LSP Accuracy Forum

[MPMP]

I'm sorry if this has already been posted, but I applied in mid/late November and have only heard back from one school so far, so I am starting to get antsy. The only decision I have received was a waitlist by a school that Law School Predictor had as a Strong Consider (64%), and the schools that I am most interested in are just 'Considers', around the 56-59% range, so I am a little worried.

In your experience, do applicants usually seem to get in +50% schools more often than not, or does the Predictor overestimate one's chances? What percentage range do you consider a target and what would you consider a reach?

Thanks guys, you have been very helpful throughout my LSAT prep and application process.

[Mroberts3]

I think it has generally overestimated my chances, but that's just my subjective feeling.

It doesn't seem like it would be too accurate for splitters, and it doesn't take a school's specific outlook. For example, I was a "consider" at UCLA, but got outright dinged pretty early on. Both my GPA and LSAT were below median, but only by a tiny bit. Even though this would put me near the 50% mark, I think my odds were much lower because it looks like UCLA is trying to up its medians this year.

I think, despite its sample size, LSN is better. It shows what actually happened to people with your combination of numbers throughout the years.

[Cupidity]

LSN can't predict YP

WL with Admit and 92% of apps below me.

[turkishangora]

I agree Did you compare it to LSAC's official calculator? My chances at some schools were really really different (i.e. a lot lower).

[dj_spin]

This is a straight-up statistics fail, and here's why.

LS Predictor says what proportion of applicants with your numbers have historically gotten into the school. If you are not accepted at a school that it gives you a +50% rating then it's likely:

1. You are biasing the sample yourself because you have a weaker-than-usual application, meaning that as far as you will see it the LSP program has messed up across the board in its predictions. In fact, you are simply part of the 30% or 8% of applicants who did not get into that school with your numbers. Since cross-admission statistics are highly correlated, this probably means this will happen more than once (ie across the board) and so you will think it overestimated your chances. In fact you had weak softs/personal statement/letters (anything other than hard numbers)

2. LSP, to be accurate, should only be right in the aggregate population, not for an individual. This means that nothing of any substance can be gleaned from individual predictive failures. Indeed, when it fails to predict, to the individual whom it fails to predict accurately, it will probably do this for most of its predictions precisely because that is an unusually strong/weak applicant with those numbers.

3. As a corollary to the above point: if, at your numbers, it says 62% get in, that means it also predicts that 38% don't. The only honest way to measure that would be if we took your rejection as part of the whole pool of accepts/rejects and looked at whether as a pool of applicants it had called it correctly.

I hope this helps.

[Panther7]

LSP doesn't account for schools trying to protect medians.

[Tangerine Gleam]

I think LSP is as accurate as anything of the sort could be, based on the fact that it uses numbers from previous cycles. But this year's cycle in particular is quite competitive, and many schools are doing things (and rejecting folks) they wouldn't have last year.

YCR has done a killer job, I think, but no one should use it too seriously. I think it gives a good ballpark of what's possible in one's cycle.

[Panther7]

YCR has done a killer job, I think, but no one should use it too seriously. I think it gives a good ballpark of what's possible in one's cycle.

This, it gives you a ballpark and a few good ideas. However, it can't statistically show things such as Indiana's 164 cutoff requirement like LSN can. You need to use more than one tool to have a good idea what your chances are.

[Flanker1067]

It also misses trends, so if this year truly is a more competitve cycle, then everyone will feel like it didn't work.

[iamtaw]

It also misses trends, so if this year truly is a more competitve cycle, then everyone will feel like it didn't work.

+1

[astro1819]

Hear it from the source:

http://www.lawschoolpredictor.com/?page_id=173

Granted, this analysis was done for last year's cycle, and this year may be more competitive/harder to predict, but I think this is as good an answer you can get to your question.

[lawhawk]

LSP doesn't account for schools trying to protect medians.

this and splitters are two things LSP cannot take into account, but LSN can