Best Guess on New Gre Percentiles

[Lox26]

"[T]op 5% of test takers" would be the fourth or fifth highest score range. I can't edit my post for some reason.

[Kitkat]

Overall, my scores have been pretty consistent. It appears that an accuracy of only 75-80% is sufficient for the highest score range in Quant, while a similar accuracy for Verbal corresponds to the top 5% of test takers. No idea how the spread at the top will look in the new scale. If Daisy's estimates are correct, those of us in the 750-800 range will score somewhere between a 155 and 170. That seems like a huge difference psychologically. I wouldn't be surprised if Karajan's hypothesis--that math scores may be on the lower end of the range and verbal on the higher end--is correct for many of us. ETS has specifically said that the new scale should allow for more meaningful comparison among top-scorers, in Quant especially. Therefore, I am afraid that a good score in the old scale (750-800) may not necessarily translate into a good score in the new scale. Best of luck!

While I completely understand you point on this, I would like to make a point of my own for it. If you look at your scores and the percentiles for those scores from the old test, the highest percentile you could get on the exam was a 94%. This was mainly because so many people were getting a perfect score for the Quant section, mainly by people going into math heavy fields. With this new test they are trying to get it so that you can start telling the difference between the people in the top 6% and more difference in the groups below. So for people who used to get into that group, will now instead get something like a 99% or a 97%.

So I don't think that everyone who is scoring in the 750-800 score range are going to be scoring on the lower range necessarily. But there is another reason why I am making this point. There are also people getting scores like 740-800, 700-800 and the like. It is not like there are specific ranges that are being given out. If you make the assumption that the top range for quant, or the verbal either, is more likely to give you a lower range, you would have to make the same assumption for all the other ranges as well. I am assuming that they calculate the ranges all the same over at ETS, to do otherwise seems like more work then is necessary. Previous people have commented that the score range is based on the data that ETS has already gotten on previous experimental sections, and will do a bell curve based on that data.

Edited October 1, 2011 by Kitkat

[crater21]

Therefore, I am afraid that a good score in the old scale (750-800) may not necessarily translate into a good score in the new scale. Best of luck!

This is precisely what I am worried about. Psychologically, I think it makes a difference. I'm applying for humanities programs, in which a high quant score is not needed. So, if one gets, say, 720 (which is 75th percentile), it looks impressive simply because it is over 700, even though the percentile is low.

On the new scale, 720 might translate into 155, which doesn't seem impressive at all. I know this makes no sense logically, but I think these subtle differences matter.

[blaspheming]

Manhattan Online Test (1210-1360):

560V, 650Q (1210)

560V, 650Q (1210)

610V, 710Q (1320)

660V, 700Q (1360)

ETS Paper Practice Exam (1270-1470):

590-690V

680-780Q

Powerprep II (1360-1520):

620-720V

740-800Q

Actual (1350-1550):

670-770V

680-780Q

[Kitkat]

This is precisely what I am worried about. Psychologically, I think it makes a difference. I'm applying for humanities programs, in which a high quant score is not needed. So, if one gets, say, 720 (which is 75th percentile), it looks impressive simply because it is over 700, even though the percentile is low.

On the new scale, 720 might translate into 155, which doesn't seem impressive at all. I know this makes no sense logically, but I think these subtle differences matter.

Yes, 155 sounds less impressive then 720. But I think that part of this is just because it is a scale that we are used to using for something like the GRE. When people think GRE they tend to compare it to something like the SAT's. But for the LSAT and MCAT, they use a different scale that is more similar to what this new scale is. I would say that impresiveness at the moment is going to have to depend on your percentiles when they come out, but the way I see it, for schools that I am looking at, percentiles, and not the score I got are what is important. Since the test is new, that is most likely what schools will be looking at. So don't worry so much about how impressive the score is, just try to think about how the school will look at it. If a 75% is good enough on the old scale, then it will be good enough on the new test.

[Lox26]

Yes, 155 sounds less impressive then 720. But I think that part of this is just because it is a scale that we are used to using for something like the GRE. When people think GRE they tend to compare it to something like the SAT's. But for the LSAT and MCAT, they use a different scale that is more similar to what this new scale is. I would say that impresiveness at the moment is going to have to depend on your percentiles when they come out, but the way I see it, for schools that I am looking at, percentiles, and not the score I got are what is important. Since the test is new, that is most likely what schools will be looking at. So don't worry so much about how impressive the score is, just try to think about how the school will look at it. If a 75% is good enough on the old scale, then it will be good enough on the new test.

What I'm more worried about is the spread. A 155 is 15 scale points from perfection (i.e., the highest possible score, which may still be achieved with 1 or 2 errors), whereas a 720 is only 8 scale points below the maximum. This difference may be more meaningful for high scorers. Assume we both score midrange for quant (no basis for this claim). A 770 is 87th percentile but just a few missed questions from 800. On a different day, with a slightly different set of questions, we may have hit 790 or 800 and would then be above 90th percentile. An evaluator might not see a meaningful difference between 770 and 790/800. In the new scale, we would score 158 or 159. While 90th percentile may still be only two points away, we no longer have a cluster of the top 6% test takers at 800. Now a 94th percentile scorer is 99th or 96th percentile. Now, we're not just below the top; we're in a completely different tier. We wouldn't get the benefit of the doubt that we're just as good as the best because it appears that we are the "second-tier" best. This of course assumes that such leeway is practiced in the current system, which it may not be.

Maybe I'm overthinking this. I understand that the GRE is mostly to make a cut-off for the department/university to which one is applying. I just worry that the context of what is achievable (e.g., actually being able to score in the 99th percentile when the ceiling of 94th percentile is removed) plays a role in how percentiles are perceived.

[Kitkat]

What I'm more worried about is the spread. A 155 is 15 scale points from perfection (i.e., the highest possible score, which may still be achieved with 1 or 2 errors), whereas a 720 is only 8 scale points below the maximum. This difference may be more meaningful for high scorers. Assume we both score midrange for quant (no basis for this claim). A 770 is 87th percentile but just a few missed questions from 800. On a different day, with a slightly different set of questions, we may have hit 790 or 800 and would then be above 90th percentile. An evaluator might not see a meaningful difference between 770 and 790/800. In the new scale, we would score 158 or 159. While 90th percentile may still be only two points away, we no longer have a cluster of the top 6% test takers at 800. Now a 94th percentile scorer is 99th or 96th percentile. Now, we're not just below the top; we're in a completely different tier. We wouldn't get the benefit of the doubt that we're just as good as the best because it appears that we are the "second-tier" best. This of course assumes that such leeway is practiced in the current system, which it may not be.

Maybe I'm overthinking this. I understand that the GRE is mostly to make a cut-off for the department/university to which one is applying. I just worry that the context of what is achievable (e.g., actually being able to score in the 99th percentile when the ceiling of 94th percentile is removed) plays a role in how percentiles are perceived.

So this is my problem with your analysis. You are implying that because there is now a 15-pt spread, instead of a 8-pt spread, that all of a sudden there are the people with similar percentiles, who will be looked at differently because it is a different spread of point. What I am trying to say is that adcomms will be comparing the percentiles, on the same score report and they will see how those two point systems will look next to each other. While we might question some adcomms judgements, I think that for the most part if they say they want scores in the 85%< range, they will still look at those scores, whether it is a 15-pt spread or a 8-pt spread. Especially if ETS is specifically saying that the extra points in that range is to help distinguish among the top most scores. I personally think that it is only fair to distinguish between those at the top. I am sure that a person who is at a 99% would like to be distinguished from a person who got a 94%.

But I think that whether it is a 94% on an 8-pt spread or a 94% on a 15-pt scale, that adcomms won't look at them very differently. They don't strike me as that arbitrary, although they seem very arbitrary about other things.

[Lox26]

So this is my problem with your analysis. You are implying that because there is now a 15-pt spread, instead of a 8-pt spread, that all of a sudden there are the people with similar percentiles, who will be looked at differently because it is a different spread of point. What I am trying to say is that adcomms will be comparing the percentiles, on the same score report and they will see how those two point systems will look next to each other. While we might question some adcomms judgements, I think that for the most part if they say they want scores in the 85%< range, they will still look at those scores, whether it is a 15-pt spread or a 8-pt spread. Especially if ETS is specifically saying that the extra points in that range is to help distinguish among the top most scores. I personally think that it is only fair to distinguish between those at the top. I am sure that a person who is at a 99% would like to be distinguished from a person who got a 94%.

But I think that whether it is a 94% on an 8-pt spread or a 94% on a 15-pt scale, that adcomms won't look at them very differently. They don't strike me as that arbitrary, although they seem very arbitrary about other things.

That's not really what I am implying. What I am saying is that for the highest scorers, the new percentiles may disadvantage those who fall in the 94th percentile and slightly below. Of course someone who is really a 99th percentile scorer would prefer to be distinguished from the 94th percentile scorer. Of course that is fair. However, it is also disadvantageous to the test takers on the lower end of the top scorers. This is not unfair, but it is unfortunate for someone in our score range if he/she had hoped for a score only 5 or so scale points below the max. But now the 94th percentile to 82nd percentile scorers are more than 5 scale points below max and so look worse by comparison.

That doesn't seem arbitrary to me. ETS is trying to achieve more of a spread to better compare applicants. This means that if one is 94th percentile (which I would be thrilled with), one is no longer a top applicant with respect to the GRE (whatever that counts for). One is a borderline top applicant. If we extend this reasoning, an 82nd percentile scorer is a borderline borderline top applicant. Right or wrong, human beings are not rational. Ninety-fourth percentile should be 94th percentile irrespective of how the information is presented. But, we cannot say with certainty that a 94th percentile applicant would be evaluated the same as in the old system, regardless of what ETS instructs.

Clearly, you and I disagree on this point. Adcoms, which are comprised of opinionated, analytical, and biased individuals such as we, may be similarly split on this issue. Hopefully not.

[Kitkat]

I think I am just looking at this from a different perspective then you are really, Lox26. The head of my department basically left me with the impression that in Earth Sciences that adcomms tend to look at the percentiles, not the number, and compare scores that way. It might be different in different fields, where other programs actually look at the score you got, and don't worry so much about the percentile. In my mind this makes sense, because I feel like the raw scores mean less if you don't look at the percentiles to give you some idea of what that score means. In m mind the GRE is all about having a four hour comparison test to other people. I understand your point on saying that people might see a 94% on a 15 pt spread differently then on a 8pt scale. But in the end isn't doing better then 94% of other test takers still doing better then 94% of other test takers? I think at those places where percentiles are what counts as the min that they want to see, I don't see those programs really changing it, or looking at something other then the percentiles.

[Lox26]

Kitkat, you may be right about that. I'm just worried that a hitherto decent-to-good score (750, 82nd percentile) might now be a so-so score (157). I see the future playing out as "Oh, I hope I don't get anything below 160" among applicants or as 160 becoming the new de facto cutoff among top programs. In other words, what if the definition of a "good score" changes after the new format becomes ubiquitous in the next few years? The hypothetical "good" cutoff does seem arbitrary, but even intelligent, thoughtful people use arbitrary heuristics at times (profs, committees, qualified applicants). I just don't want to fall behind. Ultimately, it probably won't matter, especially since I will be pursuing a masters before applying for a Ph.d. and might re-take for a higher score down the road. Still, I fret. *sigh* November cannot get here quickly enough.

Edited October 10, 2011 by Lox26

[Kitkat]

You are making the point that it is all a matter of perception of the scores themselves, and that being used to hurt the applicants. That is what is making you worried. I can completely understand that. I wish that they had made the switch to the new test sooner so that we could at least know what our final scores were, sooner then basically the deadlines for so many programs. Although I think that if they start making cutoffs for scores, they are going to first have to learn it. Yes until then they might make arbitrary cutoffs.

But I think that since as people have pointed out else where in the forums is that the results of a new version of an exam is that they tend to be lower then later scores. If that is true, then having a slightly lower score right now shouldn't hurt a person. I say this mainly because so many people are taking the test now with the new scores. As another point with this, if they only take people who make a certain score, even if that means a higher percentile, they are shrinking their applicant pool. Some schools can do this, but I don't think that all of them can.

There is also the fact that most programs don't just look at the GRE. I think that if you have a good application all around, even with a perceived lower score it should be fine.

[Lox26]

Kitkat, I have seen that claim as well. Have you been able to find support for first-round scores being lower-than-average ?

[Lox26]

http://www.yaledailynews.com/news/2010/jan/15/grad-students-react-to-new-gre/

Some Kaplan employee made that claim. Not sure how he knows this. I'm sure Kaplan has an R&D department, so it's possible their team has gone through several iterations of this process with different tests. This does not mean that new GRE scores will behave in kind, however. Sloppy argument, or sloppy reporting.

[Kitkat]

Very true, but like I said if it is true.

I still think though that a bigger point on this is to say that if they will only take a higher number on this new test, like a 160, and that is also a higher percentile then they were looking at previously, then they will be looking at a much smaller group overall for the next few years. While this might make them look more selective, I don't think that grad school is all about that. I think they are more likely to want to find students who fit the program better. After all most PhDs will be funded, and I don't think they will want to spend all that money just based on GRE scores alone. They will want to give themselves some room to work with.

[bjett003]

I just wanted to through my scores out there to contribute to the conversation. I took the test back in August and I got 660-760 on verbal and 680-780 on quant. Waiting this long for scores is driving me crazy! Thanks to everyone for posting.

[hmmatz07]

Great chain - best I've seen yet on the the new scores. Many thanks to all past posters.

It seems like folks have settled on the fact that 750-800 is the highest range you can get, which is good to know. Because that's the maximum possible range, I think it's safe to assume that the new scores will very across that full range. I'm with past posters on this front -it'd be great to have more clarity within that range, especially since that encompassed the top 15% of all old tests.

However, for reported ranges like 740-800, or 730-800, it seems like the eventual scores should be toward the lower end of that spectrum since these scores must be worse than those who achieved a 750-800.

For what it's worth, my scores were: 730-800 V, 750-800 Q.

Cheers everyone!

[resource]

Great chain - best I've seen yet on the the new scores. Many thanks to all past posters.

It seems like folks have settled on the fact that 750-800 is the highest range you can get, which is good to know. Because that's the maximum possible range, I think it's safe to assume that the new scores will very across that full range. I'm with past posters on this front -it'd be great to have more clarity within that range, especially since that encompassed the top 15% of all old tests.

However, for reported ranges like 740-800, or 730-800, it seems like the eventual scores should be toward the lower end of that spectrum since these scores must be worse than those who achieved a 750-800.

For what it's worth, my scores were: 730-800 V, 750-800 Q.

Cheers everyone!

Congrats on great scores, hmmatz. I suggested this in a different thread, but I don't necessarily think that a 730-800 will end up on the lower end of the range. My theory is that there is a roughly normal distribution of expected scores with mean=lower_bound+50 & sd~=15. With the standard deviation absorbing variation due to the weighting process (which has yet to occur). Thus, a 750-800 is more accurately a 750-850, but the top half of the distribution is right-truncated at the mean. This means that a 740-800 = 790+-, a 730-800 = 780+-, ... 690-790 = 740+-, etc. If this is true, that means there is a pretty good shot (>68%, if we assume a normal distribution), you will fall within 20 points above/below the midpoint of your range (or 20 points above/below the lower bound of your range plus 50, if you're bounded by 800).

[speculum]

Took it today!

750-800 V

640-740 Q

[alicejcw]

Congrats on great scores, hmmatz. I suggested this in a different thread, but I don't necessarily think that a 730-800 will end up on the lower end of the range. My theory is that there is a roughly normal distribution of expected scores with mean=lower_bound+50 & sd~=15. With the standard deviation absorbing variation due to the weighting process (which has yet to occur). Thus, a 750-800 is more accurately a 750-850, but the top half of the distribution is right-truncated at the mean. This means that a 740-800 = 790+-, a 730-800 = 780+-, ... 690-790 = 740+-, etc. If this is true, that means there is a pretty good shot (>68%, if we assume a normal distribution), you will fall within 20 points above/below the midpoint of your range (or 20 points above/below the lower bound of your range plus 50, if you're bounded by 800).

resource, I really hope your interpretation turns out to be correct!

I took the revised test on Monday, phew.

My ranges:

Verbal 710-800

Quant 700-800

I'm quite pleased with this considering that I only studied for about a month beforehand, and only used Princeton Review to study. I was expecting to do much worse based on the practice tests I took. I found the PR practice tests to be a good bit harder than the actual test.

I'm nervous to see how these scores translate to the new scale!

Edited October 15, 2011 by alicejcw

[George2248]

So after all this discussion I see we couldnt agree on a single theory for the results.... I cant believe we need to wait 3 more weeks!

[Kitkat]

That might be a good thing. We only have a few more weeks until the results officially come out, and once that happens it will probably be a moot point. But it is just nerve racking until we get there.

[resource]

So after all this discussion I see we couldnt agree on a single theory for the results.... I cant believe we need to wait 3 more weeks!

Science would be pretty boring if we all agreed on a single theory, don't you think?

[resource]

That might be a good thing. We only have a few more weeks until the results officially come out, and once that happens it will probably be a moot point. But it is just nerve racking until we get there.

But that's when we test our theories empirically! Not moot at all.

[Kitkat]

But that's when we test our theories empirically! Not moot at all.

Very true. But I think people will be more happy to know what they actually got.

[habanero]

I'm getting more and more nervous as each day passes! I really hope that I get above 85th percentile on Q.