Official practice test scores | GMAT Prep Mock 6

Question

Hi,

I recently took GMAT Prep mock and scored a composite of 720 (Q50, V38). This result, however, is a bit surprising because I got 28/31 correct on the Quant section and 31/36 correct on the Verbal section.

I have, in the past, scored as much as a 760 with 5 mistakes on Verbal and 5 mistakes on the Quant section. I understand that there's another layer - what was the difficulty level of the questions.

In this particular mock, I was trying to experiment a strategy - if short on time, I will guess on the last RC, but be extremely diligent on other SC and CR questions that follow. As per the plan, I guessed on the fourth RC, which had three questions, and these questions weren't easy. Two of them were hard, and one was easy. Needless to say, I got all three of them wrong (Q 31, 32, and 33). However, I got Q 34, 35, and 36 correct.

Apart from these three, I got Q3 and Q19 incorrect. I was expecting a composite of 750/760. I will always have time crunch on Verbal, as I am a slow reader. I have been training for the GMAT for almost 1.5 years, and I know what my weaknesses are. Considering the dynamics, I have mentally prepared myself to guess on about 2-3 questions.

What I want to understand is that getting three questions wrong in a row can really hurt the score. Would it be better if I space out my guesses? Or I could sail smoothly and only guess towards the end (probably the last 2 questions)?

I come to discuss this issue here because I have seen a few posts detailing experiences such as mine, a few times on GMAT Prep mock 6 only.

PS: There are no conceptual gaps as such, I am just a slow reader.

Please share your experiences and help me achieve a V42-44.

Thanks!

AliTariqPrepVista · Answer

More comprehensive post by Ian. Please refer. Some good points there.

IanStewart · Answer

It's a bit of a myth that wrong answers in succession are especially detrimental to your score. Or maybe I shouldn't say that, because in practice it's true more often than not. But many people discuss the issue as if it was some feature of the algorithm, that the algorithm specially punishes consecutive wrong answers, and that is not true at all.

The important principle is this: getting an easy question wrong hurts more than getting a hard question wrong. When you get one question wrong, more often than not (but not always) your next question is easier. So if you get that second question wrong also, you've just gotten an easier question wrong, and that will hurt you more than the first incorrect answer did.

But that's only true when the test genuinely adapts downwards after the first wrong answer. That doesn't always happen, because the test doesn't adapt in a perfectly predictable way. Sometimes you get a question wrong and the next question is harder. Then your first wrong answer hurts more than the second, if you get both wrong.

And none of this applies at all to RC, because RC  does not adapt . If you see a passage, and answer the first question attached to the passage incorrectly, that has no effect on the question you see next. All of the questions attached to the passage are chosen by the test the instant the test chooses the passage. If you're doing well, the test might pick a 'hard' passage with a hard set of questions, but RC does not adapt from one question to the next. The post currently above mine is mistaken about this point; a sequence of wrong answers in RC is not dragging question difficulty downwards, so this particular effect, that a "sequence of wrong answers is bad", doesn't happen in RC.

The problem though with guessing at RC is that you get a preset batch of RC questions, some of which will usually be below your level. If you're say at a 650 level, you might get a 650 level passage, but that passage might come with a 500-level question and three 700-level questions. If you guess at all four, you're guessing at a 500-level question, and that hurts a lot. So guessing at RC can be a bad idea (if you don't get lucky) not because you get a sequence of questions wrong, but because you're almost always guessing at questions you're supposed to get right.

Because of that, I wouldn't advise guessing at an RC passage if you're a high-level test taker (and you are). Instead, in your case, assuming you must guess at two questions, I'd either guess at two SC questions or at two CR questions, depending on which type you're better at (guess at what you're worst at, because you're less likely to get those questions right if you attempt them). If there's no obvious difference, probably better to guess at CR, because that will save you more time. Definitely try to make those guesses near to the end of the test. And you're never making a big mistake on an adaptive test if you save your guesses for the end. That's when the test is most confident about its ability assessment, so that's when the test wants to adapt least (though as I said, adaptation isn't perfectly predictable, so if you'll be guessing, no matter how you do it, you'll be running the risk of guessing at something easy). Do be sure to finish no matter what, and if you've been scoring 750-760 on several diagnostics, you should be absolutely fine, as long as you don't adopt a counterproductive strategy. Good luck!

ReedArnoldMPREP · Answer

I don't have much to add--Ian's got most of it thoroughly covered. I'd just take a second to fully emphasize question difficulty is not 'another' layer on which the score is given, the first layer being # correct. #correct is the layer, the first layer is question difficulty. It's true at higher scores (where you are, especially at a 760) that the test starts to 'shift' to counting how many questions were missed (since they can't give harder questions), but even then--missing easier questions is always more punitive.

The test doesn't really adapt in a 'right answer, harder question, wrong answer, easier question' way. That's the tendency, but that trend can be bucked. Each question has what's called an 'S curve' which is basically a graph that shows what % of scorers at certain scores get that question right. Low scorers usually have a low chance (20%, blind guessing), high scorers should have a high chance. The 'perfect' question to 'discriminate' at a score of, say, 39: 20% of people scoring below get that question wrong, 100% scoring at or above get that question right.

Of course no such question exists. The mind is not so discrete, and people's strengths vary. But generally, they know, for every question, the chance a taker at a certain score gets that Q right. Every time a question pops up, based on what the test knows about you, the algorithm predict your score, (and thus your chance of getting a question right). They have all sorts of analytics and correlations between questions to do this.

So if the test thinks you're a 38 scorer, and they give you a '33' question (a question for which people who score a 33 have a ~50% chance of getting the question right), the test might say (and this depends on that question's s-curve), "They have a 76% chance of getting this question right." If you get it right, the test takes that data and moves on, with another data point to make accurate predictions. If you get it wrong, the test thinks, "I might've overestimated," and adjusts accordingly. If you're scoring a 43 and the GMAT gives you a 50 question, it might say, "Okay, you only have a 33% chance of getting this one." And if you don't, the test isn't that 'surprised,' and your score won't drop much.

TargetMBA007 · Answer

IanStewart

Hi Ian, RC not adapting is news to me. I wonder if this is something officially confirmed by GMAC, or if its based on empirical evidence.

IanStewart · Answer

I'm not posting here any more (because of GMAT Club's new policies towards independent tutors), except to clarify earlier posts I've written. On the old GMAT, RC did not adapt question-by-question. If you got an RC passage in the middle of a test, and answered the first question on that passage correctly or incorrectly, that did not mean the second question on that passage would be harder or easier, respectively. Instead, RC adapted by passage. So if you were doing well halfway through a test, you'd normally get a hard RC passage (i.e. with a set of harder questions), and if you were doing badly, you'd normally get an easy passage (with a set of easier questions). But the set of questions you'd see was fixed as soon as the test chose the passage it wanted to give you. GMAC officially confirmed this a few years ago.

Anything I've ever posted about the algorithm is based either on official information from GMAC, or on academic journal articles about the mathematical basis of the algorithm the GMAT uses, never on empirical data (that's where most of the misinformation about GMAT scoring comes from -- people misinterpreting ESRs and experiments with diagnostic tests and other empirical data). And I'd add that when I wrote my earlier post, I was describing the classic test. The Focus test might work differently; I have no information yet either way about how RC adapts on the Focus edition.

TargetMBA007 · Answer

Thanks for your inputs Ian.

Official practice test scores | GMAT Prep Mock 6

Prep Toolkit

Top 5 GMAT Debrief Videos

615 to 715 in 15 Days (Ria's Strategies)

More than Quant/Verbal Abilities: Speed vs. Accuracy

745 with Only Self-Prep and GMAT Club

From 555 to 765: 210-point Improvement

Perfect 805 Score Debrief by Julia

My Rewards