Quote:
You need hard questions in the test to improve accuracy of prediction and accuracy of measurement because if somebody takes a bunch of easy and medium questions you don’t really know their upper level so you need a good chunk of hard questions to get that determined. But you also need medium and easy in case the person is not scoring well to potentially determine their lower boundary and then the test tries to figure out and significantly reduce the upper and lower boundaries. This is Dunn through question picking and serving and at the very end. The test gives you a single score but in reality it’s kind of a range based on the certainty of the algorithm. On the gmat prep that range is supposed to be plus minus one on each of the sections and that’s what we’re aiming on gmat Club’s new algorithm as well.
Thank you BB for such a detailed insightful response regarding the algo and score range... but I have one out of curiosity question on top of this...
BB, As we are tweaking and doing hit and trial with our algo. then we should have any rough ball park estimate as target to achieve right ? I mean to say anything like if someone have 2 wrong question in Quant he would get a score of 84 to 86 or something of that sort in each section....
If this is correct then could you please give that detail what would be the pessimistic range of score on the basis of wrong answer in each section if that is possible...
PS - Yesterday, I have asked this question in our ofc hours call but unfortunately we are unable to conclude on this anything... we all feel GMAT algo is weird and its unpredictable to get this kind of range if you have any insights on this it would be highly valuable specially for DI section 😅
Thanks !