Confused about my score

[alig35]

I took my first ever practice GMAT yesterday using the GMATPrep software.

Here's my score info:

Score: 660

Raw score:
Quant: 46
Verbal: 35

I had 14 incorrect questions on the Quant section and 32 incorrect questions on the Verbal section.

How can I have a 660 when I had so many incorrect questions? It just doesn't make sense. I figure my score would be far lower than that.

[StaceyKoprince]

First - is there a typo in your post? You said you had 32 incorrect questions in verbal. Did you maybe mean 22? I wouldn't expect 32 wrong and a score of 35. So if you really did mean 32, then I'm surprised too.

If it was a typo, though, the answer in a nutshell: pretty much everybody gets roughly half right and roughly half wrong (say between 40% and 60%) no matter the scoring level (unless you get to very high or very low scoring levels). The number you get right really doesn't change much, even as you get better. What chnages is the difficulty level of the questions you're answering. If you want more detail about how the algorithm works, do a search for "algorithm" on the board here and read some of the posts that discuss this topic. Or you can just know that you're always going to get a decent number of questions wrong. :)

Also - nice job! 660 is a great score and this was only your first test!

[Also Confused-Score.Calc]

On my first practice MGMAT I got a raw score of 45 on the math and only faced 10, (700-800) level questions. When i took my 2nd MGMAT test i got a raw score of 40, but they gave me 20 (700-800) level questions? Why the huge discrepancy? I was obviously getting most of these wrong so why would they keep giving them to me.

Also on the second test I took, towards the end of the verbal, I was given a 100 line verbal reading passage, and was only asked 3 questions
(1) I thought this was way above my level and
(2) i've never seen any passage that long before/ thought it was a bit overwhelming for someone who only scored a 36 Raw on the verbal.

[StaceyKoprince]

First, although you only see a score range of 700-800, the questions are actually rated on a 10-point scale, so you are getting a fluctuation in difficulty level even if you can't tell yourself.

Second, your score corresponds NOT to an average performance over the whole section but to your current performance at the point that the test ends. You could have tons of high level questions at the beginning and in the middle, and then get all of the rest wrong (or maybe run out of time) and your score would drop - potentially quite a bit, depending upon how many in a row you get wrong or don't answer at all.

Finally, passages are not assigned a difficulty level. Passages have a range of easy to difficult questions - the test chooses which questions to give you based on your scoring level, not which passage to give you.

Having said that, 100 lines is a little long (that probably corresponds to about 70-75 lines on the real test (the screen sizes are different) and the real test usually doesn't go much beyond about 60. I know we've got one or two passages floating around that are too long - we're working on revising them.

[Also Confused-Score.Calc]

Thanks, Really appreciate the quick response.

So i'm interpreting your response as - if I am facing a 700-800 level question, at that point in time my score is 700-800. In other words if I faced a 700-800 question, on the last problem on the section, (and get it correct) I should expect to get a 700-800 for that section. Looking back, my score makes more sense. Since i had never faced 20 - 700-800 level questions in one section my pace was off and rushed the last several questions. My last 700-800 level question was on problem #31 so if (above is correct) i probably wasn't that far off from a 700 level performance?

I'm taking a couple things away from this (if i understand your post)
(1) the fact that I was given 20 700-800 level questions (as opposed to 10 700-800 level questions when I got a raw of 45) is a sign of improvement (at least from the sense that I was able to 'earn' more difficult questions) and I should just work on pace a little bit to close out the last couple questions better?

(2) on the MGMAT online question banks, there aren't a ton of 700-800 level questions - I know you can get more challenging Quant questions if you pay for the Challenge questions, but it would be nice if you guys could throw in more tough questions in the regular package - as i said it was the first time i faced so many tough questions in one set so I was probably bound to get most of them wrong .


PS if all of what i said is remotely accurate, I'm blaming my poor performance on Kaplan who pretty much tells you to focus a disproporionate amount of your time on the first couple dozen questions - which i do and as such, i get those questions correct and the later questions wrong.

[StaceyKoprince]

Essentially, yes - though, of course, the algorithm has to factor other things in (like question type, content area, etc) - so it may want to give you a 690 DS geometry triangles question, but there may not be a 690 question of that type available, so it may have to give you a 710 instead. Things like that.

And, yes, your score was above 700 at some point since you were getting so many questions at that level. But then it dropped before you finished the section. That's why it's so important to keep going - don't get sucked into anything and blow your timing! No matter how good you get, the test will just keep giving you even harder questions. At some point, you have to make a choice: let go on this ridiculous question (that you're probably going to get wrong anyway) so that you can address the last 5 or so questions on the test, or don't let go and run out of time and have your score drop. The best option there should be obvious. :) (retty much everybody has to guess on something like 5-7 questions on the test. Your only choice is whether you want to have some control over when you're guessing.)

I agree that it would be nice for you, as a 700+ hopeful, to have more high level questions in the question banks. We do need to make sure we accomodate everyone, though, and not everyone is trying to hit that score level. It's a tough balance when you have a static pool of questions. (And the challenge questions wouldn't be a good solution anyway - those are really "900" level questions - that is, way harder than things you'd see on the real test. I tell students in my classes, who have access to those for free, not to bother with them.)

Also, the myth that the earlier questions are more important has been floating around since this test started (in fact, I worked for a different prep company then, and I was there when the myth started - everyone misinterpreted a theoretical study about CAT algorithms). But it is just that - a myth.

[still confused]

Thanks, but that doesn't jive with a Verbal Raw of 37 (83%) - the last 11 questions are outlined below

Question#31 500-600 Correct, 32. 700-800 (C), 33. 700-800(C), 34. 500-600 (C) 35. 300-500 (C) 36. 700-800 (C) 37. 700-800 (W) 38. 600-700 (C) 39. 600-700 (W) 40. 700-800 (C). 41. 600-700 (C)

700-800 level questions I got 4 out of 5 correct
600-700 level questions I got 2 out of 3 correct
500-600 level questions I got 2 out of 2 correct
300-500 level questions I got 1 out of 1 correct

given the earlier explanation (the question you get reflects your score at that point in time in the exam), i just don't get how this score is still in the mid 600 range?

thanks

[StaceyKoprince]

It doesn't technically reflect exactly your score at the end - that's just the easiest way to think about what is a very complex algorithm. It's more closely correlated with "score at the end" than it is with "average of performance throughout" (which we're more used to from paper tests) - but it's still not exactly what's going on.

I'm guessing that some of the questions in the 33 to 35 range were RC? (One passage usually has only 1-2 700-800 level level questions, so some Qs you're offered will be lower level even if you're getting harder questions right.)

I think you might also be using the percentile levels normally associated with the overall score (mid-600) interchangeably with the percentile levels associated with verbal only - which is easy to assume, since we assign these verbal questions "ratings" according to the overall score, even though they're just verbal questions. That's a flaw on our part, in terms of how we present the information to you - but they're not (necessarily) interchangeable.

If you'd like to have our algorithm guru take a look and try to explain in more detail, send an email to studentservices@manhattangmat.com, explain what you explain above, and ask whether someone could answer your question. They'll forward to the appropriate person.