Is this true? (GMAT Scoring)

[farishta]

Then there is the scoring system. Not all questions count equally. This is because the jumps in score are more extreme at the beginning of each section. In fact, two test takers can get the same number of questions right but have very different scores. Here's how it works:

Total scores on the CAT range from 200 to 800. The first question in each section is worth 80 to 100 points. You start each section with a score of 500, so if you get that first question right, your score goes up to about 600; if you get it wrong, your score goes down to about 400. The second question is worth 60 to 70 points. This means that if you get the first two questions on a section right, you're looking at a score of about 670, but if you get the first two wrong, you've got a paltry 330. Here's the odd part: If you get the first question right and the second wrong, you've got a 530, but if you get the first one wrong and the second one right, you've got a 470. The relative weighting of questions affects your score that quickly.

Of course, the scoring gets more complex as you answer more questions, but in general, after the first five questions of a section, you're just fine-tuning your score. If you answer the first five questions correctly, you'll be working at a level well above 700. If you get the first five wrong, you're well below 300 and will likely get easy questions for the rest of the exam -- the kiss of death on the CAT. You can recover from answering a few early questions wrong, but you'll need to get a long string of questions right, as later questions are worth less.

[aim-wsc]

did you copy-paste it from somewhere?

yes! what you've written here is correct. but remember It's just a hypothetical model of some prep company ... It's good to make you understand the structure n scoring algorithm of GMAT.

however you should not forget that 1/4th of GMAT items are experimental questions that are randomly placed, which dont affect the algorithm.

[Stacey Koprince]

This is only a partial picture of the way that the test works - it is incredibly more complicated than this, and the numbers posted here are not accurate.

The important things to know are:

It is a MYTH that the early questions are "worth more" (despite the fact that there are wider swings in difficulty level at the beginning) and it is also a MYTH that you can get the first 5 (or 7 or 10) questions right and then nothing else you do will bring your score down.

It would take me at least an hour to explain in depth how the algorithm works. If you're interested, you can research IRT (Item Response Theory) which is the main theory behind the algorithm. But here's an example of how it does NOT work.

A couple of years ago, I answered the first 2/3 of the math section on an official test to the absolute best of my ability (which has been documented many times at 99th percentile). I then left the remaining 1/3 of the questions blank. My final math percentile was in the low 50s. It is reasonable to assume I was in the 99th percentile before I stopped answering questions b/c that is how I normally perform on the test.

This is an extreme example, of course - few people answer only 2/3 of the questions - but I did it to demonstrate that you can be "brought back down" later in the test. You cannot "set" your score at a high level by answering some beginning number of questions correctly - I answered 2/3 of the questions correctly or mostly correctly and my score still tanked when I skipped the last third.

[mojomover]

Hi Stacey,
While I agree with your assessment that the earlier questions are not necessarily "worth more", I dont know if I agree with your "experiment".

I think the point in question is not if you left the last 1/3 blank- I think GMAC themselves say that it is better to guess than to leave a question blank, so the right test might be to guess the last 1/3 or intentionally answer the last 1/3 wrong. My hypothesis is that uanswered questions are penalized more than wrong questions.

Another way of stating the problem is- "If you are going to make say 5 mistakes in the test (based on your practice tests), then it is better to make them in the later portions of the test than the first 5.

This is only a partial picture of the way that the test works - it is incredibly more complicated than this, and the numbers posted here are not accurate.

The important things to know are:

It is a MYTH that the early questions are "worth more" (despite the fact that there are wider swings in difficulty level at the beginning) and it is also a MYTH that you can get the first 5 (or 7 or 10) questions right and then nothing else you do will bring your score down.

It would take me at least an hour to explain in depth how the algorithm works. If you're interested, you can research IRT (Item Response Theory) which is the main theory behind the algorithm. But here's an example of how it does NOT work.

A couple of years ago, I answered the first 2/3 of the math section on an official test to the absolute best of my ability (which has been documented many times at 99th percentile). I then left the remaining 1/3 of the questions blank. My final math percentile was in the low 50s. It is reasonable to assume I was in the 99th percentile before I stopped answering questions b/c that is how I normally perform on the test.

This is an extreme example, of course - few people answer only 2/3 of the questions - but I did it to demonstrate that you can be "brought back down" later in the test. You cannot "set" your score at a high level by answering some beginning number of questions correctly - I answered 2/3 of the questions correctly or mostly correctly and my score still tanked when I skipped the last third.

[Stacey Koprince]

It is better to guess simply because that gives you a 20% chance of guessing right. But you are not actually penalized more for leaving it blank than for getting it wrong - the test treats any unanswered questions as though you had answered it, but answered it incorrectly.

If you are going to make 5 mistakes on the test, it is better to have those 5 mistakes spread out rather than clustered anywhere, whether at the beginning or at the end.

If, however, you had to have them clustered at the beginning or at the end, you would want them clustered at the beginning, NOT at the end. If they're clustered at the beginning, you still have the opportunity to recover your score (and you can theoretically - it often doesn't happen in reality simply because if you make mistakes on the first 5 in a row, chances are something else is going on, either anxiety or just a low-scoring person, and that person is unlikely to recover). If they're clustered at the end, you cannot recover at all because the test is over.

You will get things wrong on the test and you will get things wrong that you knew how to do. You don't want to get a lot wrong in a row and you especially don't want that to happen at the end of the test (again, because you have no opportunity to recover).

[aim-wsc]

If you are going to make 5 mistakes on the test, it is better to have those 5 mistakes spread out rather than clustered anywhere,

I ENTIRELY AGREE 8)

[rkumar]

I want to test this and i took a long time for my first 10 ques both in verbal and quant. I ran out of time with my last 12 questions. which i guessed randomly. even though i got the first ten questions in both of them right. My score was 100 points below the normal score i was getting. So it is not true that getting the first 10 right puts you in a higher score range