Hypo method to reform LS admissions Forum

[gwuorbust]

so as I ride out WLs I thought of the following method as a possible reform.

The problem with the current system is an obsession on numbers. The difference between a 165 and a 166 is statistically meaningless but in LS admissions it can make the difference between admit and WL or 10k and 20k, etc. Obv this is all for USNWR.

I do not think admissions should eliminate numbers but it should also be able to incorporate the other parts of applicants' files.

As you all know LSAC reports score bands. So here is how I think a new system could look like:

176-180
170-175
165-169
160-164
...

where if you score anywhere within that band, the only thing LSAC reports in your file is the band. Therefore, say you get a 172. You are in the 170-175 band and the schools don't know what your real score is until after the cycle is over.

I think this could help prevent gaming the system. Your score would still measure within a reasonable degree of your abilities but it would prevent automatic cut-offs.

[yeff]

Sure would suck to get a 175, 169, 164, etc.

EDIT: Come to think of it, this would have benefited me personally, though. Hm.

Really your proposal is to use a new scaled scoring system. Instead of converting the raw score from 0-100/101 into 120-180, you'd scale it into 1-12, letter grades, whatever.

175 180 12 A+
170 174 11 A
165 169 10 A-
160 164 9 B+
155 159 8 B
150 154 7 B-
145 149 6 C+
140 144 5 C
135 139 4 C-
130 134 3 D+
125 129 2 D
120 124 1 D-

[manbearwig]

Sure would suck to get a 175, 169, 164, etc.

That would just be painful, if they did it that way. So many more people would retake. I know I would, if bumping my score one point was the equivalent of bumping it five.

[Tautology]

There's no reason to have a difference between a 170 and a 175 if you're not going to report it. If you want fewer possible scores that's one thing, but why hide the actual score until after it's served its purpose other than to piss off everyone who ended up above the middle of the band?

Oh, and you should figure out whether you're going to draw the line between xx4 and xx5 or between xx5 and xx6.

176-180
170-175
165-169
160-164

[Knock]

Although not the greatest solution, that still seems like an improvement to me. The problem is the LSAT's hyperprecision, that it can't possibly achieve. Think of NBA timing clock attempting to track things to the tenth of a second . It's just a level of hyperprecision that isn't feasible.

I think a better reform would be to make each prospective student take the LSAT 3 times. This has the incidental benefit of getting people who aren't really serious out of the pool. Each prospective student must take this new format as scheduled, ie, 3 weekend in a row, or 3 days in a row, or however they want to do it, and the new tests would be specifically designed for this 3-test = 1 score format. These 3 tests would each include the following: 1 easy RC, 1 medium RC, 1 hard RC; 1 easy LG 1 medium LG 1 hard LG; 2 easy LR 2 medium LR 2 hard LR. The ranking of the system would be decided by LSAC as the sum of the difficulty of the questions in the sections (they already rate the questions 1-5, check out SuperPrep if you're interested in getting a feel for their difficulty ratings; an "easy" section could just be a section with total difficulty of say 80, "medium" 90, "hard" 100). This keeps people from getting luckily high scores from a test perfectly designed for them, hitting all their strengths and minimizing their weaknesses, and people getting unluckily low scores, due to having that particular test hit all their weaknesses and non of their strengths.

This also reduces the hyperprecision. Now the difference between 170 and 171 isn't merely a single question, for which a person could have guessed and gotten lucky or unlucky, but at least a solid 3 questions. With this method, a single bad day wouldn't kill you, and a single lucky day wouldn't make you.

[vanwinkle]

I do not think admissions should eliminate numbers but it should also be able to incorporate the other parts of applicants' files.

They do this now. They also consider UG GPA, PS, LORs, and WE.

As you all know LSAC reports score bands. So here is how I think a new system could look like:

176-180
170-175
165-169
160-164
...

where if you score anywhere within that band, the only thing LSAC reports in your file is the band. Therefore, say you get a 172. You are in the 170-175 band and the schools don't know what your real score is until after the cycle is over.

This would make getting a 170+ even more important than it is now. It would not accomplish what you're thinking.

I think this could help prevent gaming the system. Your score would still measure within a reasonable degree of your abilities but it would prevent automatic cut-offs.

It would just change the rules, it wouldn't end the game. Schools would just focus on everyone in the 170-175 band instead of individually focusing on a narrower range (like 170-172 or whatever they usually focus on).

[Tautology]

Although not the greatest solution, that still seems like an improvement to me. The problem is the LSAT's hyperprecision, that it can't possibly achieve. Think of NBA timing clock attempting to track things to the tenth of a second . It's just a level of hyperprecision that isn't feasible.

I think a better reform would be to make each prospective student take the LSAT 3 times. This has the incidental benefit of getting people who aren't really serious out of the pool. Each prospective student must take this new format as scheduled, ie, 3 weekend in a row, or 3 days in a row, or however they want to do it, and the new tests would be specifically designed for this 3-test = 1 score format. These 3 tests would each include the following: 1 easy RC, 1 medium RC, 1 hard RC; 1 easy LG 1 medium LG 1 hard LG; 2 easy LR 2 medium LR 2 hard LR. The ranking of the system would be decided by LSAC as the sum of the difficulty of the questions in the sections (they already rate the questions 1-5, check out SuperPrep if you're interested in getting a feel for their difficulty ratings; an "easy" section could just be a section with total difficulty of say 80, "medium" 90, "hard" 100). This keeps people from getting luckily high scores from a test perfectly designed for them, hitting all their strengths and minimizing their weaknesses, and people getting unluckily low scores, due to having that particular test hit all their weaknesses and non of their strengths.

This also reduces the hyperprecision. Now the difference between 170 and 171 isn't merely a single question, for which a person could have guessed and gotten lucky or unlucky, but at least a solid 3 questions. With this method, a single bad day wouldn't kill you, and a single lucky day wouldn't make you.

And we get to fork over even more money to the LSAC. How about you eliminate the effects of one particularly bad day/section by allowing people who aren't happy with their scores to take it again. Oh wait . . .

The LSAT publishes score bands, the schools know what they mean. I think the problem you all really have is that the USNWR rankings include as a factor the median LSAT score, and so schools have to care about the actual numbers they get to report.

[Knock]

Although not the greatest solution, that still seems like an improvement to me. The problem is the LSAT's hyperprecision, that it can't possibly achieve. Think of NBA timing clock attempting to track things to the tenth of a second . It's just a level of hyperprecision that isn't feasible.

I think a better reform would be to make each prospective student take the LSAT 3 times. This has the incidental benefit of getting people who aren't really serious out of the pool. Each prospective student must take this new format as scheduled, ie, 3 weekend in a row, or 3 days in a row, or however they want to do it, and the new tests would be specifically designed for this 3-test = 1 score format. These 3 tests would each include the following: 1 easy RC, 1 medium RC, 1 hard RC; 1 easy LG 1 medium LG 1 hard LG; 2 easy LR 2 medium LR 2 hard LR. The ranking of the system would be decided by LSAC as the sum of the difficulty of the questions in the sections (they already rate the questions 1-5, check out SuperPrep if you're interested in getting a feel for their difficulty ratings; an "easy" section could just be a section with total difficulty of say 80, "medium" 90, "hard" 100). This keeps people from getting luckily high scores from a test perfectly designed for them, hitting all their strengths and minimizing their weaknesses, and people getting unluckily low scores, due to having that particular test hit all their weaknesses and non of their strengths.

This also reduces the hyperprecision. Now the difference between 170 and 171 isn't merely a single question, for which a person could have guessed and gotten lucky or unlucky, but at least a solid 3 questions. With this method, a single bad day wouldn't kill you, and a single lucky day wouldn't make you.

And we get to fork over even more money to the LSAC. How about you eliminate the effects of one particularly bad day/section by allowing people who aren't happy with their scores to take it again. Oh wait . . .

The LSAT publishes score bands, the schools know what they mean. I think the problem you all really have is that the USNWR rankings include as a factor the median LSAT score, and so schools have to care about the actual numbers they get to report.

Most people that are serious about LS fork out at least a thousand dollars + on the whole process, and many many go way above that (class, prep materials, LSAT registration fee, LSDAS fee, application fees). For people who get fee waivers anyways, it shouldn't increase the cost. For people who pay for the LS process, it would only fractionally increase the cost, while increasing the fairness by a much more significant portion.

Retakes are only offered every 4 months or so as well, and negatively impact your application increasingly with every retake. This would significantly reduce retakes as well.

[Tautology]

Most people that are serious about LS fork out at least a thousand dollars + on the whole process, and many many go way above that (class, prep materials, LSAT registration fee, LSDAS fee, application fees). For people who get fee waivers anyways, it shouldn't increase the cost. For people who pay for the LS process, it would only fractionally increase the cost, while increasing the fairness by a much more significant portion.

Retakes are only offered every 4 months or so as well, and negatively impact your application increasingly with every retake. This would significantly reduce retakes as well.

And people who work on the weekends now get to figure out how to put aside three consecutive Saturdays (or whenever the tests are offered, the test taker now has to be free all three of those days). I'm not a big fan of the "well if you're spending a lot of money anyway why not add a little more" argument, especially when not all of us did take classes or buy more than one book for prep.

This is a logistical nightmare, and excessively burdens applicants. Again, the schools know what a score means, and the LSAC is very clear about how close it is to a student's "true score" and how well it predicts law school grades. The reason the schools take small LSAT score differences so seriously is the USNWR rankings, your beef is with them not with how the LSAC runs their test.

The idea that making everyone retake the test twice will significantly reduce retakes is laughable.

[GeePee]

Generally, making marginal jumps more significant is a bad choice when trying to achieve efficiency of information. This is an example of such a choice.

The problem isn't that LSAC's reporting policy makes single LSAT points significant, it is that a rankings formula forces schools to establish hard cutoffs in attempts to make meaningless jumps in the rankings. The use of a median, while having advantages, has the disadvantage of creating black-and-white zones instead of areas of gray. L1 metrics (medians) only have 2 discernible demographics-- above the median, and below the median.

If you're angry about 1 point significances, push for the use of non-median metrics instead of trying to display less perfect information when better info is available. I'd still disagree, but you'd at least be making a plausible argument.

[gwuorbust]

There's no reason to have a difference between a 170 and a 175 if you're not going to report it. If you want fewer possible scores that's one thing, but why hide the actual score until after it's served its purpose other than to piss off everyone who ended up above the middle of the band?

well the reason would be to allow USNWR to be calculated same as it is now, expect with less gaming.

gwuorbust wrote:
I do not think admissions should eliminate numbers but it should also be able to incorporate the other parts of applicants' files.

They do this now. They also consider UG GPA, PS, LORs, and WE.

meh, not really. With a 163 I will not be accepted into Emory, almost regardless of those other things (obv there are the occasional exceptions, but you all know what I am talking about).

gwuorbust wrote:
I think this could help prevent gaming the system. Your score would still measure within a reasonable degree of your abilities but it would prevent automatic cut-offs.

It would just change the rules, it wouldn't end the game. Schools would just focus on everyone in the 170-175 band instead of individually focusing on a narrower range (like 170-172 or whatever they usually focus on).

but what I think about this is that if they tried to do this all but the truly best schools would fail. Think if Vandy only accepted kids in the 170-175 band. Most would prob end up in the t-14 and Vandy would end up w/o enf students. Therefore, after doing that once they would have to accept a good mix of both 170+ and 165-169 ppl. Except they wouldn't know if you are in the 165-169 band if you were 165 or say a 169. This would make them focus more on the holistic aspects of the application then just your numbers. Someone couldn't just write: I liked college, I volunteered once, and oh yeah my numbers are above both of your 75ths by one point so I'm an auto-admit!!!!

[Knock]

Most people that are serious about LS fork out at least a thousand dollars + on the whole process, and many many go way above that (class, prep materials, LSAT registration fee, LSDAS fee, application fees). For people who get fee waivers anyways, it shouldn't increase the cost. For people who pay for the LS process, it would only fractionally increase the cost, while increasing the fairness by a much more significant portion.

Retakes are only offered every 4 months or so as well, and negatively impact your application increasingly with every retake. This would significantly reduce retakes as well.

And people who work on the weekends now get to figure out how to put aside three consecutive Saturdays (or whenever the tests are offered, the test taker now has to be free all three of those days). I'm not a big fan of the "well if you're spending a lot of money anyway why not add a little more" argument, especially when not all of us did take classes or buy more than one book for prep.

This is a logistical nightmare, and excessively burdens applicants. Again, the schools know what a score means, and the LSAC is very clear about how close it is to a student's "true score" and how well it predicts law school grades. The reason the schools take small LSAT score differences so seriously is the USNWR rankings, your beef is with them not with how the LSAC runs their test.

The idea that making everyone retake the test twice will significantly reduce retakes is laughable.

I'm not going to sit here and argue with you. Obviously what I was proposing wasn't a final hammered out proposal to LSAC, obviously isn't a perfect solution, and obviously is unlike to occur. I believe that this would increase fairness by more accurately reflecting people's scores, and have the benefit of creating some difficulty for applicants.

[Tautology]

I'm not going to sit here and argue with you. Obviously what I was proposing wasn't a final hammered out proposal to LSAC, obviously isn't a perfect solution, and obviously is unlike to occur. I believe that this would increase fairness by more accurately reflecting people's scores, and have the benefit of creating some difficulty for applicants.

If it didn't excessively burden applicants who work on weekends then it would be more fair, while still posing a host of other problems. It would probably also increase fairness for those who don't (because they don't have the means) do a lot of prep. That I like, but I think the logistical problem kills this proposal.

[Derrex]

In actuality, at the top, law schools gaming the LSAT have very little to gain because I heard they use percentiles to calculate ranks, not the actual score. I think the truth is, schools want to show off their LSAT scores because the LSAT is standardized and the single most correlated element to grades and it gives a general sense of the intelligence of the student body.

And in terms of scorebands, those are probably most useful at 170+, but once you hit the 160s, the accuracy of the test goes up. I think the LSAT was designed to be fairly precise in the middle and less so at the top.

[skamike]

The biggest problem I see with this is that it places extra burden on the applicant for a problem that isn't theirs. Wouldn't be easier to alter the USNWR or the application process itself than to place extra burden on the applicant?

Besides, There is a difference between a 165 and a 169, and a 170 and a 175.
It's been said before. The option to retake the test exists. This alleviates any problems with a "bad luck day"

[MC Southstar]

After you get into LS, no one cares.

[YCrevolution]

..

[capitalacq]

logic fail.