Efle's MCAT 2015 to Old MCAT Percentile Comparison/Conversion Tables

[Affiche]

I am not misunderstanding percentiles, you are misunderstanding ranges. Each scaled score accounts for a range of percentiles. There are multiple raw scores on the MCAT that convert to the same scaled score. For example, a 74% correct and a 76% correct may both convert to a 31 (totally hypothetical), but they first converted to different percentiles. Each scaled score has a range of percentiles that it encompasses, but only the ceiling of that percentile range is listed.

Read the legend on the conversion chart. It explicitly states that the listed percentile is the ceiling of that score, and each percentage listed includes that percentage and those below it. Do you honestly think that on a 230 question test you can only score either 77th percentile or 80th? No, there are percentiles in between those, but they get clumped into one scaled score.

---

Members don't see this ad.

 

[deleted393595]

@gyngyn: Out of curiosity, do you see a benefit to the new MCAT? Because the claim that it results in more "well-rounded" doctors seems questionable because more content = spending more time studying for it. And that translates into less time outside in the real-world.

 

[Affiche]

@gyngyn: Out of curiosity, do you see a benefit to the new MCAT? Because the claim that it results in more "well-rounded" doctors seems questionable because more content = having to spend more time studying for it. And that translates into spending less time outside in the real-world.

LOL that is such an awesome way to look at it.

 

[deleted393595]

LOL that is such an awesome way to look at it.

I also have a much, much longer rant about it that I posted a few years back

 

[efle]

I am not misunderstanding percentiles, you are misunderstanding ranges. Each scaled score accounts for a range of percentiles. There are multiple raw scores on the MCAT that convert to the same scaled score. For example, a 74% correct and a 76% correct may both convert to a 31 (totally hypothetical), but they first converted to different percentiles. Each scaled score has a range of percentiles that it encompasses, but only the ceiling of that percentile range is listed.

Read the legend on the conversion chart. It explicitly states that the listed percentile is the ceiling of that score, and each percentage listed includes that percentage and those below it. Do you honestly think that on a 230 question test you can only score either 77th percentile or 80th? No, there are percentiles in between those, but they get clumped into one scaled score.

All 30s are listed as having the same percentile on the score report; all scores are always reported via the ceiling of their bin. When it says it includes the range from the reported ceiling down to the ceiling of the next score, it refers to the fact that 6% of test takers earned a 30

Think of it in terms of scored above. 20% of people scored higher than 30. 22% of people scored higher than 508.

 

[Affiche]

@efle one of the main goals of the new scale was to separate out all the scorers that were clumped at 30/31. Each score had a huge range of percentiles that too many people fell into. With the old scale, they were all seen as equal (I agree with you on that), but that was an issue because there was a lot of variance within that score= clumping of dissimilar scores. With the new scale, they narrow down the ranges for each score and separate everyone from one another. Thus, there are multiple new scores that convert to one scaled score on the old test.

 

[HarryH00d]

@gyngyn: Out of curiosity, do you see a benefit to the new MCAT? Because the claim that it results in more "well-rounded" doctors seems questionable because more content = spending more time studying for it. And that translates into less time outside in the real-world.

Gap years will just become the norm

 

[deleted393595]

@efle one of the main goals of the new scale was to separate out all the scorers that were clumped at 30/31. Each score had a huge range of percentiles that too many people fell into. With the old scale, they were all seen as equal (I agree with you on that), but that was an issue because there was a lot of variance within that score= clumping of dissimilar scores. With the new scale, they narrow down the ranges for each score and separate everyone from one another. Thus, there are multiple new scores that convert to one scaled score on the old test.

Not so familiar with the new MCAT, but if they're increasing the accuracy for one set of scores(30-31), they're losing accuracy elsewhere. I'm guessing that loss of accuracy takes place for high (36+) and low (24-) scores?

 

[Affiche]

All 30s are listd as having the same percentile on the score report; all scores are always reported via the ceiling of their bin. When it says it includes the range from the reported ceiling down to the ceiling of the next score, it refers to the fact that 6% of test takers earned a 30

Think of it in terms of scored above. 20% of people scored higher than 30. 22% of people scored higher than 508.

Yes, scores list the ceiling for both the new and the old test, but that doesn't mean you can disregard the percentile range that those scores cover. 6% of people may have earned a 30, but you're completely disregarding the fact that there were multiple raw scores that fell within that 6%.

Think of it this way:
Professor A is generous with his grades, so an A in his class ranges from 89.5-96.
Professor B is not as generous, so an A in his class ranges from 92-95.

Professor A has a larger range for an A grade, but that does not mean that all of the students in Professor A's class achieved higher percentages than students in Professor B's class. For example, Professor A's student scored a 91 in her class but Professor B's class scored a 93. They are both equal at an A, and just because it's listed on a transcript at "A, 96" and "A, 95" doesn't change that.

This isn't the best comparison, but it should give you a good idea of the importance of ranges.

Edit: this is actually a really poor example haha. Fixing percentages so it makes more sense.

 

[Affiche]

Not so familiar with the new MCAT, but if they're increasing the accuracy for one set of scores(30-31), they're losing accuracy elsewhere. I'm guessing that loss of accuracy takes place for high (36+) and low (24-) scores?

They shift the focus towards the middle, so the scores in the center are spread out but not so much towards the ends. On the new test, only 6 scores gave 100th percentile instead of 7 scores on the old test. If I could figure out how to attach an image I would upload my own conversion chart lol.

 

[deleted393595]

They shift the focus towards the middle, so the scores in the center are spread out but not so much towards the ends. On the new test, only 6 scores gave 100th percentile instead of 7 scores on the old test. If I could figure out how to attach an image I would upload my own conversion chart lol.

You can just copy-paste it like regular text

 

[efle]

But there is no need to make sloppy conversions like 508 -> 30 when you can just put up all the ceilings on a percentile scale. Is a 508 near/heavily overlapping with a 30? Yes. Many of the people who would have gotten a 30 will now get 508s. Are they equivalent? Not quite, as the 508 is more frequently out scored.

Your analogy doesn't work because its basedon raw score bins. Its more like saying prof A gives the middle two quartiles X and prof B gives the middle third Y. There is overlap but they cannot be said to interconvert ceiling percentiles

 

[Grace184]

You can just copy-paste it like regular text

Is that how it works?? I've been trying to figure it out...it just keeps asking for a URL and it was greatly upsetting me haha

 

[Affiche]

Your method of just looking at ceilings is sloppy an inaccurate, no offense. There is a huge range that you have to consider and you're disregarding it because you want to focus on a single number. It doesn't work that way. It's not that 508 is overlapping the range of a 30, it's completely within it! Anyone who scored in the same percentile as a 508 (77th) on the old MCAT was clumped into the 30, since it fell in the 30 range, and listed as 79th percentile. That is why the legend notes that it's a ceiling for 30 scorers, not the actual percentile for everyone that scored 30.

How do you think percentiles are generated? Our answers are graded for correct/incorrect and a raw score is generated then converted to a percentile and scaled score.

 

[HarryH00d]

Your method of just looking at ceilings is sloppy an inaccurate, no offense. There is a huge range that you have to consider and you're disregarding it because you want to focus on a single number. It doesn't work that way. It's not that 508 is overlapping the range of a 30, it's completely within it! Anyone who scored in the same percentile as a 508 (77th) on the old MCAT was clumped into the 30, since it fell in the 30 range, and listed as 79th percentile. That is why the legend notes that it's a ceiling for 30 scorers, not the actual percentile for everyone that scored 30.

How do you think percentiles are generated? Our answers are graded for correct/incorrect and a raw score is generated then converted to a percentile and scaled score.

Are you simply trying to reassure yourself at this point? All this energy...why not pre-write secondaries?... Speculation seems to be a cyclical sinkhole here.

 

[Affiche]

@ElCapone mine didn't paste haha. Hopefully this is still legible. I did this pretty quickly at work yesterday, so let me know if you see any errors/misread ranges. You'll see that multiple scores convert to one score on the old test, which helps narrow down what each score was (5 scores fall between 30-31, so no more clumping!). There are also scores that fall between old scores because the new percentiles are sometimes staggered with old ones. @HarryH00d it isn't speculation, it's reading a chart correctly. Doesn't take much energy and should be common sense.

Percentile Rank________Old Score_______New Score
100___________________45
100___________________44_____________528
100___________________43_____________527
100___________________42_____________526
100___________________41_____________525
100___________________40_____________524
100___________________39_____________523
99____________________38____________521, 522
98____________________37____________519, 520
97____________________36_____________518
95-96__________________35____________516, 517
92-94__________________34____________515, 516
89-91__________________33____________513, 514
84-88_________________32____________511, 512, 513
80-83_________________31_____________509, 510
74-79__________________30___________507, 508, 509
68-73__________________29___________505, 506, 507
62-67__________________28_____________504, 505
56-61__________________27_____________502, 503
50-55__________________26___________500, 501, 502
44-49__________________25_____________499, 500
38-43__________________24_____________497, 498
33-37__________________23_____________495, 496

 

[efle]

Your method of just looking at ceilings is sloppy an inaccurate, no offense. There is a huge range that you have to consider and you're disregarding it because you want to focus on a single number. It doesn't work that way. It's not that 508 is overlapping the range of a 30, it's completely within it! Anyone who scored in the same percentile as a 508 (77th) on the old MCAT was clumped into the 30, since it fell in the 30 range, and listed as 79th percentile. That is why the legend notes that it's a ceiling for 30 scorers, not the actual percentile for everyone that scored 30.

How do you think percentiles are generated? Our answers are graded for correct/incorrect and a raw score is generated then converted to a percentile and scaled score.

It shouldn't work that way; we should get a much more detailed distribution of exact raw scores. But thanks to the mysterious inner workings of the AAMC, it effectively does work that way. They collapse the raw bins into equivalently reported percentile ceilings, and force the only useful numbers to be ceilings (since floors are really just the ceiling of the prior score). And I also don't think it could work as neatly as you imagine. Verbal distribution used to differ from PS/BS, so do you call equal raw scores equal percentile when one got more right in Verbal and the other in BS? Regardless of what the system should be based on microdistributions within each score, what we get to work with instead is discrete ranges of equivalent scores defined by their ceilings.

I'm point out that in your analogy, what matters isn't the difference between the 89.5->A vs 92->A raw conversion methods, what matters is the number of people scoring A+/above your arbitrary A bin. Raw score conversions could be entirely different for MCAT 2015, and it matters little, because a score's worth continues to be defined by the number of people outscoring it

 

[HarryH00d]

Right... so 507< 508=30 <509

 

[Affiche]

Right... so 507< 508=30 <509

Exactly. 508 is not less than 30, but equal to it.

 

[HarryH00d]

@HarryH00d it isn't speculation, it's reading a chart correctly. Doesn't take much energy and should be common sense.

I was not referencing your inability to interpret a chart. I would hope your pre-medical curriculum refined that ability. But rather, the amount of time and energy you've spent arguing about it across 3+ different threads.

 

[efle]

@ElCapone mine didn't paste haha. Hopefully this is still legible. I did this pretty quickly at work yesterday, so let me know if you see any errors/misread ranges. You'll see that multiple scores convert to one score on the old test, which helps narrow down what each score was (5 scores fall between 30-31, so no more clumping!). There are also scores that fall between old scores because the new percentiles are sometimes staggered with old ones. @HarryH00d it isn't speculation, it's reading a chart correctly. Doesn't take much energy and should be common sense.

Percentile Rank Old Score New Score
100 45
100 44 528
100 43 527
100 42 526
100 41 525
100 40 524
100 39 523
99 38 521, 522
98 37 519, 520
97 36 518
95-96 35 516, 517
92-94 34 515, 516
89-91 33 513, 514
84-88 32 511, 512, 513
80-83 31 509, 510
74-79 30 507, 508, 509
68-73 29 505, 506, 507
62-67 28 504, 505
56-61 27 502, 503
50-55 26 500, 501, 502
44-49 25 499, 500
38-43 24 497, 498
33-37 23 495, 496

You can do this process more easily by taking my chart and collapsing the new score over, then looking at what the next old score above it is.

Eg:

Read as: The ceiling for 508 lies within the old 30 percentile bin, 2 percentile shy of the 30 ceiling.

 

[efle]

Exactly. 508 is not less than 30, but equal to it.

See above. It would have been a 30, but it does not express the same percent scoring above; it is 2%ile shy of true equivalence.

Edit: By your argument for equivalence, a 508 = 30 = 507. Do you really contest that a 508 > 507?

 

[Affiche]

It shouldn't work that way; we should get a much more detailed distribution of exact raw scores. But thanks to the mysterious inner workings of the AAMC, it effectively does work that way. They collapse the raw bins into equivalently reported percentile ceilings, and force the only useful numbers to be ceilings (since floors are really just the ceiling of the prior score). And I also don't think it could work as neatly as you imagine. Verbal distribution used to differ from PS/BS, so do you call equal raw scores equal percentile when one got more right in Verbal and the other in BS? Regardless of what the system should be based on microdistributions within each score, what we get to work with instead is discrete ranges of equivalent scores defined by their ceilings.

I'm point out that in your analogy, what matters isn't the difference between the 89.5->A vs 92->A raw conversion methods, what matters is the number of people scoring A+/above your arbitrary A bin. Raw score conversions could be entirely different for MCAT 2015, and it matters little, because a score's worth continues to be defined by the number of people outscoring it

That analogy is **** haha it was the best I could come up with while multitasking at work.

LSAT was much more transparent about scoring which is how I learned about scoring of standardized tests. I don't know much about the individual sections except that different raw scores convert to different percentiles, which makes sense. An easier section will knock you down in percentile for each wrong answer, whereas a harder section may take 2-3 questions to knock you down.

LOL I love your arrows. I understand your chart, but the setup of it implies that 508 is underneath 30 rather completely equivalent to it/within its range. The arrows help, though.

@HarryH00d yeah, I don't like to see poor logic spread around as gospel.

 

[efle]

I was not referencing your inability to interpret a chart. I would hope your pre-medical curriculum refined that ability. But rather, the amount of time and energy you've spent arguing about it across 3+ different threads.

Everyone understands her, we just disagree with her methodology. She wants to argue partial overlap in range as equivalence. We want to continue defining by ceilings like the AAMC has for years. By her method a 507 = 30 and 508 = 30, an impossibility as all 30s are listed as equivalent and this would create 507 = 508. This sort of thing is avoided (but the overlaps still recognized and understood) by using a chart like mine.

 

[Affiche]

No, because it's common sense that 508> 507, not equal to. Do you understand? There are multiple scores that translate to a 30 because the range was so large, but the new scale separates and ranks them.

You're misunderstanding; it isn't partial overlap, it's complete overlap. The range of one new score may fall completely within the range of its counterpart, which makes them equal. I never argued that partial overlap is equivalent, but total overlap is. If you look at my chart, I listed 509 as equal to both 30 and 31 because it has partial overlap, it isn't equal to either. Similarly, I listed 508 as equal to 30 because its range falls completely within the range of a 30.

 

[HarryH00d]

 

[efle]

No, because it's common sense that 508> 507, not equal to. Do you understand? There are multiple scores that translate to a 30 because the range was so large, but the new scale separates and ranks them.

You're misunderstanding; it isn't partial overlap, it's complete overlap. The range of one new score may fall completely within the range of its counterpart, which makes them equal. I never argued that partial overlap is equivalent, but total overlap is. If you look at my chart, I listed 509 as equal to both 30 and 31 because it has partial overlap, it isn't equal to either. Similarly, I listed 508 as equal to 30 because its range falls completely within the range of a 30.

I don't think I misunderstand, I think you've just thrown around "equal" too much. Agree with the bold, absolutely. My point is that translating to a 30 is not equivalence to a 30. A 507 and 508 both translate to a 30, but you need to express things as I did on my conversion to point out their difference; The ceiling for XXX is in the range of YY, Z percentile shy of the ceiling, with Z larger for 507.

 

[sazerac]

It seems like a lot of this could be avoided if instead of using ceilings, we used the middle of each range. 508 would be a 76th percentile. 30 would be a 76.5th percentile.

Alternately, we could just start translating once scale into another scale, using decimals. For instance, those old 30s translate into about a 508.3 on the new scale, because the middle of the 30 range is slightly higher than the middle of the 508 range.


Food for thought- medical schools are typically graded pass/fail. What should we do with all of those passes? They are all technically 100th percentile, because nobody scored higher than them. But in reality it represents a group that hovers around 85%, halfway between passing and perfect. All these ceilings merely distort the data.

 

[Affiche]

Agreed, but AAMC already stated that nothing on the new scale is supposed to be equal to the old scale. The best you can do is realize that there are multiple scores that fall within the range of a 30 and recognize that those scores have a ranking. To focus only on ceilings though and stagger the scores completely puts everyone with a new score at a disadvantage simply because their scores have a narrower range.

 

[HarryH00d]

Man...you kids with the new MCAT scores should just take a gap year or two until all this gets straightened out.

 

[efle]

Agreed, but AAMC already stated that nothing on the new scale is supposed to be equal to the old scale. The best you can do is realize that there are multiple scores that fall within the range of a 30 and recognize that those scores have a ranking. To focus only on ceilings though and stagger the scores completely puts everyone with a new score at a disadvantage simply because their scores have a narrower range.

It often works in the favor of the applicant as well you know. Eg if you're a 78th percentile tester, you would have been reported before in the 30/79th bin and now you are 509/80th bin.

Regardless, it makes sense to read both scores as "what percent of testers outscored them". The fairness of increasing precision in this answer is moot, it's a step forward for adcoms

 

[efle]

Man...you kids with the new MCAT scores should just take a gap year or two until all this gets straightened out.

It will be interesting to see whether the class profiles in the next MSAR have the same median percentiles for old and new mcat takers. Will the new test get more slack from being longer and tougher to study for? Or will the old test benefit from being familiar and known to correlate well to steps? Time will tell

 

[Affiche]

It often works in the favor of the applicant as well you know. Eg if you're a 78th percentile tester, you would have been reported before in the 30/79th bin and now you are 509/80th bin.

Regardless, it makes sense to read both scores as "what percent of testers outscored them". The fairness of increasing precision in this answer is moot, it's a step forward for adcoms

It also makes sense to read scores as "what percent of testers did he/she outscore?", and then you have to consider more than just ceiling percentiles. Your chart with arrows makes sense, as long as it isn't read as 507 and 508 are both less than a 30. You are looking at this as how many people outscored a 507/508/30, and that neglects analyzing how many people a 507/508/30 outscored. A 508 outscores more than a 507, but also outscores a 30 sometimes (due to the larger range and lower floor of a 30) and is also outscored by a 30 others (due to the larger range and higher ceiling of a 30). Both the ceiling and the floor matter.

Alas, I think we've both made our points. I wish you the best of luck this cycle!

 

[HarryH00d]

It will be interesting to see whether the class profiles in the next MSAR have the same median percentiles for old and new mcat takers. Will the new test get more slack from being longer and tougher to study for? Or will the old test benefit from being familiar and known to correlate well to steps? Time will tell

Scores get you to the door...ECs/PS/Interview get you through it. I don't think matriculation data will have an MCAT edition bias.

 

[Nietzschelover]

I want to take a crack at this from a non-math POV. Is it possible that the issue is that a "30" as a signifier is overimbedded with signification? Won't adcoms eventually be comparing 507s to 508s, 509s, etc, etc, such that eventually 508 will enjoy (or suffer) its own meaning?

 

[HarryH00d]

Bingo.

 

[Grace184]

Y=mX+B

 

[Grace184]

But only if m is =,<,> 508

 

[Grace184]

I hope this clears things up

 

[gettheleadout]

I am not misunderstanding percentiles, you are misunderstanding ranges. Each scaled score accounts for a range of percentiles. There are multiple raw scores on the MCAT that convert to the same scaled score. For example, a 74% correct and a 76% correct may both convert to a 31 (totally hypothetical), but they first converted to different percentiles. Each scaled score has a range of percentiles that it encompasses, but only the ceiling of that percentile range is listed.

Read the legend on the conversion chart. It explicitly states that the listed percentile is the ceiling of that score, and each percentage listed includes that percentage and those below it. Do you honestly think that on a 230 question test you can only score either 77th percentile or 80th? No, there are percentiles in between those, but they get clumped into one scaled score.

I think you're misrepresenting the scoring process here. With respect to the scoring of the old MCAT, the raw score per section was scaled to a score out of a 15 based solely on the number of questions correct out of the total. This had nothing to do with relative performance compared to test-taking peers, which is what's important when looking at percentile rank for performance. The percentile rank was generated by noting that, for example, if 12% of test takers scored 30, then the upper limit for the 30-scaled-score percentile rank is 12% above the total percentage of test takers who scored below 30. Because the raw scores vary within the group of 30-scorers, any given person with a 30 could fall somewhere within that 12% range. Some group of people got the highest proportion of questions correct (and thus the highest raw scores) possible to still scale to a 30, and some group got the lowest proportion correct possible.

 

[Affiche]

I think you're misrepresenting the scoring process here. With respect to the scoring of the old MCAT, the raw score per section was scaled to a score out of a 15 based solely on the number of questions correct out of the total. This had nothing to do with relative performance compared to test-taking peers, which is what's important when looking at percentile rank for performance. The percentile rank was generated by noting that, for example, if 12% of test takers scored 30, then the upper limit for the 30-scaled-score percentile rank is 12% above the total percentage of test takers who scored below 30. Because the raw scores vary within the group of 30-scorers, any given person with a 30 could fall somewhere within that 12% range. Some group of people got the highest proportion of questions correct (and thus the highest raw scores) possible to still scale to a 30, and some group got the lowest proportion correct possible.

Correct.

 

[Holmwood]

Postbac programs list their percentile requirements now... and it's weird. They pretty much round to tens place.

Like minimum to enter is 50th percentile, competitiveness to DO is 60th percentile, guaranteed acceptance to linkage at 70th percentile or 30 mcat...

Im too lazy to list sources right now and may do so later this evening. I think other schools would have updated their pages by then.

But in any case, it seems as if schools are not doing a direct percentile comparison (77 old = 77 new), but perhaps an arm wavy comparison (74 old = eh... 70 new).

 

[Affiche]

Lol there was a middle-tier school website that said last week that "515 is expected to be the minimum competitive score". I forgot where it was, somewhere in NY, but hopefully they cleaned that up by now! It'll be an interesting cycle, that's for sure.

 

[Holmwood]

Lol there was a middle-tier school website that said last week that "515 is expected to be the minimum competitive score". I forgot where it was, somewhere in NY, but hopefully they cleaned that up by now! It'll be an interesting cycle, that's for sure.

Well if it's a stat heavy school, maybe it isn't weird for it to be the competitive score even if it is "minimum". :/ Probably will go up or down, who knows. Not entirely unexpected for an NY school.

 

[Goro]

It will be a few years before we finally settle on a comfort number like 30 (10/10/10) was on the old exam. Again, keep in mind that the data efle was so kind to assemble was based upon only two months worth of exam scores. At some point, we'll be saying "505? Stick the to newest schools, DO schools, and these six/..." 508? try these..."

In the mean time, chill!

 

[gyngyn]

Hmmm. I understand how having screeners/readers examine apps of students who have similar gpas/mcats might bring about a more holistic review...but what about the possibility of screeners' biases towards specific ECs or experiences? Basically... where is the inter-rater reliability? Shouldn't more than one screener be assigned to a particular gpa/MCAT bin to ensure a true "holistic" review? @gyngyn

There are several more layers of evaluation, many more opinions into the mix and specific training on recognizing unconscious bias. Still, this only ensures that the pool is evaluated fairly. For individual candidates, the element of randomly getting a screener who likes (or is unimpressed by) your EC's is still present.

 

[eatandpoop]

@efle dude your real name is listed as the author for the excel file.

aaaaaaaaaaaaaaaaaaaaaaaand you look really good. dat h2o polo

 

[efle]

@efle dude your real name is listed as the author for the excel file.

aaaaaaaaaaaaaaaaaaaaaaaand you look really good. dat h2o polo

Well, ****. Taken down for now. Guess that removes any remaining anonymity I would've had, which is not much considering how I'll probably be the only person applying out of Wustl with my old MCAT next cycle anyways.

Thx

Wish I knew @gyngyn's school so I could skip applying to it...

 

[Cyberdyne 101]

Wish I knew @gyngyn's school so I could skip applying to it...

Why in the world would you not apply to his school?

 

[Holmwood]

Why in the world would you not apply to his school?

Questionable behavior on SDN, or any other media outlet, can negatively impact your application.
Gyngyn was one of the first to see the chart...

Although, I don't know if efle is naughty here in the boards.