Computer Adaptive Tests and Scoring
Believe it or not, the GMAT is a lot like a first date!
Suppose you are looking to find a long-term partner and are going on a first date with someone you've never met. What is going to be most important about your date in order to determine whether a second date is a good idea?
It will become clear why GMAT scoring is initially like a first date as we proceed. In the meantime, let's also break some myths. You have probably heard an array of theories as to what creates your GMAT score.
Which of the following determines your GMAT score?
This is also part of the reason why scores can vary greatly from one exam to another, even if you hit the same percentage of correct and incorrect. You can get a 20 in the Quantitative section, yet have gotten more answers right than someone who got a 35.
This can be demonstrated via the following graph. Notice that already after the first few questions, a similar performance can start to yield a different score:
No, this does not determine your score. In fact going too fast may well be counterproductive as you will simply make careless errors.
Well, close, but not quite.
Indeed you want to get questions right! But that is a small part of the story.
Well, you may as well simply order the check, pay your half, apologize to your date, and simply go home!
Not quite.
As you can see, a 'bad' first impression will drop the score. While it is possible to climb back up, you will need to get a very high percentage of the easier questions correct, and it is possible your score would be at a lower level than had you performed strongly on the first bunch (the first five or so).
So back to that first date analogy: A first impression is important!! The only difference is: You do not want another date with the GMAT no matter how much fun you will have this time around!
If you begin to make several errors in a row, for example, this too will hurt your score as the algorithm will begin to drop the level you were previously at. Remember your level changes as you get questions right or wrong.
Therefore, making several mistakes in a row anywhere on the exam can be detrimental to your score.
This is something you should definitely NOT do!
Leaving questions unanswered at the end carries the highest penalty. You must finish the adaptive sections.
Having the test end with a few unanswered questions at the end is worse than guessing those questions.
If you leave, say, 3 blank questions in the end, the algorithm penalizes you even **more** than if you had guessed those questions incorrectly.Cruel, isn't it?
In such an adaptive exam, the difficulty level of your question will be based on how you answered the previous ones.
Theoretically, you begin with average level questions (and score). As you get questions right, the algorithm reevaluates your level each time and sets it to a higher level, whereas if you get them wrong, it sets the score at a lower level. The GMAT algorithm will attempt to identify your true level. The more you prove that you can tackle high level questions, and the earlier you do so, the better your score will be.
No! They are important to get you to the higher levels as early as possible but they are NOT the most important questions as other factors affect your score as well.
Remember, you still have the rest of the exam to complete. You want to start off well, but you also certainly do not want to ruin all your good work. You cannot afford to sacrifice a bunch of later questions for the sake of taking (too much) time on earlier questions.
Let's look at another couple of factors that are important to bear in mind:
Guessing is better only insofar as you want to finish the exam, but you should try to avoid guessing, otherwise you risk making mistakes that would allow the algorithm to set your score at a lower level.
Later, if necessary, we will help you with proper time management so that you can maximize your score by learning to avoid guessing and figuring out which questions you should go after or avoid.
Now is the time to acquire knowledge and techniques.
Probably not. While having fun is important, you are looking for a long-term partner. Suppose you definitely want to have kids, but you find out your date doesn't; it's probably not worth going on a second date, regardless of the fun you had.
The test makers for the GMAT are a lot like you on a first date. They are trying to assess your true knowledge level. While it might seem strange, they are trying to estimate the difficulty level of a question where you have a 60% chance of getting the answer correct. That estimate at the end of the section corresponds to the score you get.
Not quite. While a good first impression can help, it isn't the most important thing. What you learn about your date throughout the whole date should inform your decision about whether to seek a second date.
The test makers for the GMAT are a lot like you on a first date. They are trying to assess your true knowledge level. While it might seem strange, they are trying to estimate the difficulty level of a question where you have a 60% chance of getting the answer correct. That estimate at the end of the section corresponds to the score you get.
Exactly! Since you are looking for a long-term partner, your date's true personality is going to be key for figuring out if a second date is a good idea.
The test makers for the GMAT are a lot like you on a first date. They are trying to assess your true knowledge level. While it might seem strange, they are trying to estimate the difficulty level of a question where you have a 60% chance of getting the answer correct. That estimate at the end of the section corresponds to the score you get.
Now think like a test maker: what would be the best and fastest way to use questions of different difficulties to figure out a test taker's knowledge as quickly as possible?
Not quite. It would be weird to start a date by asking if your date enjoys breathing. It doesn't provide any useful information because everyone would give the same answer.
Precisely!
Test makers want to avoid wasting time by giving questions that are too hard or too easy. Since most students will have a knowledge level that's close to the average question difficulty, it makes most sense to start there.
The first few questions on each section will be close to average difficulty. From there, the difficulty of each question will be determined by how you will have done on the previous questions.
Since the GMAT algorithm is trying to estimate the difficulty level of questions you will be able to correctly answer 60% of the time, what percentage of questions do you think you need to get right in order for the difficulty of questions to increase?
Exactly.
If you get more than 60% correct at a given difficulty level, the algorithm will start giving you harder questions to try and find your true knowledge level. The more questions you get correct, or incorrect, the faster the change in difficulty will be.
Not quite. If you get about 60% correct, the algorithm thinks it is close to the right difficulty for you, so it won't change the question difficulty much at all.
If you get more than 60% correct at a given difficulty level, the algorithm will start giving you harder questions to try and find your true knowledge level. The more questions you get correct, or incorrect, the faster the change in difficulty will be.
Not so. If you are getting less than 60% correct, the algorithm will give you easier questions to try and find a difficulty level closer to your knowledge level.
If you get more than 60% correct at a given difficulty level, the algorithm will start giving you harder questions to try and find your true knowledge level. The more questions you get correct, or incorrect, the faster the change in difficulty will be.
If you and a friend both get 15 questions correct on the Verbal section of the GMAT, what do you think will be true about your scores?
Actually no. This is a common misconception from students who will ask how many questions they need to get right in order to get a certain score. The difficulty levels of the questions you get right and wrong are used to estimate the difficulty level of questions you have a 60% chance of getting correct.
Not necessarily. While the first few questions in a section are important for determining the difficulty level of questions you will see in the exam, this can change quickly by getting several questions correct or incorrect in a row.
That's right! The questions you get correct will determine the difficulty of later questions. And even if you see the exact same set of questions, *which* ones you get correct will affect your final score. That's because **the difficulties of the questions you get correct and incorrect are what determine your score**.
Just as you might like your date more if they act admirably after you accidentally spill wine on them, the GMAT considers the circumstances of your performance, not just how many questions you get correct.
Here is a simple visual to show you how correctly answering 3 of the first 5 questions can result in a different score estimate depending on which questions you get correct.
In extreme cases, two different test takers can get either a top score or a very low score even if they each answer around 60% of the questions correctly. This makes the strategy for taking the GMAT, which will be explored later, trickier than almost any other exam.
Contrary to what some say, this is **not** because early questions affect the score more. Rather it's because the early questions have a larger impact on the difficulty levels seen in the rest of the section. Based on this understanding, how do you think you should approach early questions?
This is true for some students, especially those who don't make many silly mistakes.
That's the right strategy for most test-takers.
No, this isn't ever going to be worth it.
While it is still possible to get a top score if you do poorly on the first few questions, you'd need to get at least 85% or more of easy and medium questions correct in order to climb quickly enough to the hard questions that lead to higher scores. That isn't a lot of margin for error.
Spending a little bit of extra time double-checking your work on the first few questions will help you avoid getting into the easy questions without giving up too much time. You can spend 2:15-2:30 **on average** and even close to 4 minutes on a single question if you are confident you can get it correct.
If you are someone who makes a lot of silly mistakes, like calculation errors or misreading information, you want to aim for the higher end of that time estimate because you really can't afford being directed into the easy questions. Once that happens all the silly mistakes will prevent you from getting enough questions right to climb out.
Some test takers think this means they should do whatever they can to get early questions correct and get into the hard questions. What downside can you see in doing this?
Incorrect.
That's right!
While it *is* true that it is harder to get hard questions correct, it isn't a downside. You are expected to get fewer hard questions correct and the scoring algorithm accounts for this.
On the other hand, it takes longer to try and tackle harder questions. So if you spend more time on the first few questions, you risk running low on time. Being a little bit behind on time is manageable, you just need to prevent it from snowballing.
If you are behind on time, what strategy makes the most sense?
You got it!
There's a problem with this strategy.
This strategy is a recipe for disaster.
The first thing to know is that **you should never leave questions unanswered**. It has been officially stated by GMAT test-makers that leaving questions unanswered carries a higher penalty than answering them incorrectly. If you guess, you at least have a chance of getting some correct.
The problem with guessing more than a couple questions at the end is that as you miss questions from guessing, you will start getting easier questions wrong, which hurts your score severely. Additionally, you don't have a chance to recover when you guess at the end of a section.
The best option is to spread out your guesses through the middle part of the section so that you have time to answer the last 3-5 questions.
With all this talk about timing and guessing, when do you think is a good time to practice these strategies?
Now you should have a much deeper understanding of what the adaptive nature of the GMAT means for you. To summarize the discussion on adaptive scoring:
[[summary]]
Not quite. It would be weird to start a date by asking about how your date wants to raise future children because it may not be relevant at the beginning of the relationship, and it's a surefire way to scare your date away.
Definitely not. This is a very common mistake that students make. While you are learning new content, you should focus your energy on getting questions correct and mastering the material. As you learn, you will naturally get faster, and there will be plenty of time to practice guessing and timing strategies later in your studies.
Yes, but definitely only after you feel that you have mastered the content. The most important thing is to improve your accuracy first and then work on timing.
You will see the time awareness tool within the course. The goal of it is to get you to correctly *feel* how long you are taking, not to get you to answer questions in under two minutes.
Definitely! Practice exams are the best way to practice your timing and guessing strategies because it is simulating the actual test experience. You will be getting questions around the difficulty you expect to see on the actual GMAT, so try to stay at a fairly consistent pace and spread guesses out to make up time, as needed.
The GMAT may not be your true love, but you have a better chance to impress if you know how the adaptive scoring system works:
[[summary]]
Whether you have fun
Whether they make a good first impression
What you can determine about their true personality
How many questions I get right.
How quickly I finish the exam.
Which questions I get right.
So, early mistakes can ruin my score?
So are the first questions the most important?
OK, so what if I run out of time?
So guessing is OK?
Continue
Continue
Start by asking the easiest questions and slowly give harder questions
Start with medium questions and adjust based on how the student does
Start by asking the hardest questions and slowly give easier questions
Noticeably more than 60%
About 60%
Noticeably less than 60%
They will be the same
Whoever got more of the first five questions correct will have the higher score
You can't say anything for certain about the scores
The same as any other questions
Spend a little bit more time on them
Spend a lot more time on them
It's harder to get those questions correct
It takes longer to answer hard questions
Spread out guesses through the middle part of the section
Answer as many questions as you can and then guess at the end
Answer as many questions as you can and leave the rest unanswered
While learning new content
While practicing questions after mastering the content
Proceed to step
During practice exams
