Buy that special someone an AP Physics prep book, now with 180 five-minute quizzes aligned with the exam: 5 Steps to a 5 AP Physics 1

Visit Burrito Girl's handmade ceramics shop, The Muddy Rabbit: Yarn bowls, tea sets, dinner ware...

27 July 2010

Changing AP score cutoffs now that the guessing penalty is gone

In early June, the College Board released the news that, starting next year, the raw score for an AP multiple choice section will be the total number of correct questions, without a "guessing penalty."  You first heard it here, on the Jacobs Physics blog.

One of my major recommendations to AP physics teachers is to make ALL classroom tests in AP format.  This means use authentic AP questions, give the appropriate amount of time for each (one minute per point free response, 7 multiple choice every 9 minutes)... and use a scale converter to translate each student's grade into an approximate AP score.

In past Physics B exams, it's taken roughly 65% of the available points to get a 5, 50% to get a 4, and 35% or so to get a 3.  That exact conversion fluxuates year-to-year depending on the difficulty of that year's exam, but this was a good rule of thumb.  Problem is, now that the subtraction of 1/4 point for each wrong multiple choice question is gone, raw scores will be higher.  How will these numbers change?

The calculation is not quite as simple as it initially might look.  Scores of the weaker students will improve more than those of stronger students, because weaker students will miss more questions and thus fail to be penalized for "guessing" more often.  And free response scores will not go up, only multiple choice.  This all points to adding a bit to each score cutoff, with more being added to the lower cutoffs.

This year, I'm going to use these cutoffs in my class.  I'll see if they work... and I'll appreciate feedback, as well:

5  68%
4  55%
3  41%
2  30%

We'll only know more through experience, and when another released exam comes out.

GCJ

6 comments:

  1. Great post - thanks! Only place online I found this being discussed. Currently, in order to account for the new scoring method, I'm using the percentage cutoffs found in College Board's re-release of the 2004 scoring worksheet -- they mirror your cutoffs fairly closely. From the worksheet:

    5: 68%
    4: 55%
    3: 40%
    2: 33%

    What I don't know is how difficult, relatively, the 2004 exam was compared to other years. 2009 seems like it was a difficult exam -- here are the cutoffs from that re-released scoring worksheet (from the horse's mouth, taking into account the lack of guessing penalty):

    5: 63%
    4: 48%
    3: 32%
    2: 23%

    Those numbers aere even lower than the curoffs before the scoring change! Must have been a tough one. I wish I had the time to get the data together to compare the various administrations in terms of difficuly (score distribution by percentage correct for MC/FR...instead of by 1-5).

    Thanks for the post! Wish there was more out there from teacher's who give only AP questions on tests...and how they determine grades.

    ReplyDelete
  2. I have been using your score cutoffs this year. Have you been happy with them, or have you made any changes?

    ReplyDelete
  3. I've been happy with them. Guys who should be getting 5s are getting 5s. In the end, it doesn't matter much if you're off a few points here and there, but I think what I've suggested is close enough for a reasonable prediction.

    ReplyDelete
  4. what's full score?
    is it 180 still in 2011?
    or it has been changed?

    ReplyDelete
  5. Full score on the real AP should still be prorated to 180 -- one point per minute.

    ReplyDelete
  6. I use all AP exams in my class, and your cutoffs correlate pretty well to those that I use. If anything, I cut a little higher so that 70% and up are 5's - not terribly different. Anyway, last year (2011), all the students that earned A's in my class also earned 5's on the exam. The B's and 4's and C's and 3's all correlated as well. I did have three C's that earned 2's, but all the rest (about 20 in all) were very well matched. These work for me!

    ReplyDelete