Decoding the GMAT Scoring Algorithm – A Cheatsheet
Learn smart ways to exploit the GMAT Scoring Algorithm and optimize your preparation!
A lot of our students have been asking us:
“How does the GMAT algorithm work? What do people mean when they say Q49 V36?”
“With the same scores in each section, why are the overall scores for my friend and I, different?”
“If the scores are ranked out of 51, how does that account for a total score of 800?”
Don’t get confused. We will make it really simple for you in this article.
We will give you an inside peek into the working of the GMAT scoring algorithm.
And to help you get a holistic perspective, we will also explain the basic functioning of a Computer Adaptive Test (CAT) – the underlying logic behind how the GMAT works.
“How will understanding the GMAT algorithm help you?”, you ask?
It will definitely not magically improve your scores, but if you know how the algorithm functions, there are ways you can leverage the system to your advantage – we will talk about this later on.
We will be dividing the article into 6 sections:
A. GMAT Test Structure
The GMAT test is divided into 4 sections.
NOTE: We have updated this blog based on the announcement by GMAC on some major changes in the GMAT test timing and the number of questions you’re going to be having in both Quant and Verbal.The new GMAT exam will be shorter by 30 minutes from April 16th, 2018.
Now, only two sections out of the four count towards your total GMAT score – both IR and AWA don’t count (doesn’t mean you don’t prepare for it – just that it is not needed for the 3-digit score out of 800).
Which leaves us with the Quant and the Verbal sections of the GMAT – they makeup for your final GMAT score out of 800.
The following are the total computed scores for the Quant and Verbal sections:
-> Quant (31 questions) – raw score out of 51.
-> Verbal (36 questions) – raw score out of 51
These questions are designed by trained psychometricians (yes – such a profession exists) who love making things terribly complicated. What else could explain the rationale behind picking random numbers like 31, 36 and 51? 🙂
We can’t change the way the GMAT Algorithm works, but we sure can help you get an understanding of how the system functions.
Don’t get overwhelmed by the numbers you will be seeing further down in the article. We are going to peel the whole onion – layer by layer.
B. GMAT Adaptive Testing
How is the GMAT Adaptive Algorithm different from the usual tests?
The GMAT algorithm is terribly precise and does not rely on the usual linear score that we are used to.
In the linear scale, the score is computed at the END of the test by taking the number of right responses and the number of incorrect responses. So in a school exam if you answer 6 out of 10 questions correctly – you get 60%.
In the adaptive scale, the score is getting computed after EVERY question on the test. The algorithm is constantly checking where to “keep” you. Get a question correct, the algorithm will “reward” you by giving you a higher score. Get a question wrong, the same algorithm will “penalize” by dropping your score.
So the GMAT is constantly trying to test your Quant and Verbal ability, that too within a short span of time and with a limited number of questions.
To accurately assess a person’s ability from 200 to 800 – means you have 61 possible scores on the GMAT. All of this from just 58 questions (why 58 and not 78 you ask? We’ll tell you why later on in the article).
The GMAT does this incredible thing because of the adaptive nature of the test.
> Understanding the Adaptive Algorithm using an example
Let’s understand the GMAT algorithm.
First GMAT will ask you an average difficulty level question.
Why average? Because we need to begin someplace and starting at the middle is the most optimized strategy.
Based on the accuracy of your response, GMAT will either reward you by bumping you up to a higher level, keep you at the same level, or demote you to a lower level.
If you get the right answer, the next question we put forth may be of the same difficulty level, or it may increase.
If you get the answer wrong, we will ask you another question of the same level, or an easier one.
By asking such a series of questions, GMAT figures out the range or the band at which you are at. Then GMAT will ask further question to understand the specific score within that range.
In short, the GMAT exam adapts to your performance on every question.
It selects each question based on your answer to the previous one, and how you have been doing so far.
Let’s take an example.
If you look at the chart, we have 3 test-takers.
Let’s assume that all three of them have started with an average difficulty level question.
Let us look at their performance on the 1st question:
T1 answered right, and hence progressed from the score range of 500 to 600.
T2 answered wrong, and dropped from 500 to 400.
T3 also answered wrong and dropped from 500 to 400.
Moving on to the 2nd question.
T1 answered right, again – he jumped to the 700 score range.
T2 answered right, and made a small jump to the 450 range.
T3 however, answered wrong, again, and dropped further down to 300.
As you see on the graph, the trend lines are high or low depending upon the answer to the previous question.
Which also incidentally, is how the adaptive algorithm works.
So if you are on your 25th question on the GMAT, the algorithm has data about your performance on the previous 24 questions. This means the algorithm is intelligent enough to get a good sense of where you are currently and will calculate your difficulty level to provide questions accordingly.
> Analysis with the Verbal section
With the understanding of the adaptive scoring engine, it’s be easier to break the example down further – showing you what my actually happen on the actual GMAT.
We will divide the GMAT Verbal scores into 4 buckets for your benefit – though the actual GMAT will have a finer calibration.
We’re going to use two hypothetical candidates – Amit and Rupa.
The GMAT scoring rules are as follows:
1. If you get < 40% of the questions right, you get demoted to a lower bucket.
2. If you get 40% to 60% of the questions right, you remain in the same bucket.
3. If you > 60% of the questions right, you get promoted to the next bucket.
Now let us say they take the first set of 6 questions, Amit gets 3 of them correct, while Rupa gets 4 of them correct. So, Amit stays in the same level while Rupa moves up.
Throughout the test, Amit consistently performs in the 50% range while Rupa does 60% or better.
Have a look at this table:
Amit’s performance being mediocre through the test, did not progress with his scores and stayed within the 25-30 bucket.
Rupa’s on the other hand, showed progress at every stage until she crossed the 40 bracket.
Now if you look at the number of questions they answered correctly, you will see Amit’s got 18 questions right and Rupa got 22 questions right.
4 questions may not seem like a big deal, right?
If you look at the last 12 questions, Amit and Rupa got 6 answers right.
But here’s how the GMAT algorithm expects you to maintain your accuracy.
Rupa had to maintain accuracy on a much higher level.
Amit on the other hand, had to maintain the accuracy on a much lower level.
Why do you think that is?
Weighted averages – where there is a “weight” attached to every question.
So not all questions are the same – GMAT recognizes the effort for maintaining accuracy at a higher level (say a GMAT 700+) is far greater than maintaining accuracy at a lower level (say at GMAT 500+).
Now let’s move on to explain the working of the GMAT Scoring and Percentile System.
C. GMAT Scoring and Percentile system
Let’s tackle this one by one.
Now that you know how both – the Quant and the Verbal scores translate to a total of 51, let’s find out the bare minimum number of mistakes you’re allowed to make to get a high score.
How many mistakes can you afford to make on the GMAT?
It’s really hard to not make mistakes on the GMAT. So let’s say – it’s safer to make mistakes in intervals rather than continuously.
So in the first 15 questions, instead of getting question number 4,5,6,7 wrong – you get questions 4,8,12,15 wrong.
Making mistakes in a continuous string reduces your accuracy drastically.
Here’s a secret – You can actually make a few mistakes (say 2 in Quant and 1 in Verbal) and score a perfect 800!
Notice that even a solid score like a 710 – with a split of Q49 V38 – means you would have made over 20 mistakes! All the more reason not to fear intelligent guessing on the GMAT.
Moving on to the percentile system.
What do percentiles mean?
The percentile rank of a score is the percentage of scores in its distribution that are equal to or lower than it.
A test score that is greater than 75% of the scores of people taking the test is said to be at the 75th percentile, where 75 is the percentile rank.
For example, in the GMAT, if you’re in the 90th percentile, you’ve scored more than 90% of the people taking the test with you.
Let’s now have a look at the percentile charts for Quant and Verbal.
Notice a few “weird” things over here:
If you get a 45 out of 51 in verbal or above, then you’ll still land in the 99th percentile. That same score of 45 out of 51 in Quantitative is considered 57th percentile. This is because a lot more students are scoring a 51 in Quant than in Verbal.
However on the flip side, with the Quant percentile starting at 96%, every mistake you make drastically drops your score. For the Verbal spread of percentiles it is not that vast; this means the scope for improvement in Verbal will not be that steep.
So that is all that you need to know about the scoring system while preparing for the GMAT.
Let us now deep dive into how GMAT takes your sectional raw scores and computes the final GMAT score.
D. How is the total GMAT score calculated?
Let us look at a few things:
The total GMAT scores range from 200 to 800.
The Verbal and the Quantitative scores range from 6 to 51.
Both the Verbal and the Quant combine to give you the total GMAT score.
Here is a chart that you can use to check your final GMAT score, if you know your raw scores in Quant and Verbal.
So for you to score a 700 you need to have a total raw score total of 86. There are a few ways you can score this:
Strong in Quant
Q50 V36 (percentile)
Strong in Verbal
Q44 V42 (percentile)
Equally strong in both Quant and Verbal
When you start your GMAT prep, you might realize which one of the two areas you are better at.
And no, being an engineer doesn’t automatically endow you with superior quant skills 🙂
Here, is the GMAT final score Percentile Ranking Chart:
Note this: even if you don’t score in the 99th percentile on individual sections, you can score in the 99th percentile on the GMAT.
Let’s see how.
Say, you get an 86th percentile in Quant (Q 50), and a 96th percentile in V (V42) – you will still get a 99th percentile overall (760). This is because there are fewer people scoring such high percentiles in BOTH the sections.
These charts are updated every year and you can head over to the GMAC website to know more:
We just helped you understand how the scoring chart works – hopefully now you don’t really need to break you head over it 🙂
E. Commonly Asked Questions about GMAT Scoring
a) Does GMAT take time into consideration while calculating the score?
Remember, we are not calculating the time you take to answer each question, it does not have an impact on your final score.
b) If I am seeing easier questions, does it mean I am not doing well?
No! Firstly, you don’t really know if it is an easier or harder question (GMAT can make very tough problems deceptively simple). Secondly, it could be an experimental question, which means it is not based on your performance.
c) Does GMAT take the position of the mistakes into consideration?
If you get questions wrong one after the other, you are in greater risk than if you distribute your mistakes over a range.
For example, let’s say, from questions 21 to 30 there are 2 candidates X and Y and their frequency of mistakes is:
X marks the wrong answers for questions 22, 26 and 29, while Y marks the wrong ones for 23,24,25.
Then Y would be penalized heavier than X.
d) Does it mean that the first 10 questions important on the GMAT?
There are 2 fallacies here:
a. Spending more time will improve your accuracy. If you do spend more time in the first 10 questions, (and not improving your accuracy much), then you are actually robbing other questions of the time they rightfully deserve.
b. You control whether you get the question correct. Actually, if someone is smart enough to get first 10 right – isn’t he/she smart enough to get the rest of the questions also correct?
e) If I keep getting questions correct, will GMAT start giving me impossibly tough questions?
Yes, the level of difficulty does increase, but at a more controlled level. Also remember, that the questions are not “easy” or “hard” by themselves but they are “easy” or “hard” for the test taker at a given level.
f) Can you skip questions on the test and come back to them later?
Skipping questions on ṭhe GMAT is not an option.
If that option were to exist, it would go against the concept of the adaptive testing method.
g) What if I do not have ṭhe time answer all 37 questions on quant and 41 questions in verbal?
If you run out of time towards the end of the test, your overall score reduces.
The GMAT marks all unanswered questions as wrong, thus reducing your overall score.
h) Are the questions within an RC, computer adaptive?
No, the questions within the RC is not computer adaptive. The difficulty level of the question is pre-determined based on your performance up until that point.
i) Do continuous errors adversely affect your score?
Yes. The GMAT being a computer adaptive test, a string of errors will reduce your overall score.
j) Is the first question all test takers get of the same difficulty level?
No. The first question for all test takers are not of the same difficulty level. Since the GMAT is an adaptive test, you can get a question of a random difficulty level.
k) Does your performance in the quant section affect the difficulty levels of the question in the verbal section?
No – it does not.
l) Why are the percentiles ranks different for verbal as compared to quant?
More people are scoring higher in Quant than a few years ago, and fewer people are scoring as high in Verbal now than a few years ago.
With more people taking the GMAT from India, China and Asian countries the average Quant scores are going up and Verbal scores are going down.
You can read more on that here : https://gmat.crackverbal.com/gmat-verbal-new-percentile/
m) Do the AWA and IR sections also contribute to the overall score?
No, the AWA and IR sections do not contribute towards your final GMAT score.
n) Why is the GMAT score on a scale of 200 – 800?
This is a question a lot of people have been asking. Unfortunately, no one has an answer other than GMAT.
o) How do I know how many questions are experimental and which questions are experimental?
Refer to the section above on experimental question, we’ve provided a table with an explanation. And as for knowing which questions are experimental, there is no way of knowing.
p) How do I know which difficulty level a question falls under on the GMAT?
Again, there is no way of knowing the difficulty level of a question.
We hope this article helped you understand the GMAT algorithm.
One important piece of advice before we wrap this article up, understanding the GMAT algorithm will certainly make you aware of your scoring pattern, and the existence of experimental questions – but then again – it won’t help you beat the GMAT.
You can also watch CrackVerbal’s Founder and GMAT Expert – Mr. Arun Jagannathan explain the algorithm in the video below:
If you need any help with your GMAT prep, you can sign up for our online demo session :