ETS converted my old GRE scores to the new format

[blaspheming]

I logged into MyGRE today and realized that ETS rolled out a new score report page. The new format converts your old scores to the new format. I did terrible on it back in 2008, but I'll post my results,

V: 490 is now 152 (60th percentile)

Q: 570 is now 147 (40th percentile)

Again, these are official ETS conversions from https://mygre.ets.org . My new scores aren't up yet.

[habanero]

Yesss!! Everybody needs to log in ASAP and give us their old scores with new conversions!!!!!!!!!!!!!!!!!!

[eomentl]

Interesting... This proves that the scores do not scale linearly between formats. If they did, your new scores would be V: 149, Q: 155. Linear conversion formula is New=(Old+1750)/15 for those interested.

[cunninlynguist]

V: 590 = 159 (84%)

Q: 650 = 151 (59%)

[Kitkat]

My old score was:

V640 = 162(92%)

Q670 = 152(63%).

Looking forward to the new results!

Edited October 28, 2011 by Kitkat

[wassermelone]

V: 720 = 168 (98%)

Q: 770 = 161 (87%)

[Tdearr]

Hi everyone,

We are starting to analyze the new data. Here is our early take: http://www.manhattanprep.com/gre/blog/index.php/2011/10/28/friday-surprise-ets-releases-newly-converted-gre-scores/

And for the record My scores:

V: 700 = 166 (97%)

Q: 780 = 163 (89%)

Best,

Taylor

[kolja00]

So it is no longer possible to get a perfect on the Quant? lol.

[Tdearr]

It will still be possible, just harder. And a "perfect" new quant will put you in a higher percentile than a "perfect" old quant. The "perfect" score will be less common.

Best,

Taylor

[kolja00]

I'm not sure if I understand how it is possible to score above a 166 if an 800 is scaled to that? How is it possible to get more points on a test if you already answered every question correctly. Or is it possible to get an 800 by missing a few questions?

[blaspheming]

I was ok with getting 680-780 quant on the old scale, but it looks like it will hurt me a bit on the new scale.

[Tdearr]

I'm not sure if I understand how it is possible to score above a 166 if an 800 is scaled to that? How is it possible to get more points on a test if you already answered every question correctly. Or is it possible to get an 800 by missing a few questions?

The old test was a CAT (computer adaptive test) so you could miss questions and still get an 800. However, on the old test an 800 was the 94th percentile (six percent of test takers got a "perfect" score). ETS wants more differentiation at the top so they are allowing for room for 97th and 99th percentile scores by moving an old 94th down in terms of raw score. Essentially this is more of a signal that perfect on the old test was easier than perfect on the new test. It is a measure of difficulty more than a traditional score for number of mistakes. That is just our first take on the meanings of the new scores. We will have to see if ETS releases information to give more insight into this.

[habanero]

So I guess that my experience of the new math being harder might be an accurate assessment.

Wow, Taylor--Manhattan didn't hire you for nothing! Boi, u speedy: http://www.urch.com/...gre-scores.html

Edited October 28, 2011 by habanero

[Lox26]

I never took the old test, but my SO's scores are below:

V: 570 = 158 (80th percentile)

Q: 780 = 163 (89th percentile)

[Grunty DaGnome]

The new scales seem to make much more sense.

Now, if with all the interactive software, GRE can figure out how to create a separate math tests, appropriately scaled for English major test takers vs. Economics or Engineering majors. I'm not complaining about the math sections that test things like statistics and correctly interpreting tables of historical data -- it seem that English majors should know things like this in order to avoid drawing and perpetuating false social theories -- but do I really need to spend 2 month brushing up on factoring polynomials? Seriously, I learned how to do that in the 7th grade. I think I got a shiny gold star on my homework that week, and I have never in my life had occasion to factor a polynomial since then. Why give engineering schools a false picture of their applicants, all scoring in the 99% for math, because they beat a bunch of English majors on factoring polynomials or solving for fractional exponents and simultaneous systems of equations? Seems like if you took liberal artsers out of the mix, you'd get a much more accurate picture of the applicants abilities who are actually going to use these skills.

[kolja00]

The new scales seem to make much more sense.

Now, if with all the interactive software, GRE can figure out how to create a separate math tests, appropriately scaled for English major test takers vs. Economics or Engineering majors. I'm not complaining about the math sections that test things like statistics and correctly interpreting tables of historical data -- it seem that English majors should know things like this in order to avoid drawing and perpetuating false social theories -- but do I really need to spend 2 month brushing up on factoring polynomials? Seriously, I learned how to do that in the 7th grade. I think I got a shiny gold star on my homework that week, and I have never in my life had occasion to factor a polynomial since then. Why give engineering schools a false picture of their applicants, all scoring in the 99% for math, because they beat a bunch of English majors on factoring polynomials or solving for fractional exponents and simultaneous systems of equations? Seems like if you took liberal artsers out of the mix, you'd get a much more accurate picture of the applicants abilities who are actually going to use these skills.

I do think it would be interesting if the scores were scaled to intended grad major. That would be a much better way to compare candidates IMO.

[radioalfredio]

Okay, so I totally bombed the GRE the first time that I took it, but my new scores were estimated to be much higher.

V: 570, 158, 80% (the new estimated score is higher than predicted using eomentl's formula)

Q: 470, 142, 22% (the new estimated score is lower)

[balderdash]

I took it July 2010.

V: was 770, now 170, 99%.

Q: was 800, now 166, 94%.

Edited October 28, 2011 by balderdash

[habanero]

So if the old 800Q is a 166 (94th percentile) and the new 800Q is a 170 (99th percentile), does that mean that scores like 750 will be around 90th percentile? Gosh, this is just so interesting!

[emmm]

No, scores of 750 were below the 90th percentile. Why would that change?

[Grunty DaGnome]

What is interesting is that my old scores as reported on my 2006 mailer put my percentile rankings at V:94% Q:81% and AW: 77%.

Now, they are V:94% Q:80% and AW:84%.

Weird that my "5" on the written test changed quite a bit, and that's the numerical score that does NOT change under the new scale ?!?

[resource]

Some more data:

Quant: 760 old -> 160 new -- (84%)

Verbal: 540 old -> 156 new -- (73%)

[resource]

What is interesting is that my old scores as reported on my 2006 mailer put my percentile rankings at V:94% Q:81% and AW: 77%.

Now, they are V:94% Q:80% and AW:84%.

Weird that my "5" on the written test changed quite a bit, and that's the numerical score that does NOT change under the new scale ?!?

On the ETS score report screen:

PERCENTILE RANK (% BELOW)

The percentile ranks in this report indicate the percentage of examinees who scored below your score. Note that these percentile ranks may be different from those that applied when the scores were originally reported to you if the scores were earned prior to July 2011. This reflects annual updating of these data to permit admissions officers to compare scores, whenever earned, with those for a recent reference group.

[MartianQ]

I took it July 2010.

V: was 770, now 170, 99%.

Q: was 800, now 166, 94%.

Wow, that is just amazing. I mean I get the logic of the new system, but when a perfect math score on the old system translates into a less than a perfect score, and by 4 points at that, on the new system, and a less than perfect verbal on the old system translates into a perfect score on the new system, you know there is/was something wrong with one or the other. I guess percentiles were always the most important measure and now finally after all these years they have decided to reflect them in the actual scores

Congrats on an outstanding score by the way.

Edited October 29, 2011 by MartianQ

[resource]

I am doing some back of the envelope calculations and, based on the notion that ETS was trying to design a normal distribution, it looks like they did just that.

So based on the scores reported so far, I can ball park the median at 50%. If we assume scores are normally distributed, this will also be the mean. Since we all know the standard bell curve, we can take an educated guess at the standard deviation and narrow it down to approximately 9 points.

For quantitative scores, with mean=150 & sd=9, the top 2% should have a score of 150+8*2 = 168. We know in actuality that 166 (800) translates to the 94th pctle, so this guess seems pretty accurate.

For verbal (and here's where it gets fun), the new scores translate almost exactly in percentile terms to the quant scores. Again, assuming mean=150 & sd=9, we get the top 2% at 168 -- where a 168 (or 720) is exactly the 98%ile . In terms of lower ranges, one sd above the mean would be predicted at the 84%ile -- we have a data point: 159==590==84%ile.

Thus, it seems that the ETS did design the new score based on a normal distribution centered at the midpoint of the score ((170-130)/2)=150 with a standard deviation of 9 points (not a linear conversion).

Edited October 29, 2011 by resource