Verbal Accuracy Metrics

achilles · Posted: Wed May 23, 2007 5:54 am Post subject: Verbal Accuracy Metrics

Needed some stats on general accuracy level in verbal, from experts & ppl who have taken umpteen CAT's from various sources.

Say to score a gud 45+ out of 60, how many question can i afford to get wrong out of 41, assuming 50% of the wrong answers WEIGHTED very-difficult and rest difficult. Similarly stats for 43+ score etc.

This, or any sort of metric for verbal accuracy will be really helpful, especially frm ppl having given GmatPrep Tests, so respond plz.

Stacey Koprince · Posted: Mon May 28, 2007 11:26 pm Post subject:

I don't have specific stats for you, but scoring up to 90th percentile, you should expect to answer 50-60% of the questions correctly. Scoring between 90 and 99 percentile, you should expect to answer 60-85% of questions correctly. The range is wide because the algorithm is tied very much to the difficulty level of the question, rather than the strict percentage correct.

Short answer: no matter how well you score, expect to get a lot of questions wrong.

achilles · Posted: Sun Jun 03, 2007 3:20 am Post subject: Verbal Accuracy Metrics

Thanks Stacy, that really helped.
However as I have noticed in some CAT's that I have given from Kaplan, Princeton, scoring is much more frugal than stated by you. For Eg. on an avg 10-13 mistakes (70-75% correct) in verbal test fetches me only 32-35 score, much less than 90 %ile threshold. Sad

. So just wanted to know are these practice CAT's more stricter than the actual exam.

Stacey Koprince · Posted: Sun Jun 17, 2007 11:16 pm Post subject:

I don't know how other companies have constructed their algorithms, so I can't comment on that.

On the real thing (and on our test), different people could all get, say, 60% right and yet score anywhere from 30th percentile to 80th percentile (for example), depending upon the difficulty levels of the questions they are answering.

Oh - and on practice tests, there is one big additional variable that's not on the real test: practice tests count all of the questions you answer, while the experimental questions on the real test are not counted toward your score. So that adds some complexity to our attempts to mimic the real algorithm (though our company, at least, is also working on building experimentals into our practice test).