAT sometimes does start with an easy or hard question, so it's not even certain to be true that your first question is "500-level".
The test does adapt immediately; after each question, the test estimates your level based on all of your answers so far, and uses that estimate as one factor to select your next question. It's not the only factor, so you can get a question right and then get an easier question next, but most of the time, if you get something right, your next question will be harder. I have the impression from various test taker data, that Quant often adapts quickly, so if you get two or three early questions right, you'll often be seeing some very difficult questions. That doesn't seem to be true quite as often on Verbal, though, and it won't always be true in Quant either.
I'll add, in case this is an assumption behind your question: the GMAT does not classify questions into discrete categories like "600-level" and "700-level". I've recently seen some dangerously misleading prep company "strategy advice", from more than one source, that claims the algorithm divides questions into five or so difficulty levels, and assigns points to each level. These sources claim that hard questions are "worth more points". Literally none of this is true. GMAT question difficulty is measured on a continuous scale (if you could see inside the algorithm, you'd see difficulty levels between roughly -2.5 and +2.5, which correspond to how many standard deviations above or below average a question's difficulty is, so a hard question might have a difficulty level of +1.87, and a fairly easy one a difficulty level of -1.07). The questions are not divided into discrete difficulty "bins". Nor is it true that the algorithm assigns "points" for right answers. There are no "points" anywhere in the scoring algorithm; the algorithm instead produces scores using probability theory. And anyone who understands the math behind the algorithm will tell you that "harder questions are worth more points" is pure nonsense, and is dangerous nonsense to believe if you want to do well on the test. Getting easier questions wrong on an adaptive test is very damaging to your score, because high-level test takers almost never get easy questions wrong, so as soon as you do, the probability you're a high-level test taker becomes low, in the eyes of the algorithm. You have a lot of latitude to get hard questions wrong, however, especially those at the very top of the scale, because even top-level test takers sometimes get the very hard questions wrong.
But in general, and this is the most important point (and I mention this in part because I've recently encountered a handful of videos and similar material about "strategy to get a Q51") -- the "strategy" to use if you want a Q50 or Q51 is to understand math well. There are no secrets about the scoring algorithm that a Q47-level test taker could learn that will raise her score to Q50 (or even to Q48, unless the test taker believes some persistent myths, e.g. about the importance of early questions, and is using a bad timing strategy because of that). Getting a Q50 or Q51 is not about "strategy". I've said this before, but a test taker who knows math perfectly, and avoids careless errors, will get a Q51 even without knowing the first thing about "strategy", or about the algorithm. A test taker who knows every "strategy" in every prep book, and who knows exactly how the algorithm works, but who knows no math, will get a Q6 unless they have some good luck.
Posted from my mobile device