Re: GmatClub Mocks
[#permalink]
20 Jun 2021, 15:58
I can't comment about any particular company test, though I would only trust scores from official practice tests. Still, a V45 is so strong that you can take it as a very positive sign, even if it's from an unofficial test (and company tests more often produce scores that are too low than too high, so maybe you're doing even better than that!). On the real GMAT though, it's definitely normal to see questions at various difficulty levels, even when you're doing very well. That's true in either section - in Quant, you might be getting most things right, and then see an easy question in the middle of the test. In Verbal, I've recently seen a few ESRs from very high-scoring test takers that illustrate that a high-level Verbal test taker can sometimes get a test that contains a lot of medium-level (not high-level) questions throughout, even near the end. I can't say how common that is -- the ESRs were brought to my attention precisely because they illustrated that phenomenon, so that might be rare, or might be somewhat common -- but it definitely happens.
Most prep company materials claim the algorithm adapts very predictably. A lot of prep books use a 'tree diagram' ('you get your first question right, the next question is harder'). Those explanations are definitely misleading (and people draw a lot of incorrect strategy conclusions from them). The test doesn't need to adapt all that well to produce a reliable score. It generally wants to give you questions around your level, but the test's estimate of your level is not the only criterion it uses to pick your questions. There are also content balance restrictions (e.g. you're supposed to get a fixed number of DS problems, and a fixed number of problems on each topic), and the question bank isn't huge, so often the test doesn't have a question precisely at your level. And then there is an exposure restriction -- the test doesn't want to use any one question too often, in case people dishonestly learn about the questions a friend saw on their test. If certain questions were very likely to show up on anyone's test, that dishonesty would be more likely to be rewarded. So if the test wants to give you some 700-level Geometry DS question, but that question has been used too many times already, it might need to substitute in a 450-level question instead. That question will still count, and as long as you answer it normally, your score will end up the same. Verbal adapts even less predictably than Quant, at least in a certain sense -- RC doesn't adapt question-by-question. The test will choose your next RC passage and all of the associated questions all at once. So early in the test, it might want to give you a 500-level passage, but that passage might come with two 600-level questions and then two 400-level questions, and those questions are fixed as soon as you get the passage; even if you get the first three right, your fourth question is still that 400-level question that comes with the RC passage.
One important consequence of all of this: there's no good reason, in the middle of a test, to try to guess how well you're doing by the difficulty level of your questions. It's almost impossible to do. Another consequence: if you see an easy question in the middle of your test, don't be alarmed, and be sure to get it right. Getting easy questions wrong is the worst thing you can do on an adaptive test, and if you see an easy question, it very well might be a question that counts.
Of course you'll also see a small number of diagnostic (sometimes called experimental) questions scattered throughout your test, and if you're a high-level test taker, those will mostly be easy for you, but what I describe above would be true on a GMAT test that didn't even include diagnostic questions.
Good luck!