Accuracy vs GMAT Score / Percentile

[ManishH404]

Hi Stacey

I am 4 weeks into my Full Online class and ~2 mths away from actual GMAT. Your blog posts and responses on the forum have been immensly helped in understanding what GMAT tests and how to prepare for it (2nd level learning).

There is one question, however, I have been searching for an answer to, but haven’t been able to find neither on MGMAT nor on other GMAT forums. My question is
What is the accuracy rate of students who score 99 percentile on GMAT? (in other words – relationship between accuracy and GMAT scores)

Reading your post http://tinyurl.com/kbaodqz, it is crystal clear is that (1) GMAT is about making decision on where to guess and where to solve, and (2) No test-takers answers all questions correct (apart from some or all GMAT tutors). But Whenever I make a guess n a question, my confidence dips and anxiety upps a notch, which makes me nervous and the spiral goes deeper as I progress. However, If I knew that an average 99% with say 60% accuracy, I am not drastically wrong if I guess between 2 good answers (50% hit rate).

To become smarter in time management, I need to get myself comfortable with educated guessing . And I think getting some statistics, and your view on this would make that acceptance easier.


Thanks again
Manish

[StaceyKoprince]

The GMAT doesn't publish this info, but my best guess is that someone scoring in the 99th percentile (that is 760+!) is probably answering about 70% to 90% of the counted questions correctly.

Remember that up to about 10 questions in each section are experimental questions (that don't count towards your score). You can get every single one of those wrong and not have your score affected one bit. Of course, you don't know which ones count and which ones don't...

If, on a question that you cannot do, you can narrow things to 50/50, then you are in a GREAT position - because all of the questions that you do know how to answer, you are answering at greater than a 50/50 rate!

But Whenever I make a guess n a question, my confidence dips and anxiety upps a notch, which makes me nervous and the spiral goes deeper as I progress.

This is because, at heart, you still believe that this is an academic test, not a business test. You need to get to the point where you don't just intellectually know that the GMAT is a business / decision-making test, but you actually feel it.

Read this article every day for the next 2 weeks:
http://tinyurl.com/executivereasoning

Come back after the first read and tell me why I told you to read it. Then do the same on day 7 and day 14.

And use this analogy to help: You're running a division at your company. You have a quarterly budget. Consider these scenarios.

(1) You spend the entire budget 7 weeks into the quarter. What is your boss going ot think?

(2) Your subordinate Jim wants to go to a conference next month that will cost $1,500. If you allow him to go, then you can't send 2 other subordinates to a different conference that costs the same total amount.

(3) Your subordinate Jim wants to go to a conference next month that will cost $1,500. You're not sure how else you might spend that money this quarter, but the conference is tangential to the company's core business and you're not entirely sure that Jim will actually make relevant contacts.

Answers:
(1) Your boss is going to fire you. :)
(2) There isn't necessarily one right answer here - there are so many variables involved - but you have to make the best call that you can with limited info. If you assume that your goal is to train the greatest number of people possible, then you'd rather spread that money (time!) over two people (questions) vs. just one.
(3) You can't know for sure that there's going to be a better place to spend that money (time!) in future, but you already have a sneaking suspicion that the current conference (question!) is not a fantastic use of that money (time!). So tell Jim no.

Basically, there are always trade-offs in business. Your constant task is to figure out the best way to spend limited resources (time and mental energy, in the case of the GMAT), knowing that you will never have enough to do everything (as in business).

School tests were built for you to be able to do everything in the time given (if you learned the material well enough). The GMAT was NOT built that way because this test is NOT an academic test!

[ManishH404]

Stacey - thank you so much for taking time and effort to respond to my question to this detail.

Your explanation make so much sense. I have read so much on GMAT tests, but no one put it like this. Your other posts did game me some perspective but this one makes it crystal clear. Knowing that GMAT is not all about memorising (was never good at it) but more about identifying strengths and making smart decisions (don't waste time where you can't add value) gives me much confidence.

I will re-read your blog post as advised and will update you.

Many thanks Stacey!!
ps: 70% accuracy for 99 percentile sounds unreal, but does make sense.

[StaceyKoprince]

You are very welcome! Let me know how it goes.

[JaydeepS583]

Hi Stacy,
i also started to give my mgmats now. But have couple of questions with the mgmats scoring and questioning techniques.

Before i mention them, I'll give a quick overview of my journey so far. I have been trying to give gmat for the past 2 years now (with doing job and all to which I was giving higher priority so far and taking gmats lightly), where for some reason, I seem to be stuck at 650 notch, with always scoring 48/49 on quants but only 30 on verbal. I seemed to take it a wee bit lightly and being a bit of a lazy bum, never took the effort to improve but continued to give exams after very little effective reading, so far.
Now, i started to buy exams and gmats. Earlier, I purchased princeton 4 tests where I somehow reached 34-35 which also seemed to reflect in the 1st 3 of my mgmats. Here, my only source of learning was the explanations against the failures in the above mentioned tests.
I intent to hit a minimum of 42 in verbals such that when paired with 48/49 q, it will result to score around 750, a minimum of which I am targeting.

Coming closer in the overview to my questions, I would mention this as well that for the 1st 3 tests in mgmats given so far, I was primarily focusing on the verbal questions only, due to which gave all of them without the IR, posted 1 liners in AWA and tried to rush through the quant sections as quickly as possible, trying to answer as much correct as possible (without trying to invest time in getting them correct). This kept me at 43q, but it is currently out of my scope with an assumption that I will keep my 48/49q calibre in real gmats. I intended earlier to give the whole exam with proper answering for IR and quant for the last 3 mgmats but it seems from where I am now, I will get to do that only in the 6th mgmat.

Now the 2 questions I have with mgmats:
1. Does mgmats have the experimental questions like real gmats? The reason for this question is that I am wondering where do I actually stand in terms of verbal with actual gmats (with experimental question included)?
2. The mgmat verbal question for the 700-800 difficulty (which I tend to get the most in my tests) are how much comparable to the difficulty levels of the real gmats?

Hope to get a response soon as I really want to be done with gmat exam soon with a score, somewhere above 750 since gmat is all I have been doing for the past 2 years.

[StaceyKoprince]

The most important question to answer is one that you didn't ask. Stop wasting practice tests - don't rush through the earlier sections and take only verbal seriously! There aren't that many high-quality practice tests out there. Plus, when you rush through the three earlier sections, you're not getting a true gauge of your verbal score, since your score is affected by the mental fatigue associated with doing the earlier three sections!

We do not include experimental questions, but this doesn't impact your result - when you take tests under official conditions, your scoring on our tests is fairly close to your current scoring level if you were to take the real test. (so this does not apply when you mostly skip earlier sections and only do the verbal - that artificially inflates your score.) Our verbal questions are also generally well-calibrated to the difficulty on the real test (at all levels, not just the 700-800 level). This is part of what allows our tests to predict your real-test score with good accuracy (within about a 50-point standard deviation, compared to a 30-point SD on the real test).

Next, let's talk about how you learn: it's NOT by taking practice tests (or even doing practice questions) and learning from the written explanations of those tests. That's maybe 15% of your possible learning. Here's how you actually learn:
http://tinyurl.com/2ndlevelofgmat

Coupled with this mindset:
http://tinyurl.com/executivereasoning

If you'd like to get more detailed analysis / advice, first think about how what you've been doing does and doesn't match up with those two articles and how you may need to change your approach accordingly.

Then, take a FULL MGMAT CAT (including essay and IR) and use the below to analyze your most recent MGMAT CATs (this should take you a minimum of 1 hour):
http://tinyurl.com/analyzeyourcats

Based on all of that, figure out your strengths and weaknesses as well as any ideas you have for what you think you should do. Then come back here and tell us; we'll tell you whether we agree and advise you further. (Note: do share an analysis with us, not just the raw data. Part of getting better is developing your ability to analyze your results - figure out what they mean and what you think you should do about them!)

[JaydeepS583]

Hi Stacy,

Thanks for the response and information provided.

I went through the links provided by you in the reply. I understand completely and agree to what you are saying.
Now, so far, i have already practiced beyond 1000+ verbal practice questions across sections, using various books and apps tried.

However, based on my history, the quant didn't really seem to be a problem for me when I took the GMATs (q49, then q48 and q49), but I seemed to be stuck at 30v (v30, then v27 and v29). Same was the case of the practice tests given, regardless of weather I give the full exam or just the verbal section, rushing through the rest.
Also, another point of observation was that the wrongs were not because of lack of fundamentals or something but actually because of not getting the answer correct, the GMAT way (eg: questions like inference should be drawn completely from the para available and not use any gk assumptions which we might derive based on understanding, 'like' vs 'as' questions, not using 'whom' for living things other than humans, etc.), an information, which I couldn't seem to find in any books but only in explanations.
I would really appreciate if you can point me towards a book or some material that provides this info.

Based on this analysis only, I had come up with a plan, such that I'll give the practice tests and learn the GMAT ways first, get my standalone verbal score to a v45/ 46, post which I'll start giving the full cats, to get final output as q47/48/49 combined with a drop in verbal, due to fatigue, to a score of v42/43, whose effective output would be somewhere around 750!, a score I desire.

For some reason, this strategy seems to work for me, such that after giving 11 gmat practice tests (2 free gmatpreps, 4 princeton & 5 mgmats), rushing through the quant and focusing on verbal and its explanations, taking the fatigue factor out, I am currently ranging somewhere around 670, with a q43/44 (despite rushing through quants) and a gradual consistent increase to v37 in the 5th mgmat from v31 in the 1st.

In this course, the segregation of difficulty was as below:
v31:
Row Labels Correct Wrong Grand Total
300 - 500 1 1
500 - 600 5 2 7
600 - 700 13 10 23
700 - 800 3 7 10
Grand Total 22 19 41
v37:
Row Labels Correct Wrong Grand Total
500 - 600 1 1 2
600 - 700 7 1 8
700 - 800 18 13 31
Grand Total 26 15 41

Please advise if you still feel that I am on the wrong track and need to change something, and if yes then what and how of it?
Also, based on the information above, do you feel that in the gmat exam, from where I am right now, with the experimental questions included, will I be able to stay on atleast v37 or will there be a + or - ? I am assuming that the score obtained with focusing primarily on the standalone verbal of the mgmats would stick close since the lack of fatigue factor will be counter-balanced with the availability of the experimental questions. Am I right with this thought?
Also, unfortunately, I seem to have exhausted all of the mgmat exams and I need more. Can you please help me point out where can I get them as well? For now, I am re-setting and continuing with the hope that only some and not majority of the questions will be repeated such that I continue on the learning curve. Thoughts?

[StaceyKoprince]

(eg: questions like inference should be drawn completely from the para available and not use any gk assumptions which we might derive based on understanding, 'like' vs 'as' questions, not using 'whom' for living things other than humans, etc.), an information, which I couldn't seem to find in any books but only in explanations.

All of these examples that you gave are in our books. Inference questions for both RC and CR are taught this way - only use the source text; the test is not actually asking you to come up with a real-world conclusion, but with a "what must be true based on the given information" deduction - like a scientist concluding something, etc.

Like vs. as is discussed in the comparisons chapter of our SC guide; it's got its own sub-heading in the chapter. Who/whom is discussed in the modifiers chapter of the book (first sentence page 61 of 6th edition: "The pronouns who and whom must modify people.").

So part of the issue may be that you are not studying from the books in a way that allows you to fully retain this information when you get to the test. How do you study from the text of the books? You read it obviously. Then what? Take notes? Highlight / underline? Make flash cards? Etc? (Is part of the issue that you've been focused so much on doing tons of questions that you're not actually able to make great memories on all of this stuff that you're trying to learn?)

get my standalone verbal score to a v45/ 46

The two sets of scores (Q and V) are not really scaled the same. This is the 99th percentile on verbal. In other words, your goal is to get your verbal score to be essentially higher than quant, your big strength. Probably not the best way to set the goal!

I'll give the practice tests and learn the GMAT ways first

Yes, this can work. For most people, it's an inefficient way of studying (because you're basically reinventing the wheel - you have to learn everything on your own vs. first learning the content and strategies via books or lessons designed to teach you this stuff without you having to "derive" the lessons on your own). But people who have generally liked and been very good at standardized tests in the past can find this a good way to study. (These tend to be those annoying people who always did very well on standardized tests without really seeming to try...like me.

So if that is working for you, great. But two things:
(1) You may learn some material better (and faster) if you concentrate a bit more on the books / lessons that already distill the necessary information for you; this is true for most people and it sounds like it may be true for you on verbal, given what you told me about not knowing certain rules or strategies that actually are in our books...
(2) V44/45 is a crazy high score. Go for more like a V38 to 40, which is roughly equivalent to your quant level. If you score Q49 and V40, you're at about 720, more than good enough for any program on the planet. (And even if verbal drops to 38, you'll still be at about 710.)

On your experimental / fatigue question: no, the experimental questions are not going to balance anything. Our tests still predict your real test level, even if we don't have experimentals on ours. So assume that, because you are not taking those verbal sections under normal conditions, your scores are somewhat artificially inflated. (This may not be true for you - some people can maintain that mental energy across 4 hours of testing. Most people can't, though, and are negatively affected by mental fatigue. So assume the worst, just in case.)

Also, the best practice tests available are the 4 GMAT Prep tests and our 6 tests. You can keep re-taking the tests once you've used those up, but you do run the risk of seeing repeated questions. You can also try other company tests, of course. They just aren't as good as ours or GMAT Prep.

[RebeccaF792]

Hi Stacey,

Thank you for the thread. Is there an article or breakdown of how the combinations of Quant and Verbal scores sum to the overall score? For example, how you are able to know that Q49/V40 leads to an overall score of about 720?

Thanks,

Rebecca

[StaceyKoprince]

There isn't an official publication of this - I just know because I track the scores students report to me.

We're talking about making something that we can publish. There's just one complex bit that we haven't figured out how to display in an easy-to-read way: a particular pair of Q and V scores doesn't always resolve to just one three digit score. There's variability. For instance, Q49 / V40 could lead to a score of 710, 720, or 730 even from three people taking the test on the exact same day.

Your next question is: wait, how is that possible? (And that doesn't seem fair!) If you really want to know, keep reading. But feel free to skip this paragraph and just believe me that it happens and it is fair. The 2-digit scores are not what's used to calculate the 3-digit scores. Rather, you earn certain raw scores for Q and V, and those raw scores are combined to produce both the 2-digit scores and (separately) the 3-digit score. There are more 3-digit scores available than 2-digit scores - so, basically, they can be more precise with the 3-digit scores. A simpler example: You take a raw score and map it first to a scale of 1 to 10, in 1 point increments. Let's say two people each earn a 6. Then, you map that same raw score to a bigger scale of 1 to 100, also in 1-point increments. Now one person might earn a 5.8 and the other might earn a 6.4. Both would earn a 6 on the 1-10 scale, but one actually did better than the other and earned a higher score on the 1-100 scale.

Anyway, so any particular Q/V combo appears to be able to map to 2 or 3 different 3-digit scores...and that's pretty hard to show visually. So we're still working on how to dispaly that in a way that makes sense!

[JaydeepS583]

Hey Stacy,

I did read quite a number of books (didn't have the mgmat books since never knew about them till late) already and did the notes, flash cards stuff, etc. but to my unfortunate fact, I never came across of these logics in the books I read. Hence, the decision of learning through practice questions. So memories, etc. is not the issue.
Now, coming to the repetition of tests part. I re-used the tests that I had given and scored a 720 (v42) in a test. However, on analysis, I found out that out of 41 questions, 23 questions were repeated. So now I am confused as to what would my real score have been? The breakup was something as below:

-------------------------------Repeated--------------------------------New-----------------------------------Grand Total
Row Labels----------------Correct-------------Wrong-----------Correct--------------Wrong
CR-----------------------------7----------------------0-------------------3----------------------4----------------------14
600 - 700-------------------0-----------------------0-------------------0----------------------2----------------------2
700 - 800-------------------7-----------------------0-------------------3----------------------2----------------------12
RC----------------------------9-----------------------3--------------------0----------------------0----------------------12
600 - 700-------------------3----------------------0---------------------0----------------------0----------------------3
700 - 800-------------------6----------------------3---------------------0----------------------0----------------------9
SC-----------------------------2----------------------2--------------------9-----------------------2---------------------15
500 - 600--------------------1---------------------0---------------------0----------------------0----------------------1
600 - 700-------------------0----------------------0---------------------1----------------------0----------------------1
700 - 800-------------------1----------------------2---------------------8----------------------2----------------------13
Grand Total--------------18----------------------5--------------------12---------------------6----------------------41

However, I guess I sadly have no other option now but to see off mgmat tests and buy the 4 gmat preps and study there. Do they provide as good explanations like you guys do?

Any thoughts? Suggestions?

I have enrolled for the 6 mgmats. So do I get any sort of question bank of any sorts to practice and continue with mgmats?
i have invested quite a lot so far so I don't think I will be able to buy the books now as I believe the 3 books related to verbal alone cost somewhere around 5g inr.

[StaceyKoprince]

Ah, okay - sorry, I thought you meant you couldn't find that stuff in our books.

So it turns out that buying one book gives you access to the 6 practice tests - so if you want the CATs, you might as well buy just one book and then you get the CATs too.

I would go for the SC book, because it would be hard to get all of those lessons by going through all my blog posts. For CR and RC, try these (they won't replace the books, but they'll give you the main lessons):
RC: https://www.manhattanprep.com/gmat/blog ... rehension/
CR: http://www.manhattanprep.com/gmat/blog/ ... reasoning/

Also, you can buy e-versions of the books instead of physical versions. They're less expensive (and no shipping costs).

Okay, 23 repeated questions is a lot. There isn't a way to tell what your score would have been without all the repeats, but one thing you can do is this:
- how many did you get right that you aren't sure you would've gotten right if the question had been new?
- how many did you answer much more quickly than you would've if the question had been new?

If the answer is only a few, then your score wasn't greatly inflated. If the answer is a lot, then your score was really inflated.

The GMATPrep tests do not offer any conclusions. You can search the GMATPrep forums here for solutions.