Leveraging the GMAT Scoring Algorithm to Your Advantage

Yes, please don't take my advice literally by taking exactly 2.5 minutes on each question in the first 1/3 of questions. What I mean is "an average of 2.5 minutes per question." So if you need 4 minutes to solve a particularly work-intensive question, it's OK so long as you answer another question in just 1 minute (maybe because you've "punted," a.k.a. taken a wild guess, due to a lack of understanding, or maybe even a feeling/gamble that it's an experimental question).

I doubt this is helpful to many people. If you can actually control the correctness of your answers, you'd be better off simply answering all questions correctly.

Also, I suspect this thread will undermine and confirm many theories regarding the GMAT algorithm, so I think we should wait before announcing anything strategy-related based on the algorithm: information-on-new-gmat-esr-report-beta-221111.html

Yes, that's true, and obvious--of course you don't know for sure which ones you are getting right and wrong, and yes, in an ideal world you would just get every question right.

On the other hand, we do have some sense of how well we are doing on the GMAT, especially the high scorers--it's not like we can't tell the difference between a question that is hard for us and a question that is easy for us. The point is that if you know suspect that you've struggled on the last couple of questions, then you need to really hunker down on the next one, and vice-versa (if you're confident that you've gotten a few questions right in a row, then you can afford a wrong answer or two).

So yes, you are operating in the dark--sometimes the hardest GMAT questions are the ones whose answers look easy--but you can still utilize a standard operating procedure to leverage the algorithm, using a combination of strategy, intuition and common sense.

Hi Senthil7,

I wouldn't trust the Manhattan GMAT scores as anything more than wild-guess estimates. Not only are they not real GMATs, but they don't have access to the real GMAT scoring algorithm, either. Fake questions + fake algorithm = unreliable score estimates.

Feel free to use the MGMAT CATs for additional practice and learning / strategy, but take the score results with a giant grain of salt. The only scores you can trust are from the 6 official GMATPrep tests from GMAC.

It's true that we at Manhattan Prep don't have access to the official GMAT algorithm (no one else does, either), but GMAC has published a great deal of research on that algorithm, so there are some things we can be pretty confident in. For instance, our CATs do not weight later questions more heavily. Our CATs mimic the real test, which generally weights questions about the same, regardless of position. We also know that the algorithm is complex enough that it's impossible to reverse-engineer by taking 5 or 10—or 100—tests yourself.

Last year, Dr. Lawrence Rudner, former Chief Psychometrician of GMAC, spent several months reviewing our algorithm and test questions, including thousands of data points of our students' performance on our own CATs compared to their scores on the real test. In his official endorsement in August 2015, he said (among other things): "I can further attest that Manhattan Prep’s GMAT practice exams do an excellent job of predicting a student’s score on the actual GMAT examination." https://www.manhattanprep.com/gmat/blog ... -endorsed/

I'm sorry, but this doesn't pass the smell test:

1) MGMAT is your employer, so you are inherently biased on this issue.
2) Any corporation with money to spend can hire an "expert witness" to say whatever they want the witness to say. Just because the scientist hand-picked and paid by the "Dream Team" said that O.J. didn't do it doesn't prove anything.
3) The algorithm is irrelevant if the questions are not real in the first place.
4) Earlier questions are definitely weighted more than later questions. Check out this detailed analysis of the GMATPrep software from Bunuel (and the corresponding Verbal analysis from Vercules) if you don't believe me. No, the GMATPrep algorithm is not the same as that of the real test, but it is our most reliable source of information about the GMAT algorithm.
5) The GMAC (Mr. Rudner's former employer) loves to claim that earlier questions are not weighed more than later questions, so it is not surprising that he would stick to this story...but the results I've seen on real GMATs have not supported these assertions whatsoever.

mcelroytutoring,

You are actually wrong on many counts and I find many of your posts on GMAT Club misleading and very one-sided. I have been growing weary of your provocative claims and get-high-score-fast schemes that imply a high score without doing the actual work. I am also disappointed at the statements you have made above - they are shortsighted and illogical. It really saddens me to see anyone say things like that, and brings even more sadness when I think whey they would do that. It casts a pretty big shadow on your capabilities as a tutor and a fellow human capable to coexist.

Here are some of your most prominent faults:
1. GMAT Exam packs came out just a few years ago - you may sound belieavable to those taking the GMAT now but 3 years ago, there was only 2 GMAT Prep tests and that's it. Are you implying that people only took 2 CAT's for their prep? What should a person do who needs more than 6 CAT's? Based on your claim - they should not take CAT's but instead study questions that are missing even basic explanations. I am sorry but that's insanity.

Let me pop your bubble - GMAT and GMAT Prep do not use the same algorithm. They are different algorithms and there is a pretty big difference between GMAT Prep Test questions and questions you get in EP2. I have seen people get 670 on Test#5 and 600 on test #6 within a few days. How do you explain that? Or the fact that GMAT has been changing questions and had more meaning-based SC's and other questions that are not really present in the GMAT prep. In either case, you don't see that much variance in other products..... I think by blindly putting your faith only in the official things, you are drinking way too much coolaid but you are motivated to do that to sell your services. Otherwise, why would anyone hire you if they could just use MGMAT or Kaplan, or Veritas Prep books?

Your reference to the Bunuel's post, only proves that 10 of the first questions in the GMAT prep are important - it does not prove the same about the GMAT. You would need to analyze a bunch of the ESR's to figure that out. We are not there yet.


2. Are you saying that everyone preparing in the last 10 years using Kaplan, Princeton, Manhattan, Veritas Prep was blindly fumbling around? I did not even use the official stuff and I got 750. If anything, the official products are the lowest quality - look at the Official Guide and lack of hard questions or the meagerness of math and verbal information in it or the Exam Packs or GMAT Prep that do not even have explanations. They have other values but such spartan tools are only good to those who are at the end of their prep, not starting out. You are suggesting a very biased direction that is suitable for 1%, who probably don't even need these tools in the first place. In addition, there are no stats/analytics/explanation in the official CAT's - they serve only as a measurement/predictor of the final score - that's what they are good at but that's it.


3. It is possible to create an accurate algorithm that follows the GMAT scoring patterns with other questions - even we at GMAT Club with our meager resources have been able to do that for the Quant tests - all you need is a bunch of people who have received an official GMAT score and who have taken your tests - you can then correlate the results to accurately reflect one's abilities and get a GMAT score. We also regularly evaluate it against scores people report from taking the official GMAT and tweak it to adjust to the latest trends.


4. You have taken the easy way out of poo-pooing everybody else's work without creating anything of your own besides a speculative post that perhaps you can game the test - that's a get-rich-quick promise that has been around for ages and those who fall for it pay dearly. It is funny that you accused someone else of being biased when you are much more so. Not sure if I am the only one seeing it but you are clearly have been campaigning for people to use just the official materials, which are scant and are as helpful in prep as a thermometer for a fever.


5. There is a use for the official materials - it is good to use OG with your guidebooks or a course; it is good to use GMAT Prep tests when you need to get a score estimate but they are not sufficient to get a score higher than 600 or so - we have thousands of people who have fallen into that trap on GMAT Club saying that they have solved 3,000 questions and their score is still bad. Solving questions is not what gets your score up - it is a part of it, but a small part. The biggest score boost comes from knowing the principles behind the questions that is the most reliable way to get a good score.



P.S. This discussion is now locked. I am not concerned with the replies or comments but rather don't want to see this dirt brought up again. I am sad I had to write this post but I see no better way putting closure here.