Which LSAT is the "easiest?" Forum

[MachineLemon]

There is absolutely, unequivocally, beyond a shadow of a doubt no advantage to be gained (in terms of difficulty) by taking the LSAT in one month and not another.

The statistics about which months have the most forgiving curves is accurate-- and it is meaningless. Don't draw any conclusions from it.

While it may be true that the curve perfectly compensates for the difficulty of the test, it does not follow that an individual's chances of doing well do not vary month-to-month. Even if the overall distribution of scores were the same for all administrations, this distribution would still fail to capture the % rank for individuals across those administrations. For example:

Let's say that I have some background knowledge of meteorites. This may enhance my performance on RC sections involving meteorites, and thus, increase my overall score. However, so long as there is another test-taker who is less familiar with meteorites, the overall distribution won't change. Nevertheless, I would in fact stand to gain from taking that particular administration, all things being equal.

Similarly, for someone who has a significant and fixed rate of stupid mistakes, an easier curve will be more forgiving since his or her errors are not wholly connected to the difficulty of the test. However, there are presumably enough people who make fewer dumb mistakes, but are more sensitive to difficulty, whose score will suffer in a harder administration. Of course, the benefits/harms would be reversed at an easier administration with a tougher curve. So long as these benefits and harms are roughly proportional and so long as the numbers of people affected by them are roughly equal, there will be no change in the overall distribution.

I found that, near the end of my studying, nearly all my mistakes were stupid. This resulted in better scores on tests with easier curves since my mistakes were unrelated to the test's difficulty.

[JasonR]

If the curve is based on how well people score on that administration on average...

It isn't.

Well, then what is it based on then? If it's based on the "difficultly" of the test, then the only way to measure that is to look at how people scored on it. If people statistically do worse on a given test, then it will be assessed as a more "difficult" test. I'm sure they get some useful data from the previous use of the questions in experimental sections, but they'd better not be solely relying on that!

And why had they better not rely solely on data from experimentals?

That's what they do. That's where all of their data comes from. It's what ensures that an aberrant test-taking population has zero effect on the scoring scale. The scale is determined before the test is ever administered.

[tomwatts]

The logic of scaling based on experimentals is that they have the performance of people with given scores on the section. That is, they assign someone a 160 based on the other sections, and they also get to see how that 160 person does on the experimental section. They assign somebody else a 155 based on the other sections, and they also get to see how a 155 does on the experimental. So they get to compare a 155 and a 160 (and every other score they assign) on this particular section and see how they would have to scale the section in order to predict a score of 155 or 160 (or whatever) in the future.

There's some after-the-fact examination of the data, but not much. The scoring grid (the "curve," though it's not a curve) is almost entirely done beforehand, as you would expect. And when they offer a test again (an unreleased international or February test), they use the same scoring grid.

[3v3ryth1ng]

The logic of scaling based on experimentals is that they have the performance of people with given scores on the section. That is, they assign someone a 160 based on the other sections, and they also get to see how that 160 person does on the experimental section. They assign somebody else a 155 based on the other sections, and they also get to see how a 155 does on the experimental. So they get to compare a 155 and a 160 (and every other score they assign) on this particular section and see how they would have to scale the section in order to predict a score of 155 or 160 (or whatever) in the future.

There's some after-the-fact examination of the data, but not much. The scoring grid (the "curve," though it's not a curve) is almost entirely done beforehand, as you would expect. And when they offer a test again (an unreleased international or February test), they use the same scoring grid.

Hmmm...

Not sure it's been discussed yet, but do you know if tests are derived from experimentals administered during the same months of previous years? In other words, was the December LSAT made of experimentals administered during a past December LSAT, or several past December LSATs?

That might make a difference because if the sample of test takers they're "equating" a score to is inherently different at different times of the year, that could affect the "curve" they make in the end. For example, what if December test takers are (for various reasons), on the whole, less studious, more anxious, or in some way more inclined to miss questions? To account for this possibility, wouldn't they want to make December tests with past December experimentals?

Just a thought...

[tomwatts]

Not sure it's been discussed yet, but do you know if tests are derived from experimentals administered during the same months of previous years? In other words, was the December LSAT made of experimentals administered during a past December LSAT, or several past December LSATs?

That might make a difference because if the sample of test takers they're "equating" a score to is inherently different at different times of the year, that could affect the "curve" they make in the end. For example, what if December test takers are (for various reasons), on the whole, less studious, more anxious, or in some way more inclined to miss questions? To account for this possibility, wouldn't they want to make December tests with past December experimentals?

No, it doesn't work like that. A 160 is a 160 is a 160, no matter when it's earned. The same goes for a 150, and so on. And even if people do score a few points lower in February than in October, they still hand out a full range of scores in every single administration. (Also, they can offer the same experimental a bajillion times, because not everybody gets the same one.) So they will always have data on how anybody of any given score level will perform on any given question. And if they don't, they just offer the experimental again until they do have a full range.

Put another way, if the February test-takers score 4 points lower on average, but a 160 is a 160 regardless, then they'll have fewer 160's to get data from, but they'll still have them. Likewise for all other test dates and all other scores.

[clarified_butter]

Maybe February...but we will never know...because it's not disclosed

[EarlCat]

I'll take Useless Things LSAT Students Waste Their Time On for $500, Alex.

[3v3ryth1ng]

I'll take Useless Things LSAT Students Waste Their Time On for $500, Alex.

Congratulations, Mr. Connery! You've picked the Daily Double:

"This word refers to the waste of time between birth and death, encompassing, among other useless charades of meaning, the LSAT, marriage, and showering each morning."

[kimchiman87]

This is a great thread. I especially am intrigued by the theory of uniformly administering experimentals in a given month... Eg using only Dec exp to equate future December tests... I highly doubt lsac is dense enough to allow for that kind of variability in the test. After all dec test takers are probably the last minute types who would not prepare thoughtfully anyways. But this would explain why dec is statistically more lenient than june or oct... good theory 3v3ryth1ing!

[KevinP]

Experimentals most likely come from a combinational of all the months; they definitely don't come only from the same month. For example, Jewett was an experimental in September of 2009 and then showed up as the real deal in December of 2010.

[clarified_butter]

I'll take Useless Things LSAT Students Waste Their Time On for $500, Alex.

I have to agree with you on this.

[Jeffort]

I don't know why these threads make me so mad, but they do. They do.

I'm with you SA. These threads filled with misconceptions and false assumptions bother me too.

The frequent use of the phrase "the curve" is similar to hearing fingernails scratching a chalkboard. The LSAT is NOT graded on a curve.

Dudes, it's a scaled equated test so that each scaled score, no matter which administration or test-form you take represents a particular ability/performance level that is comparable to all achieved scores over a three year time span no matter which test-form or administration you take.

The easiest LSAT is the one you take when you are properly prepared to do your best.

/end thread

[EarlCat]

I'll take Useless Things LSAT Students Waste Their Time On for $500, Alex.

Congratulations, Mr. Connery! You've picked the Daily Double:

I daily double on your mother.