Grading

Let's think about grades for a minute. The percentage of the total points on a test, quiz, project, etc. seems to be what everyone thinks they should give their students for a grade. But this carries a lot of assumptions, which IMHO make little or no sense. Percentage grading assumes:

• You can write a test or judge another type evaluation so that a grade of ,say, 80% means the same thing each and every time, and that the minimum "passing" grade is that much better than one percent less than that grade.
• Each test has the same difficulty, so that 80% always means the same thing through the year.
• That the teacher across the hall teaching the same course, will give percentages that mean the same thing as yours, and the teacher in the next school district, the next county, the next state ...

Then what happens? All your grades for a particular test are low. Does this mean the test was too difficult or that the students did not learn all they should? What to do? Some folks "scale" or "curve" the these tests (but not the ones where everyone does well), or they make sure the next test is "easy." Then of course you have to justify scaling only the "hard" tests and not the ones where everyone does well. (This in fact assumes that the test was too hard -- else why reward kids who didn't learn what they should have -- and proves you can't judge the difficulty of your own tests.)

Percentage grading only seems to work for very experienced teachers who use the same tests from year to year and have adjusted them so the grades come out the way they would like.

There are various ways to scale or curve tests.

• One that has been discussed recently is the "square root curve" where the grade is 10 times the square root of the percentage. This raises everyone's grade; those in the middle a lot, those on either end a little. (Max/min question: which percentage gets raised the most? Which the least?). I can see even less justification for this approach -- except to raise everyone's grade.
• Other schemes are to list the grades in order and look for natural breaks. Highest group gets A or 90s, next get B or 80 etc. Somewhat arbitrary, but may make more sense than anything above.
• Find the z-scores and force the grades into a normal distribution -- this assumes your kids are normally distributed and in an AP class that may not be the case. Besides with this approach you will always have to fails some one even if they are all pretty good students.
• Throw the test up the stairs ...

So what to do? After all you do have to assign some grade. My suggestion is to read "Assessing True Academic Success" by Dan Kennedy. A perfect approach, probably not. But it will give you some things to think about and suggest a curving scheme that you might like and that has some thinking behind it. 

The take-away from this article is "You control the grading algorithm."