Stumped by the MGMAT CAT scoring algorithm:

Okay, I took a look. I can see why you found your results puzzling, but nothing went wrong. Here's what I can tell you:

First, the simple answer is that your final score is not simply a function of how many you got right/wrong in each difficulty category. The algorithm is a great deal more complicated than that. For one thing, the difficulty ratings you see are just overall categories. The individual questions are not rated as "600 level" or "700 level"--that information is just there to provide a general sense of the difficulty of each problem. Therefore, there’s no way for you to actually calculate the average difficulty of the problems you got right or wrong. Even if you could, the order of responses matters. The algorithm is building up a model of your ability, so elements such as *when* you get a problem wrong, or how many in a row you get wrong, can make a difference.

Of course, we don’t have access to the *real* GMAT algorithm, but we’ve learned a great deal about it and we get as close as we can. Even the real thing is going to have some variation. In fact, it may interest you to know that your two scores, 81% and 79%, are actually statistically identical.

So the main thing for you, of course, is to look at how to improve from here. You’re starting out strong on these verbal sections, and then at some point you start to get in trouble and your accuracy rate drops substantially. In some cases, you miss many problems in a row. Why is that happening? To some extent, it may be the content—you’re having a hard time with high-level CR, so maybe once your score has worked its way up, the difficult CR drags it back down. There may also be an endurance element to consider. Maybe by the end of the test, you’re just tired. That’s the one way in which a tough math section can affect your verbal score!

Keep practicing, paying particular attention to maintaining your focus as you get tired. What can you do to refresh your concentration, or keep your mind on an especially troubling passage? Also, take the time to do some thorough review of the harder CR you’ve seen. Are you missing the conclusion or passing over a crucial assumption? Getting caught up on trickily-worded answers? Maybe some of the actual questions are confusing, and you need to practice ID-ing different question types. If you can find where the trouble is and push that accuracy up a bit, it should help your overall performance.

Thank you for looking into my CATS Dmitry!

It makes sense that whatever difficulty level that each question might be grouped into to show to the candidate, the algorithm may not have the same exact calibration. So, as much as I might have wanted to, I cannot reverse engineer the algorithm. I've got to move on and look to improve.

Indeed, I am guilty of some things that you mention I could be doing wrong. You give me some great ideas on where I could look for my problem areas and improvements, thanks a bunch for that.! MGMAT has been great at answering my queries on this issue. Kudos.

Hi Dmitry,
Could you please let me know whether you've categorization of OG/Verbal review CR questions according to difficulty level and solutions for the same?

Appreciate your feedback.

I'm experiencing a similar thing on my Manhattan GMATs. Here's a bit of data about my last quant scores:

CAT 1: 0 700-800 questions, mostly 600-700 questions. Score: 47%, 20 right total
CAT 2: 26 700-800 questions (11 out of 26 right). Score: 47%, 17 right total

I understand GMAT algorithm works in mysterious ways, but it makes little sense to me that a test feeds you a ton of 700-800 questions (and by a miracle I still got quite a few right ), yet gives you the same score as for a test that mostly tested 600-700 and 500-600 questions. I understand that I got less questions right, but 700-800 questions are much harder than 600-700 questions. Any reasoning behind this? Thanks.