As a famous physicist once said, everything is relative.
All grades are relative. The only question is, relative
to what?
All grades are on a curve of some kind. Anybody who
doesn't realize this isn't paying attention.
Yet whenever I tell students this, they look stunned.
They think I'm from Mars. They "know" that 90% is
an A, 80% is a B, and 70% is a C ... it's the zeroth
commandment.
But let's get real. It would be the easiest thing in
the world to design a test where everybody in the class
got a perfect score:
True or false: p = m v
True or false: p obeys a local conservation law
Et cetera.
This test would fulfill some objectives but not fulfill
others.
-- It would be great for building up the confidence
of students who needed that: "Look, Ma, I got 100% on
the physics test!"
-- It would not be very informative; it would not tell
the teacher or the students very much about how well
the students were progressing.
Similarly it would be easy (using time pressure
and/or otherwise) to devise a test where the typical
student would get an arbitrarily low score. The Putnam
is well known for its challenging questions and Draconian
grading: http://www.maa.org/awards/putnam.html
A third option would be to try to make the test maximally
informative. This would involve a mean near 50% and
the largest possible variance. This would, of course,
not maximally fulfill the confidence-building objective.
A fourth option would be to decide how many As you want
to give out, and how many Bs et cetera, and design the
test so that 90% corresponds to an A, et cetera. With
enough experience and skill, you can do this "open loop",
but more commonly it will require some _a posteriori_
adjustment factors.
A fifth option would be to adopt some "standard" for
how much progress is expected, and test accordingly.
Even this contains a great deal of arbitrariness;
I assure you that the standards for grading freshman
physics classes change a lot as you move from MIT
to the University of Southern North Dakota at Hoople.
Finally: With a modicum of effort, you could design
a pair of tests such that the score on one was
anti-correlated with the score on the other.
=========================
On a statewide or national test, they commonly report
the percentiles and leave it at that. This of course
comes with its own set of advantages and disadvantages.
The alternative is for somebody (possibly the teacher,
possibly some governing board or accreditation body,
or some self-appointed testing service) to establish
a standard and hold people to it.
========================
In my role on various committees, I've seen lots of
transcripts. It's super-obvious that its all a game,
and there is no consensus about how the game is played.
It varies from institution to institution, and varies
from class to class within an institution.
Suggestion #1: Find out the *local* tradition and
don't deviate too far from it. In particular, if you
were to grade harshly compared to the local tradition,
some people would avoid your courses, because they
don't want to "ruin their GPA" (whatever that means).
Suggestion #2: Tell kids not to take grades too
seriously. The important thing is to learn the
material. Some kids are motivated by grades, some
aren't. To the extent that grades help motivate
people, fine -- but grades are not and end in themselves.
There are lots of things that optimize grades at the
expense of really knowing the material. This includes
cheating, plus lots of things that aren't cheating but
aren't very wise, such as cramming for tests. Tell
kids that if they pull such stunts, "the truth will
out" sooner or later, usually sooner, such as when
they take the next course in the series, or when they
go to a job interview.
Suggestion #3: The school should have a fairly
liberal policy allowing electives to be taken pass/fail.
That way an astronomer can take a serious music-theory
course, and a musician can take a serious astronomy
course, without fear of "ruining their GPA".
Suggestion #4: It's not always possible, but when
possible it is nice if the course is "goal oriented".
The goal might be to prepare students for the AP or
the IB or the GRE, or just to prepare them for the
next course in the series. If there is a goal set
by some outside authority, it tends to drive the
class and the teacher together: "let's work together
to meet this challenge". In contrast, if the teacher
is also the tester, it tends to set up an undesirable
adversarial situation: the class has to "overcome"
the teacher. You *must not* let such an adversarial
situation develop.
Related point: if you are designing the grading scheme,
you might get in trouble for giving As to everybody
in the class. But if you are teaching an AP-prep
course and everybody gets a 5 on the AP, that's no
trouble at all!