Dan Crowe and Michael Edmiston make some good points. But here are
some other factors to consider:
(1) The fact that interim grades fluctuate doesn't concern me too
much provided those interim grades that I officially report are
collected at the right point in the cycle of flucutation, "right"
being defined as representative of the other interim and final
grades. In other words, I need to schedule tests at the right time
relative to when official interim grades are posted.
(2) I damp out some of the fluctuations by having more frequent
tests. This past semester I had 5 (not counting the final exam), ie.
about one every 3 weeks covering about 3 chapters. I think this is a
good testing pace for an intro survey course, the primary
disadvantage obviously (!!) being the grading load.
(3) If you really think homework doesn't represent student work
accurately, count it for fewer points. My numbers were just examples.
(I also think time-driven randomized online homework systems such as
WebAssign help somewhat both in that different students get different
numerical values and in that I don't have to grade them so I can
assign a lot more of them: 45 sets this past semester.)
(4) You didn't really address my primary concern with the
average-weighted system of grades: A student can bomb the first test,
get their act together ("get religion") and do much better on the
next test. In your system, their grade is going to be low at the
first interim report if there's only been one test by that point in
time. In my system, homework and labs can "prop up" a poor first test
and yet tests overall still end counting for the most weight as the
course progresses.