Manhattan prep Cats

Question

I just gave the manhattan prep CAT exam and score 640(V31 Q47). Shocked to see the verbal score because i have been constantly scoring 38 and 40 in Kaplan and Princeton review CATs. Even the last manhattan prep cat i gave some months ago i scored 36 in verbal. Since then i have significantly improved in verbal and it has been showcased in other cat exams. When i reviewed my current manhattan prep CAT with the previous manhattan cat prep i saw a discrepancy in the no. of difficult question. In my current manhattan prep cat i got 23 no. 700-800 level question, 12 no. of 600-700 questions and rest were 300-500 level, whereas in my previous mahattan prep cat,in which i scored 36, i got 13 no. of 600-700 level questions, 14 no. of 700-800 level questions and 7 no. of 500-600 level questions. It can be seen that in my previous cat the level of questions were scattered according to my level, but in my current mock i got an disproportionate no. of high difficulty question. Even after getting multiple questions of 700-800 level questions wrong in a row i was still given the next question of high level. I'm highly demotivated by my score,because i cannot believe that i stand on verbal 31. Any manhattan expert please clear my doubt and explain whether the test mechanism was faulty or it is my actual level. I have attached some screenshots of the test to show the discrepancy of the system in throwing questions in two cats.

Attachments

previous cat2.png [ 118.19 KiB | Viewed 2713 times ]

previous cat.png [ 116.55 KiB | Viewed 2710 times ]

current cat 2.png [ 137.61 KiB | Viewed 2689 times ]

current cat.png [ 116.85 KiB | Viewed 2722 times ]

bb · Answer

Hi. I do not think the test mechanism is faulty.... just saying.... but looking at how many you got wrong upfront in verbal.... 6 out of 12 are wrong and in groups of 3 indicating further drops in score.... I mean the whole thing is red... you can't read a table and give a score but in this case, I can say it is not V41 material. Sorry.

shob22 · Answer

Hey BB. The reason i got 6 questions wrong at every interval of 12 is that i was getting more no. of high difficulty question despite getting questions incorrect. I have uploaded the ss of my previous cat in which i scored 36, to make a comparison between the two cats. I scored 36 on my previous cat and got less no. of 700-800 level questions and in my current cat i scored 31 and got almost 65% of 700-800 level questions. So, if my score is low then then no. of difficult question should also be on the low side. Ik that my level is not 45 but still i was expecting a 36 atleast.

shob22 · Answer

What i am trying to say is that test was not that adaptive. I have given veritas,gmatprep,kaplan and princeton cats and i think their adaptive nature was good. Even the manhattan prep cat which i gave earlier was adaptive. But this particular cat shocked me with unsual no. of high questions. In every other test if i am getting questions wrong my level will drop with the drop in the level of question , but here the case is different, my level came down but the level of questions remained the same

ReedArnoldMPREP · Answer

Hey there,

It's definitely an unusual outcome, based on what I've seen. But it's important to know something about how the algorithm works.

The algorithm will give you a question and predict your chance of getting that question right, based on what it knows about that question and its correlation to other questions, and based on your performance on previous questions so far.

So when the test starts out, it will give you an average question (that, say, 70% of people get right) and, since it knows nothing about you, will give you the average chance (70% in this case) of a correct answer. As the test goes on, the prediction mechanism 'refines' based on your answers. So at question 14, based on your performance on the 13 previous questions, it will make a prediction for your chances of getting question 14 right. If you've gotten several hard questions right, it might give you a question that it is quite sure you will get right.

The algorithm has troves of data of other test takers, and has a decent sense of the correlations between questions, particularly when there are several questions. For instance, if I give a test taker one hundred 300-500 level questions and they don't miss, I'd give them a higher chance of getting a 700 level question right than a test taker who missed eight.

Similarly, if I give a test taker one hundred 700 level questions and they get 35 right, I give them a higher chance of getting a 300 level question than a test taker who got 28 right.

Now, what if one test taker went 92 out of 100 on 300 level questions, and another test taker went 35 for 100 on 700 level questions--who gets the higher score? Well, I don't know, I'm not the algorithm. I don't know the correlations well enough to make a determination. But the algorithm does know. The algorithm knows how to compare these two students, even though they had totally different tests, because the algorithm knows--from the thousands and thousands of other test takers--how these 100 questions correlate to the thousands of other questions the test uses.

My point: yes, on the most recent test, you saw a lot more 700 level questions. Largely because you got a few right up front, so the algorithm adapted and started giving you those questions. Then you started to miss them, several at a time. So based on that data and all the correlations the algorithm has, it gave you a higher score on the test you saw fewer 700 level questions.

The other thing that is important to consider is that the range '600-700' and '700-800' is pretty broad. There's a big difference in a 610 difficulty questions and a 690, and a 710 difficulty question and a 750. A 690 and a 710 have much more in common with each other than they do with questions at the other end of their respective ranges, but they get lumped into the buckets they get lumped into. So it's very possible that while you saw more 700 level questions on the most recent test, many of them were pretty close to 600 level questions, and the 600 level questions on the previous test were pretty close to 700 level questions.

At the broadest level, I give this advice to all students in your predicament: don't worry about the mysteries of the algorithm, and never let one single test score get you down. Just learn what you can from every practice test and keep moving forward. Improvement is a trend over time, not a linear process guaranteed to be noticed in every next test score.

shob22 · Answer

Manhattan prep Cats

Prep Toolkit

Top 5 GMAT Debrief Videos

615 to 715 in 15 Days (Ria's Strategies)

More than Quant/Verbal Abilities: Speed vs. Accuracy

745 with Only Self-Prep and GMAT Club

From 555 to 765: 210-point Improvement

Perfect 805 Score Debrief by Julia

My Rewards