Higher accuracy, but lower score

[EmanueleB476]

Hi Everyone!
Can someone please help me explain why I got a lower Q score on the real thing, achieving an higher accuracy?

08/12/19 CAT5 50% 50% 88% 43% 42
14/12/19 Official3 63% 75% 50% 43% 42
19/12/19 Real TEST 1 71% 57% 71% 57% 39

CAT5 is from Manhattan
Official 3 is the 3rd official simulation test from GMAC
Real Test 1 is my first attempt on the real exam.

I hope I can learn something for the next time!
Thank you very much in advance!
Manu

[StaceyKoprince]

Hi! The scoring on this test is weird.

It's not based on percentage correct, which is what our school tests were based on—so basically, pay (almost) no attention to how many you got right. It's very common to see a higher percentage correct but lower score (or vice versa).

The test is re-calculating your score after every question you answer and using that data to determine which question to give you next—so everyone is literally taking a different test. It's not like school-based tests, where everyone was taking the same test so your teachers could just add up the # you got right.

The GMAT is using an algorithm to calculate your score based upon the difficulty levels of the problems that you got right vs. wrong. One way to think of that is that the algorithm is trying to find your "60% level"—the scoring level at which you can consistently answer about 60% of the questions correctly.

The test is also what we call a "where you end is what you get" test, so you could lift up to a higher scoring level in the middle of the section, but if you drop down by the end of the section—where you end is what you get.

So basically, you're trying to have a pretty steady trajectory across the exam (so that you don't crash at the end), while not missing too many "lower-level" / easier-difficulty questions (because that will reduce the difficulty of the future questions you receive—and will therefore reduce the score that you can earn).

In terms of interpreting the data that you shared, I'm not sure what the data is—for the official ones, are you showing the percentage correct for each of the 4 quadrants of the exam? (But I'm confused because we don't provide this data for our own exams—so I'm not sure what data you're showing for our exam. Ours will tell you, for example, percentage correct for DS and PS—which the official report also tells you—but that's only 2 data points, not 4.)

This blog post shows how to analyze the official Enhanced Score Report:
https://www.manhattanprep.com/gmat/blog/enhanced-score-report-part-1/

I'm also happy to help you try to interpet if you tell me what the data is that you've included.

[EmanueleB476]

Hi Stacey,
and thanks a lot for your answer!!

Yes, I know you don't provide those results, but I kind of tried to understand more about the algorithm, and so I also analyzed your exam using excel, so yes those are approximately the quarter percentages, which I wanted to compare with my exam (ESR). Do you see anything useful? I have my new exam in two weeks and your advice is more than welcome!:-)

Thank you very much in advance!
Manu

[StaceyKoprince]

Ah, ok! Understood.

For that set of data points, I'd also need to see the data for avg time per quadrant and that difficulty-level chart—basically whether it was going up or down from quadrant to quadrant. Since the algorithm isn't based just on % correct, you need to look at all of those data points together to try to get a sense of what was going on. (I talk about that a bit in the article I linked last time.)

I'm going to use just the data from the official and official practice test, as those are based on the same algorithm. (We had to build our own, and while we try our best to mimic the real test, it's still not the same algorithm.)

It looks like you started off well in both tests—63% or 71% correct (which probably represents only 1 extra question right for the higher number). In the official practice test, you likely continued to lift in the second quadrant because you also had a high percentage right there—while on the official test, the % correct dropped in this quadrant, so you either stopped lifting your score or (if you missed any lower-difficulty-level questions) the score actually went down a bit.

It seems like that would balance out in the third quadrant (because the practice test dropped but the official test jumped up)—but remember that you are at a higher level on the official practice test. So you're answering harder questions in that third quadrant. You did have a drop in % correct in the final two quadrants, but if you managed to get the "lower level"—of the questions you were offered in those two quadrants—problems right, and just missed the "even higher / this would lift your score more" ones, then you'd be roughly maintaining where you were after lifting high by the mid-way mark.

On the official, you likely did still lift your score somewhat in the third quadrant, but it was not quite enough to get you all the way up to where you were performing on the official practice. It may be that, in the third and fourth quadrants, you missed some easier ones and that limited how much you could lift your score.

Your final scores were not that far apart (39 and 42). I'd call that "statistical noise"—within the major of error. There's a good chance you can already hit a 42-level score on the real thing, too, but you'll have to make sure that you aren't (for example) spending too long on certain problems and then having to rush others, which can result in missing lower-level questions due to careless mistakes. (And that will bring your score down.) It also means making sure that you don't have any holes in your foundation—that you aren't going to miss lower-level problems in certain areas of weakness, and then if you happen to get a few more of those on the real thing...now your score is 39 rather than 42. That kind of thing.

What do the timing and difficulty level charts show? And what about the rest of the data in that report?

[AndrewH809]

Hi,

Just a follow up on this question on the "where you end is what you get"

Would it be more beneficial if I spend moderate amount of time for the first 15 questions of a Quant or Verbal section (let's say 1:30 average each), and then really focus on the last 15/20 questions?

If I focus too much on getting the first 15 right - I might be wearing myself out / running out of time towards the end.

The way you described the testing methodology, it seems to be more beneficial to focus on the last 15 or 20 questions. Is that right?

[StaceyKoprince]

What's really best is to have a pretty steady performance across the entire section. If you go too quickly in the first 15 and don't lift your score as much as you're capable of, then you'd find that it's harder to lift your score as high as possible when you only have ~15 questions left.

The basic idea is this: You're going to see a range of questions, some of which you can do easily, some of which you can do in roughly normal time but with some effort, some of which you could do with a lot of extra time, and some of which you really can't do no matter what.

For the ones you can do easily: Don't miss them. Be super systematic with your thought process and your scratch paper work. You don't need to spend extra time on them but don't rush them and risk careless mistakes.

For the ones you can do in roughly normal time (within about 30-45 seconds of the average time for that question type) but they're not easy for you: Again be very systematic / careful. Here's where you may want to invest an extra 15-30-45 seconds at times / when warranted.

For the ones you can do but only with a lot of extra time (1+ min above the average time for that question type)—not worth it. See whether there's a good way to narrow down answers, but make a guess and move on before you've spent extra time on this problem.

For the ones that aren't happening no matter what, try to recognize this within the first 30 or so seconds—and bail immediately. Save that time and mental energy for the second category above.

What I just described is harder than "spend more time on <earlier / later / middle / whatever> questions"—because you actually have to think / evaluate for yourself. You know...just as you have to evaluate various business questions / opportunities in the real world. There's no such thing as a "consistent, easy way to success"—for that path, there are just people who got lucky but can't replicate that kind of success another time.

[KiraL598]

Thanks for the help

[StaceyKoprince]

Do you have a question for us?