Scoring Algorithm and Weight of Verbal Score

[MSJ]

I found a few posts related to the MGMAT scoring algorithm and how it differs from other commercial CATs. I'm still a little confused. For instance, compare these two CAT results of mine:

Kaplan CAT
Quant: Score of 29 (18 incorrect)
Verbal: Score of 36 (13 incorrect)
Total: 550

Manhattan CAT
Quant: Score of 42 (18 incorrect) 63rd percentile
Verbal: Score of 42 (15 incorrect) 93rd percentile
Total: 700

How could the two scores vary so much? Is it because MGMAT weights harder questions more and I just happened to get a lot of the harder questions right on the MGMAT?

Also, is it known if the real GMAT always gives you a harder question after answering the previous one correctly? Based on the MGMAT questions, it looks like this is not always the case (eg after answering a 600-700 difficulty question correctly, I was given a 500-600 level question).

Finally, this article surprised me. This was the first I heard of the verbal score actually contributing more to your final score than the quant score: http://www.vault.com/nr/newsmain.jsp?nr ... at_id=3071
Is this actually true for the real GMAT? If so, perhaps this explains the discrepancy between my Kaplan and MGMAT scores if Kaplan weighs the two scores equally.

Thanks for your help!

[StaceyKoprince]

I can answer most of your questions with what we know via lots of research, but most of that info is not released by GMAC.

It does appear to be the case that the verbal score factors more heavily into your overall score. GMAC has not officially confirmed this but this can be seen if you examine a lot of records - and we've done this with our students' records.

The correct/incorrect numbers you report have very little to do with the score calculation - either on our test or on the official test. In other words, don't use those numbers to assess anything and don't look for much correlation there between higher and lower scores and higher and lower #s of problems right.

Most people will get roughly half the questions right and roughly half wrong, though this starts to skew a bit at very high and low levels, as you can see from your 93rd percentile verbal score. You also see, though, that you still get a large number of questions wrong regardless of your scoring level. The single largest factor contributing to your score is the difficulty level of the questions you're answering. People at the 25th percentile can get just as many questions right as people at the 75th percentile.

The algorithm does not SOLELY depend upon difficulty level as the determining factor for the next score. The algorithm also has to give you a certain mix of different question types (eg, DS and PS), different content areas (eg, geometry vs. algebra) and even different content sub-areas (eg, circles vs. triangles). And, on the real test, the algorithm will also give you about 10 experimental questions, which are not tied to your performance at all. As such, it is not necessarily the case that the difficulty level will strictly go up or down depending upon whether you got the previous question right or wrong, respectively.

Also, although your assessment reports only show you 100-point ranges for those questions, the difficulty ratings are actually much more granular than that. So the 600-700 level question you had might have been rated a 610 and the 500-600 level question might have been rated a 590. Those are within one standard deviation of each other, so statistically speaking, we can't say that the 590 is easier than the 610. Essentially, after you got that 600-700 level question right, the algorithm looked for another question within a range of difficulty levels around where you were currently scoring that also met some of the other parameters it needed to hit (in terms of question type and content area) - and that 500-600 level question came up.

I'm probably getting into more detail than you want to know. Suffice to say the algorith is VERY complex. :)

Oh, and I can't actually say anything about Kaplan's algorithm because I have no idea how they make it! I only know ours and the official one.

[odonnelle22]

Hi, if you are already in the 99th percentile in verbal but are still getting a couple of questions wrong per practice exam, can you still increase your overall score as greatly by getting those last few questions right? or is it once you're in the 99th percentile getting another question (or 2 or three) right doesn't have as much effect on overall score? since verbal seems to carry so much weight...

thanks.

[StaceyKoprince]

Probably not going to change much, no. At that point, if you're still looking to increase your overall score, then you need to be looking more toward the quant... :)