(Un) Scientific analysis of gmatprep

installing and reinstalling won't do anything..the program is stored in your registry..you would have to find a way to delete from the registry

Best thing to do is the take an exam and don't review what you got wrong/right. You can review what you got wrong, as far as question number, but don't review the answers.

Is there a lot of overlap, when the two tests are taken for the second time?

What I read on this and other forums, you can take test 3-4 times easily without much overlap (i.e. 4-5 questions). GPrep has good amount of questions.

In general, I have notified an obvious correlation of getting a substantially higher score the less goofy I am on the first 9 questions.

The myth seems somewhat logical to me if I bring it into real life perspective. Although I know the logic and coding behind GMAC's software for determination of assessment, I think its somewhat common sense that initial assessments carry more weight rather than fine tuning one's ability past the first 1/4 to 1/5 of each respective section. Meaning the if then statements (obviously probably much more complicated in reality) matter more initially because thats a relative of how fast you can rise or fall in a mere few questions in the beginning. Kind like the phrase

initial impressions most?

Out of curiosity have any of you tinkered with the myth of the initial 5 or 8 or 10 questions on the GMAT Focus CATs? (Been hearing those are 2nd to none for Quant prep.

now it'd easier to contact Princeton (or who it may be) to probe the binary\statistical algorithm its based on... but must admit the dedication and energy on this thread... I could think some of they guys could become great security experts and the other bit more edgy evil minded could become hackers...

Nice read really.

This was great analysis and effort devoted is highly appreciated.

What I believe is they have hired smart people as we pay a lot to give gmat lol.
So, they must have used complex algorithms or artificial intelligence programs to understand and mould the level of question anytime.

Eg. They will throw question randomly from level 550-650 first and based on first few question..say 4-5 or 7-8 they will guess the level. Though after every wrong/right ans level would have been varied.

Now algo had decided we are 650 level, now if we ans question above 650 level we will get actual + bonus mark( these marks are not actual marks, they might be anything used for analysis by the algo) and if we ans the question wrongly below 650 we lose actual + penalty mark. And side by side our initial level is also varied and they keep on throwing questions above/below depending upon our ans.
And these bonus/penalty marks must have been different for different level and number of question answered right/wrong in a row. So just by getting 4-5 right and then guessing a question will not harm.

Now its obvious number of question with difficulty level in their DB must be like this
750+ < 700+ < 650+ < 600+ < 550 +
So more the number of questions in a particular level we exhaust, more likely to have that level in the analysis.

I would like to request BSD to experiment this::: in quant we have 37 questions.. after every 4 right, ans one wrong and in total you will have around 9 wrong.
again after every 8 right, ans 2 wrong in a row and again you will have 9 wrong.
again after 2 right one wrong, 6 right one wrong and again you will have total 9 wrong.
Just compare the results.

Getting 12 wrong answers is not the same as getting 12 consecutive wrong answers. Your claim is valid only if you miss 10 consecutive ones in the middle/end and see totally different results compared to missing 10 in the beginning.

I'm not surprised by your Verbal experiment's outcome. You are essentially saying that one of SC or CR is more important than the other. However, that can't be the case, because an easy SC question should have the same weight as an easy CR question. Likewise, a hard SC question should have the same weight as an equally hard CR question. So when you get all SC correct and the rest wrong, you're essentially getting relatively the same amount of hard questions right as when you get all CR correct and the rest wrong.

Hope the above makes sense.

excellent work!! keep on posting updates

In verbal we have sc, rc and cr.
However, in math we have different topics.
But what if you get wrong a geometry question, sometimes, I think that later appears another geometry question.
What do you think?
Don't you think that sometimes there are some basic topics that are always cover and if you get one of this topics wrong, another question of this topic is going to appear in the exam.

anyway, great work and very interesting results!!

great post buddy.

I'm wondering. I kaplan free practice test, I got 4 incorrect in quant. it showed 45 scaled score. I was expecting atleast 49 here, but my first question was wrong. Does 1st incorrect answer scale down your score 4 points? or is just kaplan's strategy to enroll me for the course?

Just got a 50q on the GMAT Prep. I took it without trying to find patterns. Also, I only ran into 2 questions I had seen previously. One I got wrong and one right, but not on purpose. I took the questions as I got them. Here is some info. in case it helps.

2 out of the 1st 10 wrong: #3 and #8
Then got on a hot streak until I reached #20, which I missed
I also missed #22
Then I missed #28, 29, 31

Total 7 Incorrect:
More importantly 2 out of the 1st 10 were incorrect. Still got a 50 so don't know if that proves it's a matter of streaks more than anything else (or not missing too many consecutively)

1st two verbal questions were very easy. I then skipped the rest of the questions.

7 incorrect still 50, that's great.

Best we could do is to take one extra minute just after the question we are sure of being incorrect or randomly guessed.

Wow guys, this is awesome. Now I won't stress quite as much when I'm panicking over an early quant question.. wait no, I still will :[

But I would still suggest you to kill the initial questions with extra caution and make sure you don't miss consecutive questions. It should be intersperse throughout.

I took 2 GPREPs in the last 3 days.

GPREP 1 - Q43 - Overall 20 Wrongs. 7 Wrongs in the 1st 20. Last 10 (28 - 37) - All wrong. Ran out of time, had to guess them.

GPREP 2 - Q42 - Overall 13 Wrongs. 8 Wrongs in the 1st 20. Only 3 Wrongs in the last 10. None consecutive wrong in last 10.

I had a good feeling about Quant this time. I was expecting 46 - 47. Bump, it was 42 and 1 less than previous attempt despite fewer Incorrects.

This goes to say that 1st 10 or maybe 1st 20 matters, cuz the software fairly fixes your range by this time and your performance in the last 10 or 7 doesnt really help or hurt much.

One more observation in both Quant attempts, I got 2 consecutive wrongs 3 times in the 1st 20. Maybe consecutive wrongs further helped to downgrade my Quant level - range.

GPREP 1 - V38 - Overall 14 Wrongs. Got all RC right. SCs - 8 Wrong. CRs - 6 Wrong. I spent lot of time on RCs, hence ran out of time towards the end and had to guess like 10 Questions.

GPREP 2 - V35 - Overall - 12 Wrongs. RC - 4 Wrongs. SCs - 6 Wrong. CRs - 2 Wrong.
I wanted to replicate the same here. Get all RCs right by spending more time so that I can assure myself of V36 atleast. But 1 RC came after 33 and had to guess those 3 or 4 of those RC Qs and ended up with 2wrong out of it. I think if I had replicated the same here, like all RCs right, I would have hit V38.

What do you guys think of this Verbal strategy? I know not the best of strategy, but if RC was my strength, why not put that wee bit extra time to minimize the RC errors like only 1 or 2 RC errors.

GMATPREP1 I have taken last Sunday (without the AWA):

Q50- 6 errors: 2 in the first 15Q's, 1 between 16-30 and 3 between 31-37. (3PS, 3DS)
V42: - 5 errors: evenly distributed: 3SC, 1RC, 1CR.

As a rule of thumb, when you are hitting the top scores in Quant, around (+-2) 12 wrong questions give 49, 6 wrongs questions give 50 and 3 wrong questions 51.