Can someone explain the LSAT curve Forum

[SJU2010]

My understanding of the LSAT grading system is that its based on how many you get right however, I keep reading and hearing of the "LSAT curve" and I am absolutely clueless to how the curve on this test works. I asked my Kaplan teacher about it but she said no such curve exists so now I am bit confused. Can anyone clarify? Should I contact LSAC about it?

THANKS!!

[Eruannon]

My understanding of the LSAT grading system is that its based on how many you get right however, I keep reading and hearing of the "LSAT curve" and I am absolutely clueless to how the curve on this test works. I asked my Kaplan teacher about it but she said no such curve exists so now I am bit confused. Can anyone clarify? Should I contact LSAC about it?

THANKS!!

It is most definitely scaled. Hence the raw score to 180 score conversion charts that are released every year. It doesn't change very much year to year however. On these boards the scale changes are referred to by how many you could get wrong that year and still get a 170, believe for the Sept09 LSAT that was -11... not sure though.

[pjo]

is the curve by year posted anywhere?

[Eruannon]

Just google or forum search for the raw score conversion chart for the specific test you are looking for.

[SJU2010]

Just google or forum search for the raw score conversion chart for the specific test you are looking for.

You are right I just looked over the Kaplan conversion charts for each of the tests and the curve is quite significant. On some tests as much as 4-5 points. How do they calculate it? Based on how people score on that specific test?

Is there a strategy to getting an x amount right and wrong that bumps your score?


Thanks

[rayiner]

Just google or forum search for the raw score conversion chart for the specific test you are looking for.

You are right I just looked over the Kaplan conversion charts for each of the tests and the curve is quite significant. On some tests as much as 4-5 points. How do they calculate it? Based on how people score on that specific test?

Is there a strategy to getting an x amount right and wrong that bumps your score?


Thanks

"The curve" is simply a mapping from number correct to a 120-180 score. The curve allows the results of different exams with different difficulties to be compared against each other.

The curve is determined based on the results of the experimental sections. For any given 3 year interval, the LSAC has target percentiles for each numeric score (eg: 97.5%-ile for a 170). It looks at how many correct answers people get on an exam when it is given as an experimental section, and then sets the curve accordingly.

[chewdak]

no

[Big Dog]

The use of the word "curve" is really a misnomer. The test makers know going in what the score distribution will be if a gazillion folks take the test on the same day. That's why each test has experimental sections -- to prescreen a question and its 'hardness'. Thus, the test maker knows the near certainty what % of testers will miss that question. A test with a bunch of "hard" questions, will appear to have an easier "curve." And vice versa.

btw: on a rare occasion, a question may get thrown out at the last minute. For example, question on xx passes by the experimental prescreens, but when fully implemented is missed by 99% of the testers in Seattle. Or, by females in NY. Or....

Similar to the SAT/ACT, LSAT raw scores are scaled, not "curved".

[SJU2010]

The use of the word "curve" is really a misnomer. The test makers know going in what the score distribution will be if a gazillion folks take the test on the same day. That's why each test has experimental sections -- to prescreen a question and its 'hardness'. Thus, the test maker knows the near certainty what % of testers will miss that question. A test with a bunch of "hard" questions, will appear to have an easier "curve." And vice versa.

btw: on a rare occasion, a question may get thrown out at the last minute. For example, question on xx passes by the experimental prescreens, but when fully implemented is missed by 99% of the testers in Seattle. Or, by females in NY. Or....

Similar to the SAT/ACT, LSAT raw scores are scaled, not "curved".

That clarifies it A LOT thank you!

[rapstar]

This is how it works:

You take the 100 or 101 question test and the number you get right is your raw score. Your raw score also represents what percentile you place in. For example, if you answer 90 correctly that is your raw score and you may have done better than 98% of test takers. LSAC takes this raw score percentile and coverts it to a score ranging from 120 to 180. Any particular score, such as 170 represents basically the exact same percentile from the last 3 or 4 tests. So, depending on how hard your test is and how well other test takers do determines how many you can miss to make that 170. For example if you only get 86 questions right but this is better than 98% of everyone who took your test then you will get a 170. A 170 from june or sept or dec or feb all represent basically a 98% on that test. (People often confuse this and think that previous tests determine the curve, when in reality only your test determines the curve. Previous tests only ensure that a cetain score equates to a certain percentile. For example a 171 is always 98%. If 2% of the test takers get 14 or fewer wrong then that will be the curve. It's not like the curve is predetermined and sometimes 10% of test takers miss less than 14 and LSAC still gives them all a 170. A 171 or so means that you got 98% on your test, and someone with a 171 from a different month also got 98% on their test. The tests range in difficulty so you can miss more or less to still beat 98%)

[coyote]

What does it mean when people say that the December LSAT had a -14 curve?

[09042014]

What does it mean when people say that the December LSAT had a -14 curve?

That the December test was hard.

[coyote]

What does it mean when people say that the December LSAT had a -14 curve?

That the December test was hard.

But why -14?

[barnum]

I feel like I have explained this before, but here it goes again.

As one poster mentioned it is not really a "curve" but a scaled score. A curve would mean that your score is increased or decreased depending on how others who took the same exam did. This is not the case. The scale is pre-determined before you ever walk into the test. It would have to be in order to be standardized.

Based on experimental sections and a ton of psychometrics, LSAC is able to determine the relative difficulty of the exam they create and determine a scale that balances that difficulty against past exams. A harder test has a more forgiving scale and an easier exam has a less forgiving scale.

The percentiles are not tied to the raw score (the number of questions a tester gets correct), but to the scaled score (120-180). The percentiles are based on the last three years worth of test-takers. Meaning that if 172 is the 99th percentile, then in the last three years only 1% of the testing population has managed to score a 172 or higher.

[rapstar]

-14 means you can miss 14 and make a 170.

The curve is not predetermined. By your very own logic a 172 is always 99% regardless of when you took the test, so if it was predetermined that the curve was going to be -10 for a 172 then it is very possible that more than 1% of the test takers would get fewer than 10 wrong, and therefore a 172 would no longer represent 99%. (in fact if 3% got 10 wrong it would represent 97% which on another test 97% would be a 170) The only thing that is predetermined is what scaled score corresponds to what percentile. i.e. it is predetermined that 172 will represent 99% on the june test, dec test, feb, test, etc. LSAC has an idea of how difficult the test is but there is no way to predict exactly how the raw scores will be dispersed.

[09042014]

-14 means you can miss 14 and make a 170.

The curve is not predetermined. By your very own logic a 172 is always 99% regardless of when you took the test, so if it was predetermined that the curve was going to be -10 for a 172 then it is very possible that more than 1% of the test takers would get fewer than 10 wrong, and therefore a 172 would no longer represent 99%. (in fact if 3% got 10 wrong it would represent 97% which on another test 97% would be a 170) The only thing that is predetermined is what scaled score corresponds to what percentile. i.e. it is predetermined that 172 will represent 99% on the june test, dec test, feb, test, etc. LSAC has an idea of how difficult the test is but there is no way to predict exactly how the raw scores will be dispersed.

You've got that wrong. The scale is predetermined, and the percentiles do change test to test. 172 is not defined as 99%.

[rapstar]

edited

[rapstar]

a 172 represents something. it represents that you did better than a certain percentage of test takers. your claim would mean that a 172 is sometimes better or worse than another 172. that makes no sense. a 172 from june is just the same as a 172 from september. you did better than 99% of test takers. (it may be 99.1 vs 98.9 depending on the month but thats only because it is based on a discrete number wrong. not a continous scale where you can miss 9.19372 questions. otherwise the percentiles would be exactly the same month to month. lsac will obviously give all raw scores from one test month the same scaled score which may cause your exact percentile to vary just slightly from test to test)

[Woozy]

a 172 represents something. it represents that you did better than a certain percentage of test takers. your claim would mean that a 172 is sometimes better or worse than another 172. that makes no sense. a 172 from june is just the same as a 172 from september. you did better than 99% of test takers. (it may be 99.1 vs 98.9 depending on the month but thats only because it is based on a discrete number wrong. not a continous scale where you can miss 9.19372 questions. otherwise the percentiles would be exactly the same month to month. lsac will obviously give all raw scores from one test month the same scaled score which may cause your exact percentile to vary just slightly from test to test)

You are right in your first sentence, rapstar, but the rest is wrong. 172 represents a certain level of LSAT performance. It is a number which is standardized across different forms of the test so that scores from different administrations may be compared with each other. The LSAT is not graded on a curve. Theoretically, 100% of the takers in a single month could score 172 or above. The percentiles are simply calculated by looking at the score distributions over the 3 years up to and including any given administration. They in no way determine the scores allocated during an administration.