Hello everyone, I took the June 2019 SAT and a lot of test takers have been saying that the curve will be harsh. Do you guys think the college board will pull the same scaling as the June 2018 sat and kill our scores?
Is it curved? Can you explain how it works?
Jenn
The College Board says “We use a process that adjusts for slight differences in difficulty between various versions of the test (such as versions taken on different days).” Basically, if the test is easier at a certain sitting, your score will suffer more by missing fewer questions. If the test is harder, you can miss more and achieve a higher score. Something like that!
Don’t go by the rumor mill on Reddit. It’s just an echo chamber of anxiety and complaints.
Overall, I will say this: the “harsh curves” only matter if you are trying to score in the 700s (especially the mid to upper 700s) on a given section. After that point, curves are pretty constant from one test to the next. Now, even in the lower 700s, the curve isn’t likely to affect anyone that much. A harsh curve is generally indicative of an easier test. The main problem with a harsh curve is at the tippy top of the score range. The problem here is that if you make a couple careless mistakes it can be a really big deal. Lower in the 700s, the problem is not likely careless mistakes (if you are making 6 careless mistakes, that will affect you on any test). The problem is you got some questions wrong that you didn’t know how to do.
If you look at the analyses I provided below, you will see that the curves are not consistently different from what they have been previously (only one is abnormally harsh). Rather, there are a variety of curves. This is why it is always worthwhile to take the SAT multiple times–you are more likely to hit a favorable curve.
There is also an assumption on here that tougher curves are bad. This is true if you know the material through and through. I have some students, though, who perform better on the tests with harsher curves because the tests are easier.
Also, the ACT isn’t more generous than the SAT when it comes to curving. The problems on the ACT are generally easier than those on the SAT and the test is curved accordingly.
Here are what the curves were like on the recent tests for which there are score charts.
March 2019:
-math slightly harsher than normal; 10 wrong was 670 (normally expect 690-700), 1 wrong was 790, 2 wrong was 770, 4 wrong was 740; not overly harsh as far as I am concerned, probably just reflective of easier test
-reading was completely normal and typical, nothing overly harsh to report
-writing was maybe a touch harsher than normal, there was a jump from 30 points off for 3 wrong, to 50 points off for four wrong, nothing too unusual though, again probably reflective of a slightly easier test
October 2018:
-math definitively harsher, 10 wrong=660, still relatively fine at top, 1 wrong=790, 2 wrong=770, 3 wrong=750 (20 point jumps happen, the weird part is having 2 in a row)
-reading was a slippery slope for the top 40 points (albeit not abnormally so), after that pretty typical
-writing was very precipitous for the top 60 points, 3 consecutive 20 point gaps at top
-May 2018
-overall similar to March 2019
-very generous at tippy top, though, 1 wrong on math=800, 2 wrong on math=790, 1 wrong on reading=0 points off, 3 wrong on reading still only 10 points off, 6 wrong on reading only 30 points off, 2 wrong on writing only 10 points off
March 2018 was pretty typical, a bit generous on math and reading, decidedly ungenerous on writing.
Hope this helps.
Before every sat is administered they do a test run and compare it to other sat tests. They then scale the test BEFORE it is administered and the easier the test, the harsher the scaling. So on a hard test, 1 wrong might still be an 800, but on an easier test, 1 wrong will take you down to a 780 or for the June 2018 SAT, to around a 770(from what I’ve read/heard).
DS20 has taken Dec 2018 test, 7 wrong in writing 31, 2 wrong in reading 38. Reading/writing 690. 7 wrong in math 680. I think those are all very harsh.
D17 took Jun 2016 SAT, 2 wrong on reading, nothing wrong on writing 750.
D17 said she decided to take Jun 2016test (new SAT available March 2016) as historically June curve is better, not sure if that is the case in the new SAT. Thoughts or comments?
I think it’s important to understand that a student who finishes with a raw score in the top x% will receive a score of Y, regardless of the number of questions correct or incorrect.
Statements that “Z questions wrong resulted in a much lower score” have an implicit assumption that students would have had the same number of questions correct/wrong on every test. It just doesn’t work that way. If the same number of wrong questions resulted in a lower score, then the test was easier and that student fell further down in relation to other students taking that test. There is not a fixed raw:scaled score correlation. here is a percentile:scaled score correlation.
Take any SAT test. Add 10 IMO-level math questions to it. Top students will get 12 wrong instead of 2 wrong. Middle students will get 19 wrong instead of 9 wrong. They performed “worse”. And everyone will get the exact same score. Because they performed the same relative to other test takers.
Add 10 basic addition problems. Top students now get 68 of 70 rather than 58 of 60. Middle students get 61 of 70 rather than 51 of out of 60. They performed “better”. And they will get the exact same score.
It would certainly be easier to have every test be exactly the same difficulty and every raw score would always result in exactly the same scaled score. It just doesn’t work that way. They run questions in experimental sections to attempt to calibrate each test. But test calibration is not an exact science.
(And the reality is that it’s more complicated than this - it’s not just others on the specific test date, it’s the entire universe of tests taken over time. So the “take in when the Talent Search 7th graders take it so you’ll score better” advice doesn’t hold either.
College Board has published technical white papers on their process if you really want to understand it and are up for concepts like chained equipercentile linking and r-biserial correlation)
@TheSATTeacher what about the June 2018 curve? I had read that was pretty harsh. Your opinion on that?
The scoring chart for that wasn’t released (to the best of my knowledge) so I can’t say anything definitive about it in particular. I think my point stands: there are some more harshly curved tests now, but there are also plenty that aren’t. Take the test several times and you’ll be fine.
Personally, not losing extra points for a wrong answer has inflated scores compared to the old days.
Outrage about curves is misguided. If the curve is harsh, that means the test was easy, so a test with a more generous curve would almost certainly result in more wrong answers. It’s a pretty effective system, if annoying at times.