Acceptance Matrices using C/O 2016 Medians and 25th/75th Forum

[cotiger]

For all of the T14 that have posted medians, 25th, and 75th data for the class of 2016, I went to mylsn and found the likelihood of acceptance in the 2012-2013 cycle based on location in the categories of 0-24th, 25-49th, 50-74th, and 75-100th percentile for LSAT and GPA.

Edited: GPA and LSAT floors are now taken as the lowest non-URM score to get at least a waitlist in order to weed out useless data on the low end.

As an example, here's Harvard:

Harvard 3.30-3.76 3.77-3.87 3.88-3.94 3.95-4.33
164-169 0% 0% 4.8% 13.3%
170-172 8.6% 7.4% 57.1% 83.3%
173-174 20% 76.2% 83.3% 100%
175-180 35.3% 81% 93.8% 100%

The spreadsheet is here: https://docs.google.com/spreadsheet/ccc ... ZN2c#gid=0

Feel free to make it prettier if you're good at that kind of thing.

Note: URMs and ED excluded

[jrsbaseball5]

This is awesome. Thanks!

[whippersnappery]

This is awesome! Did you exclude or include URMs?

[cotiger]

This is awesome! Did you exclude or include URMs?

Excluded URMs and ED

[wolfgang]

this is rad.

Any way to incorporate assumed GPA floors into the matrix? I mean, I know that nothing is going to be perfect, and we could subdivide this further and further forever.

But in this case, given the fact that most schools seem to have an unofficial gpa floor, I think that by taking this into account you gain a significant amount of accuracy on the low end for the cost of only one additional column.

Take duke, for example. the low GPA column goes from 2.0-3.58, whereas someone with a 2.0 has really zero chance and someone with a 3.58 has a pretty good chance depending on lsat score.

Their gpa floor looked like it was about 3.3 last year (with one lower outlier), so if the column is added that separates the "2.0-floor" gpas from the "floor-25th) gpas, you end up with all people under the floor—who, while definitely not offering or intending to offer perfect percentages, are statistically very similar—lumped into one percentage range that will reflect the fact that a) outliers do exist, and b) despite this, your chances really do approach zero at a certain gpa. You also lump all people from the floor to the median GPA, and again, while not perfect, they all stand chances that are much closer together than a range that dips below the floor.

I just think that, since this is a matrix, the attempt is to consider all significant cutoffs, and I would argue that the gpa floor is one. It will really only improve the accuracy of the low gpa/high lsat box (i think!), but for such a small price, I think it's worth it.

The percentages boxes for 170-180 lsat for duke currently run:

39.9 - 93.1 - 97.3 - 87.6

If the subdivision is made, taking data for the last 2 years for the lower end data (URM excluded), it will look more like this:

13 - 53 - 92 - 96 - 93

That's a significant division.

I didn't want to just run in and change stuff on someone else's work! But if this is something that you think sounds good, i'd be happy to do it or to compile to data and send it over, or whatever. Or you can just ignore me if you think it's good enough

And, really, it is good enough. This was a killer idea.

[cotiger]

this is rad.

Any way to incorporate assumed GPA floors into the matrix? I mean, I know that nothing is going to be perfect, and we could subdivide this further and further forever.

But in this case, given the fact that most schools seem to have an unofficial gpa floor, I think that by taking this into account you gain a significant amount of accuracy on the low end for the cost of only one additional column.

Take duke, for example. the low GPA column goes from 2.0-3.58, whereas someone with a 2.0 has really zero chance and someone with a 3.58 has a pretty good chance depending on lsat score.

Their gpa floor looked like it was about 3.3 last year (with one lower outlier), so if the column is added that separates the "2.0-floor" gpas from the "floor-25th) gpas, you end up with all people under the floor—who, while definitely not offering or intending to offer perfect percentages, are statistically very similar—lumped into one percentage range that will reflect the fact that a) outliers do exist, and b) despite this, your chances really do approach zero at a certain gpa. You also lump all people from the floor to the median GPA, and again, while not perfect, they all stand chances that are much closer together than a range that dips below the floor.

I just think that, since this is a matrix, the attempt is to consider all significant cutoffs, and I would argue that the gpa floor is one. It will really only improve the accuracy of the low gpa/high lsat box (i think!), but for such a small price, I think it's worth it.

The percentages boxes for 170-180 lsat for duke currently run:

39.9 - 93.1 - 97.3 - 87.6

If the subdivision is made, taking data for the last 2 years for the lower end data (URM excluded), it will look more like this:

13 - 53 - 92 - 96 - 93

That's a significant division.

I didn't want to just run in and change stuff on someone else's work! But if this is something that you think sounds good, i'd be happy to do it or to compile to data and send it over, or whatever. Or you can just ignore me if you think it's good enough

And, really, it is good enough. This was a killer idea.

Yeah, that's a good point. I didn't do anything like that initially just because that would require more work and I was just doing it to kill time in the airport. I don't like just eyeballing a floor, though. So here's my thought.. use the score of the lowest waitlisted non-URM as the floor. If they're waitlisted there's at least a theoretical chance of acceptance, so the lowest scores there sort of proxy for the lowest scores possible to have some sort of chance at acceptance.

[lawbeahs]

delete

[cotiger]

I'm sure I'm missing something, by why is it that the highest GPA/ highest LSAT bracket is often less likely to be admitted than the highest GPA/ second-highest LSAT bracket? Is this YP?

Yeah, I'd say so. Especially looking at UVA, the most egregious and obvious YP school ever.

Notice that everyone still gives the most money to that highest GPA / highest LSAT bracket, just that some of them are hesitant about handing out an acceptance to someone who probably won't come. (Berkeley doesn't count because they're weird/stingy about money anyway)

[John Winger]

This is absolutely fantastic! Mad props my man.

[wtrc]

This is great! Thanks.

I wish my GPA was 0.02 higher

[RetakeFrenzy]

amazing! Thank you

[gobosox]

This gives me hope. At least I have an 8.6% chance at Harvard...

[Happy Gilmore]

Yale is the worst.

[t-14orbust]

damn, too bad my GPA isn't lower. lol yale

[MoMettaMonk]

damn, too bad my GPA isn't lower. lol yale

Lol Yale, indeed.

[AT9]

This is awesome. Gives me a glimmer of hope for my Duke ED app.

[cotiger]

damn, too bad my GPA isn't lower. lol yale

Clearly yield protecting

[cotiger]

Updated with Columbia. We have a new winner for importance of being at the LSAT median. Although based on the crazy high acceptance rate for NYU at below 25th GPA but above 75th LSAT, it might not stay that way.

[cotiger]

Now that all of the stats from last year are in, I thought I'd update this and make a few awards:

Most Splitter Friendly (>75% LSAT, <25% GPA)
1. Northwestern
2. NYU
3. Georgetown

Least Splitter Friendly (>75% LSAT, <25% GPA)
1. Stanford
2. Yale
3. Penn

Most important to be at or slightly above median LSAT vs slightly below
1. Columbia
2. NYU
3. Georgetown

Least important to be at or slightly above median LSAT vs slightly below
1. Berkeley
2. Yale
3. Stanford

Accepts the highest percentage of at or above both medians applicants
1. Duke
2. Harvard
3. NYU

Accepts the lowest percentage of at or above both medians applicants
1. Yale
2. Stanford
3. Berkeley

Friendliest to Reverse Splitters (<50% LSAT, >50% GPA)
1. Cornell
2. NYU
3. UVA

[Vincent]

Now that all of the stats from last year are in, I thought I'd update this and make a few awards:

Most Splitter Friendly (>75% LSAT, <25% GPA)
1. Northwestern
2. NYU
3. Georgetown

Least Splitter Friendly (>75% LSAT, <25% GPA)
1. Stanford
2. Yale
3. Penn

Most important to be slightly above median LSAT vs slightly below
1. Columbia
2. NYU
3. Georgetown

Least important to be slightly above median LSAT vs slightly below
1. Berkeley
2. Yale
3. Stanford

Accepts the highest percentage of above both medians applicants
1. Duke
2. Harvard
3. NYU

Accepts the lowest percentage of above both medians applicants
1. Yale
2. Stanford
3. Berkeley

Friendliest to Reverse Splitters (<50% LSAT, >50% GPA)
1. Cornell
2. NYU
3. UVA

I'd take this as further evidence that in order of black-boxiness,
1. Yale
2. Stanford
3. Berkeley

[xylocarp]

Holy yield protect, UVA

[lawschool22]

Great contribution!

[philly93]

AMAZING JOB!

seems to me that NYU pretty much only cares about LSAT (something i'm happy about!!)

[lastsamurai]

this is possibly the most useful thing since LS22's spreadsheets

[lawschool22]

Feel free to make it prettier if you're good at that kind of thing.

Prettied it up just a teeny bit