760+ and the Myth of Level of Difficulty?

[cetspeck]

Hi,

This will be a long post, but I would really like some knowledgeable feedback about very high scores on the GMAT.
This is really important because my current full-time position will be ending soon, and I'm trying to get a 760 so I can get employed by a test-prep company that pays really well (any suggestions, lol?) and am right at the 750/760 mark.

I've seen posts about number of questions and level of difficulty, but the data below (sorry, takes a while to wade through!) suggests that the number of questions answered at 760+ makes a huge difference between 750 and 760+.

GMATPrep 1 (Q48: 13 questions missed, due to some unfamiliar problems and some careless mistakes, knew during the section that I was struggling)
GMATPrep 2 (Q48: 11 questions missed, ditto)
GMAT Actual Test 1 (Q48: at least 6-10 unfamiliar problems I knew I was lost on and probably missed; of course, it's hard to tell how many of them were experimental; still, pretty much knew during the section that I was about where I was for GMATPrep1 and GMATPrep 2, i.e., struggling, so Q48 not a surprise)

(2 months later)
GMATPrep 3 (Q50: 5 questions missed, knew I was doing well, much better than Prep 1, Prep 2, and Actual 1, lots of familiar but challenging problems)
GMAT Actual Test 2 (Q50: knew during the section that I was doing really well, very similar to GMATPrep 3, so Q50 not a surprise)

It seems on these 5 tests -- and also the 6 Manhattan GMAT tests I've taken -- that the difference between a Q48 and a Q50 has less to do with an increasing level of difficulty and more to do with getting an additional 5-8 really hard problems correct. It seems like not much return in terms of score increase, but maybe that's why the percentile difference between Q48 and Q50 is something like 80th percentile and 92nd percentile?

GMATPrep 1 (V46: 3 questions missed, knew I was doing well)
GMATPrep 2 (V46: 2 questions missed, ditto)
GMAT Actual Test 1 (V46: again, knew I was doing well, except for a few at the very end that suddenly seemed more difficult than the rest)

GMATPrep 3 (V42: 5 questions missed, knew during the section I was missing a few more than in previous tests, but down to a 42?)
GMAT Actual Test 2 (V42: the first half I knew I was doing really well, and then the verbal section got significantly more difficult -- CR questions almost seemed like some LSAT questions, and one RC passage was almost incomprehensible; I knew I was missing a few more questions than when at V46, so V42 was not a surprise because this mimicked GMATPrep 3 but still, it seemed really unforgiving)

Again, it seemed that just a few questions and suddenly the verbal starts plummeting. Here are the combined scores:

GMATPrep 1: 760 (Q48, V46)
GMATPrep 2: 760 (Q48, V46)
GMAT Actual 1: 750 (Q48, V46 -- 98%, argh!)
GMATPrep 3: 750 (Q50, V42)
GMAT Actual 2: 750 (Q 50, V42 -- 98%, double argh!)

It seems, then, that in order to get a 760, you need to:
(1) only miss 3-5 math questions to get a Q50, or maybe 6-8 questions to get a Q49?
(2) only miss about 3 verbal questions to get a V46, or maybe 4 questions to get a V45/44.

Maybe nobody really understands the algorithm and Person #1 got a V44 and missed 7 questions, etc., and yet it seems that the GMAT at these very high levels is unforgiving, much like the SAT Math, where if you miss just 2 question on the entire Math portion, you can be down to a 780 or even 750.

The other interesting thing I noticed -- and this happened a lot when I took a bunch of GRE tests -- was that on GMATPrep 1, GMATPrep 2, and GMAT Actual 1, the math seemed more difficult than the verbal. And vice versa for GMATPrep 3 and GMAT Actual 1.

Again, any thoughtful and knowledgeable feedback would be great. Thank you.

[StaceyKoprince]

You're absolutely right that the test is super-unforgiving at the highest levels and that you do have to start getting a higher percentage of questions right at the highest levels. That's not true for most of the test levels, but those factors do start to come into play at the very top.

We don't know the exact numbers, though, and there's also the whole experimental wildcard. We can't draw conclusions 100% directly from GMATPrep because GMATPrep uses only fully tested questions, not experimentals.

Plus, on the real test you could miss every single experimental and several more, so you could miss more than just a few and still get a 99th percentile score. (Though the odds are that you will get most experimentals right - if you are at that high a level, the odds are that most experimentals will be below your level / easier for you.)

In terms of pushing yourself over that last little step, have you read this article?
http://www.manhattangmat.com/articles/7 ... erence.cfm

Really, you're there - it's a matter of being able to put together both performances on the same test day. You may also want to take a look at this:
http://www.manhattangmat.com/blog/index ... you-crazy/

And make sure your timing is DEAD on; you can't afford even slight timing problems at the 99th percentile level!
http://www.manhattangmat.com/blog/index ... anagement/

Good luck - let us know how it goes.

[StaceyKoprince]

oh, and one more thing. REALLY try not to focus on the # right or even thinking about that at all. There's NOTHING you can do about that in the middle of a test. You either know how to get it (in a reasonable amount of time) or you don't. If you don't but obsess about how many you've gotten wrong already or whatever, you'll just blow time and mental energy, which will make it less likely that you'll get your score in the end!

[cetspeck]

Hi Stacey,

Thanks for the info about the test being super-unforgiving. That's good to know and I hadn't really run across that info before. The GMAT in that sense is a lot like the SAT math, where 1-2 wrong answers can drop you from a 800 to 750-780.

I had already read both articles you recommended, but it never hurts to re-read.

Thanks again.

[StaceyKoprince]

glad to help!

the super-unforgiving thing is also why it's so critical to maintain your time, especially if you're going for a 99th percentile score.

If you are a 780-level tester, and you spend enough extra time on some earlier questions that you have to guess on more than 1 question at the end (and those 2+ questions aren't experimental and you get them wrong), your score will drop out of the 99th percentile.

Crazy!!