90abyss wrote:
I've been taking Manhattan mocks and I've observed a very strange pattern in percentiles. For example I was on qs #15 and currently on 99%, then I got an RC passage which I completely screwed up ending up with 3 continuous mistakes which reduced my percentile to 80%.
Also I've noticed that even if I go wrong or right many times in the last 10 questions of any section then my percentiles are hardly affected. So is it a just estimation to conclude that first 15 questions matter more than last 10?
It is so unfair to see one fumble in an RC passage decreasing your percentile by 20!
I can't speak to how
MGMAT specifically calculates their percentiles, but I can say a little about percentiles in general.
A percentile is a very tricky thing. Ostensibly, it's easy --- if you are in, say, the 93% percentile, you have a score or a level of achievement above 93% of the population. So far, so good. Notice, it does not function the way an ordinary point total functions ---- the difficulty of an individual question or task, as well as the combination with other questions and tasks, will impact percentile in a complex way.
In your overall GMAT score, for example --- if you get, say Q50 and V60, that will not be the same percentile as Q60 and V50 --- in other words, it depends on the percentage of the population who scored that particular combination. One point up or down in either Q or V could result in seriously non-linear moves in percentile. In another example, it could be, say, that your Q is 95th percentile, your V is 95th percentile, but your overall is 98% percentile. Again, the percentile number is never a simple average. Even if only 5% of the population did as good as or better than you did on the Q, not all of them did well in verbal; even if only 5% of the population did as good as or better than you did on the V, not all of them did well in math; therefore, the combination of strong in Q & V is a higher percentile than the percentile rank in either of them separately. Percentiles never "average". They never go up or down in an easy to understand pattern. They are, by their very nature, something incredibly complex, because they are data-driven, and anything based on data can be complex and anti-intuitive.
I can't speak to your specific question, but that might give you some insight into what is difficult about understanding percentiles.
I have some very practical advice.
Absolutely ignore your question-by-question percentile rank! If there's some way to turn it off, then turn it off! Paying attention to that as you work through your GMAT sound like absolute mind-poison. Any emotional energy & focus you invest into worrying why your percentile rank is this and not that is emotional energy & focus you no longer have to devote to performing well. Yes, in some general sense it's good to have a ballpark idea about how well you are doing, but detailed attention to ups & downs of that is 100% unproductive. All your focus should be on the content and strategies that will actually result in higher performance on the GMAT. Investing any emotional energy into your score/percentile along the way is a pure waste of time that doesn't do diddly-squat to make you one iota better in how you actually perform.
It's this eternal student paradox: in order to get result of high performance (good grades, good scores, etc.) you have to focus totally on the task at hand and let go, as much as possible, of any attachment to the external measures (grades, score). Emotional investment in the external results tends to inhibit achieving those results.
Does all this make sense?
Mike