Events & Promotions
|
|
GMAT Club Daily Prep
Thank you for using the timer - this advanced tool can estimate your performance and suggest more practice questions. We have subscribed you to Daily Prep Questions via email.
Customized
for You
Track
Your Progress
Practice
Pays
Not interested in getting valuable practice questions and articles delivered to your email? No problem, unsubscribe here.
May 1408:30 AM PDT
-09:30 AM PDT
Most GMAT test-takers are intimidated by the hardest GMAT Verbal questions. In this session, Target Test Prep GMAT instructor Erika Tyler-John, a 100th percentile GMAT scorer, will show you how top scorers break down challenging Verbal questions..
May 1212:00 PM EDT
-11:59 PM EDT
Make the most of your break with the most realistic GMAT™ prep. Take up to $700 off select products.- Jun 10
06:00 AM PDT
-06:15 PM PDT
Register for the GMAT Club Virtual MBA Spotlight Fair – the world’s premier event for serious MBA candidates. This is your chance to hear directly from Admissions Directors at nearly every Top 30 MBA program..
18 May 2019, 14:06
Kudos
Bookmarks
Hi all,
I just took the official GMAT practice exam #6 and am disheartened by my verbal score. I got 2 questions wrong (#29 and #33). I found both of those questions on GMAT Club and each was listed as 700 level. Pretty good I'd say! Well, my verbal split was a 42. Is it really possible to drop to a V42 having only gotten 2 questions wrong - both difficult questions near the end of the test? I'm very confused by this.
The verbal split on the other practice exams I've taken lately are:
Official Practice Exam #3: V47 (2 incorrect)
Official Practice Exam #4: V42 (5 incorrect)
Official Practice Exam #5: V51 (0 incorrect)
So.. based on these results it's hard to believe that 2 incorrect answers could result in a V42 as it did on Practice Exam #6. I'm pretty worried about this algorithm now...
Does anyone have similar experiences or any wisdom? Thank you!!
I just took the official GMAT practice exam #6 and am disheartened by my verbal score. I got 2 questions wrong (#29 and #33). I found both of those questions on GMAT Club and each was listed as 700 level. Pretty good I'd say! Well, my verbal split was a 42. Is it really possible to drop to a V42 having only gotten 2 questions wrong - both difficult questions near the end of the test? I'm very confused by this.
The verbal split on the other practice exams I've taken lately are:
Official Practice Exam #3: V47 (2 incorrect)
Official Practice Exam #4: V42 (5 incorrect)
Official Practice Exam #5: V51 (0 incorrect)
So.. based on these results it's hard to believe that 2 incorrect answers could result in a V42 as it did on Practice Exam #6. I'm pretty worried about this algorithm now...
Does anyone have similar experiences or any wisdom? Thank you!!
18 May 2019, 15:59
Kudos
Bookmarks
I wouldn't worry about that 42 at all...that verbal performance you've had over the last 4 official tests is pretty incredible and suggests that that only verbal score that really matters - the official one - is going to be really high (provided you execute the way you have been, of course).
My somewhat-educated theory (I've gotten to know Item Response Theory pretty well and have had a few happy hours with the guy who oversaw it for GMAC for a long time) on why that happened:
-To some extent adaptive test scoring is dependent on the algorithm having access to enough high-quality questions to accurately delineate between ability levels. For example if the test only had 300-level questions it would have a hard time differentiating at the 700 level.
-Within the last few years, GMAC pretty rapidly increased from just the two official tests and a fairly static, updated-every-so-often Official Guide series to 6 official tests, OGs updated every year with at least some percentage of new questions, etc.
-With a smaller pool of questions for each practice test to draw from the tests aren't as robust as the official test (which has a larger pool to pull from particularly at the higher-difficulty range) it stands to reason that the practice tests may not be as suited to differentiating a 96th percentile V42 from a 99th percentile V45. You may by GMAT Prep #6 just end up with enough "meh" questions (low weights, middle-of-the-road difficulty) that some of your right answers didn't move the needle as much as a high-powered, official test day question might, so your 28-question win streak to start the test didn't build as solid a 99th percentile "lead" as it might on the official test.
Anyway, my hypothesis here isn't that the algorithm did anything wrong...just that with these practice tests the pool of questions might be just shallow enough to be pretty darned accurate in the middle of the curves but have a slightly wider error margin in the tails of the curve than you'd see on test day. And the larger message is "don't stress a total of 9 wrong questions over 4 official tests" - take a look at those mistakes to see if you can shore up 1-2 things heading into test day, but don't let that 42 shake your confidence at all. With your ability going into a test that has a deep, powerful pool of questions, you'll be able to demonstrate your superior ability even with a couple missteps along the way.
My somewhat-educated theory (I've gotten to know Item Response Theory pretty well and have had a few happy hours with the guy who oversaw it for GMAC for a long time) on why that happened:
-To some extent adaptive test scoring is dependent on the algorithm having access to enough high-quality questions to accurately delineate between ability levels. For example if the test only had 300-level questions it would have a hard time differentiating at the 700 level.
-Within the last few years, GMAC pretty rapidly increased from just the two official tests and a fairly static, updated-every-so-often Official Guide series to 6 official tests, OGs updated every year with at least some percentage of new questions, etc.
-With a smaller pool of questions for each practice test to draw from the tests aren't as robust as the official test (which has a larger pool to pull from particularly at the higher-difficulty range) it stands to reason that the practice tests may not be as suited to differentiating a 96th percentile V42 from a 99th percentile V45. You may by GMAT Prep #6 just end up with enough "meh" questions (low weights, middle-of-the-road difficulty) that some of your right answers didn't move the needle as much as a high-powered, official test day question might, so your 28-question win streak to start the test didn't build as solid a 99th percentile "lead" as it might on the official test.
Anyway, my hypothesis here isn't that the algorithm did anything wrong...just that with these practice tests the pool of questions might be just shallow enough to be pretty darned accurate in the middle of the curves but have a slightly wider error margin in the tails of the curve than you'd see on test day. And the larger message is "don't stress a total of 9 wrong questions over 4 official tests" - take a look at those mistakes to see if you can shore up 1-2 things heading into test day, but don't let that 42 shake your confidence at all. With your ability going into a test that has a deep, powerful pool of questions, you'll be able to demonstrate your superior ability even with a couple missteps along the way.
18 May 2019, 17:13
Kudos
Bookmarks
I've coded an IRT based algorithm (the same kind of algorithm the GMAT uses), and everything Brian says above sounds correct to me. In the additional official tests (the Exam Pack ones), I haven't analyzed the pool of Verbal questions, but it does seem that in Quant, sometimes the overall difficulty level of the questions can skew very easy. And as a consequence, you need a lot of right answers to get even a halfway respectable score. Conversely, on the older free tests, which had a huge supply of very hard questions, you could afford a ton of wrong answers and still get a great score, provided those wrong answers were only on the hard questions.
So what's likely true is that you predominantly were answering easy questions (correctly), which persuaded the algorithm you were very good, but there weren't many questions to let the algorithm differentiate between very good and extraordinarily good. That's a fault of the question pool on those particular tests, not of the algorithm (which is simply probability theory - it really can't be wrong). That fault won't be replicated on test day, because they'll control more carefully what questions are available in the pool on the real test.
The old rule of thumb should still apply - provided the wrong answers are on hard questions (which they almost always will be), normally one wrong answer in Verbal produces a V49 or V50, and two wrong answers a V47 or V48. With three or more, you'll get a V46 or lower. Your Verbal scores have been outstanding, and I think you should likely view this particular score as a V46-V47, and not as a V42, since if you were to perform identically on the real test, you'd likely be in the 99th percentile range.
And I'd add, it does sound like a big difference, between say V45 and V49, but those are both 99th percentile scores. I don't think the test is designed to differentiate clearly between test takers at that high a level in Verbal - there just aren't enough questions that V45 test takers usually get wrong but V49 ones usually get right. I'm not even sure what such a question would look like. That's not at all true in Quant, where there's a very big difference between a Q45 and a Q49, a difference the test is very good at recognizing, since there are lots of math questions a Q45 test taker will find difficult that a Q49 test taker will usually answer correctly.
Anyway, I wouldn't worry about this particular score, and since it's extremely hard to improve at Verbal once you're scoring in the V46+ range, as long as your pacing is fine, I'd say you're very ready for your test, at least in the Verbal section. Good luck!
So what's likely true is that you predominantly were answering easy questions (correctly), which persuaded the algorithm you were very good, but there weren't many questions to let the algorithm differentiate between very good and extraordinarily good. That's a fault of the question pool on those particular tests, not of the algorithm (which is simply probability theory - it really can't be wrong). That fault won't be replicated on test day, because they'll control more carefully what questions are available in the pool on the real test.
The old rule of thumb should still apply - provided the wrong answers are on hard questions (which they almost always will be), normally one wrong answer in Verbal produces a V49 or V50, and two wrong answers a V47 or V48. With three or more, you'll get a V46 or lower. Your Verbal scores have been outstanding, and I think you should likely view this particular score as a V46-V47, and not as a V42, since if you were to perform identically on the real test, you'd likely be in the 99th percentile range.
And I'd add, it does sound like a big difference, between say V45 and V49, but those are both 99th percentile scores. I don't think the test is designed to differentiate clearly between test takers at that high a level in Verbal - there just aren't enough questions that V45 test takers usually get wrong but V49 ones usually get right. I'm not even sure what such a question would look like. That's not at all true in Quant, where there's a very big difference between a Q45 and a Q49, a difference the test is very good at recognizing, since there are lots of math questions a Q45 test taker will find difficult that a Q49 test taker will usually answer correctly.
Anyway, I wouldn't worry about this particular score, and since it's extremely hard to improve at Verbal once you're scoring in the V46+ range, as long as your pacing is fine, I'd say you're very ready for your test, at least in the Verbal section. Good luck!
