Thank you for using the timer - this advanced tool can estimate your performance and suggest more practice questions. We have subscribed you to Daily Prep Questions via email.

Customized for You

we will pick new questions that match your level based on your Timer History

Track Your Progress

every week, we’ll send you an estimated GMAT score based on your performance

Practice Pays

we will pick new questions that match your level based on your Timer History

Not interested in getting valuable practice questions and articles delivered to your email? No problem, unsubscribe here.

Thank you for using the timer!
We noticed you are actually not timing your practice. Click the START button first next time you use the timer.
There are many benefits to timing your practice, including:

Question about how they score the GMAT [#permalink]
04 Apr 2007, 11:00

OK, so this may sound like a silly question, but I'm a bit confused. I just took my first GMATprep and scored a 600. Now, don't get me wrong, I am thrilled with this score after a month of studying and my first time taking a timed CAT version. My question is, how is this possible when I got 18 questions wrong in the Q section and 10 wrong in the V section?

Yes, I understand that they take off fewer points if you get a hard question wrong than if you get an easy one wrong, so does this mean that (especially in the Q section), I was just getting all the super hard ones wrong? I guess I'm just questioning this because 600 seems to be a high score considering I got almost 1/2 of all the Q questions wrong!

I wouldn't worry about the number of questions you get wrong. No one knows how the algorithm works (outside of Pearson / ETS etc), but its definitely possible to get high scores with multiple mistakes. 49Q is still possible with over 10 wrong questions. You also have to bear in mind some of the questions are experimental.....

As I understand it, the CAT doesn't assign points based on the difficulty of a question. Instead, the CAT will select questions as you go along, based on how you are doing so far. The traditional view is that the first question in a section will be a medium difficulty question (call it 500 level). If you get that right, you might get a really tough question (call it 700+ level). If you get that wrong, you might get something in between, say a 600 level.

By midway through the test, the theory is that the CAT will have ascertained the general level that you are testing at, and will then just ask question to fine tune the final score. Say you do well early in the test and are at a 700-750 range. It will then throw a series of questions in this range at you. If you get 1/2 right and 1/2 wrong, it will know that you should be scored in that range. If you get a bunch right or wrong, your score could get shifted to a higher or lower range. Conceivably, if you do well early on, you could miss 1/2 the questions in the last 2/3 of each section and still get into the 600 range.

I think this is why many people have the believe that the final few questions can't really do that much to swing your score. Even if you get the last 3 right or wrong, your score will probably only move 10-20 points because your score range has already been identified. This is just a theory so I wouldn't advise any strategies that ignore the last part of each section.

the algorithms for scoring GMAT are relying on a statistical theory called "item response theory".

every question in the GMAT database is associated with few parameters the describes its level of difficulty and other statistical characteristics.

throughout the test the computer evaluate (and re-evalute) your score based on your answers. it is easiest to think of it as if the computer calculate a "confidence interval" for your score. so at the beginning the onfidence interval might be 200-800 (or slightly less on each side).
each answer provide's the computer data (based on the answer and the questions' statistical parameters) to adjust this confidence interval, and the computer tries to choose questions that will gve it maxinmum information possible. so if your cuurent scre interval is 500-600 the computer will not give you a question targeted at 700 scorers...

each question, beside moving this confidence interval up or down, will also make the range itself smaller, until at certain point it will get to a single number.

the number of question required to achieve that may be different between people, and depends on their answering patterns.

for example, suppose you got two questions, one at a level of 600 and one at a level of 700. if you got the first right and the second wrong then the computer, with more "confidence" assign you a score in the range 600-700. but if it was vice versa (you got the 700 question right and the other wrong) - then the computer might give you a higher/lower score (dependig on other parameters) but with less confidence, and require more questions to get the required level of confidence.

this also explains why you can get some wrong questions without you score being penalized - those question were used mainly to enhance the level of confidence (sometimes measured by confidence intervals or standard deviation or other statistics), rather than affecting directly on your score.

the problem with all the details of the theory is that they DONT HELP YOU PREPARE for the test, or make a better strategy.

the guidelines that you should try harder to get the first questions correct are true: any "learning theory" will say that most learning and adjustments are made based on the initial information, while later information is used mainly for reinforcement. this is true for item-response-theory and GMAT as well.

but beyond that, i find that there are almost no "tricks" to fool the algorithms...