Testing strategy

My colleague and I collaborated on a mid-term exam, which ended up being pretty enormous (five pages, free response). Some of the questions were very straight-forward (what's the slope of the line between these two points? What's the y-intercept of the equation y=3x-2), but many more were multi-part, rich questions.

As the day went on, I realized that the majority of students weren't going to come anywhere near to finishing this beast of a test. So, as I was grading the 2nd period tests, I developed a points system (different questions, or parts of questions, worth different numbers of points), such that the entire test ended up having 400 points. That done, I gave the last class a rubric (including the points-value of each question): 100 points to get a D, 125 for a C, 160 for a B, and 200 for an A (yes, a non-linear function, which tops out at 50%).

Seems to be pretty well calibrated: there's a LOT of different content, and a LOT MORE different presentation of similar content, on this test. By providing an unrealistic number of opportunities to demonstrate what a kid knows, and indicating that I don't expect anyone to finish the entire thing, I ended up providing options to students, so that they can really show me what they know well (by choosing the problems that "look easy", whether that's the visualization, or the abstract, or the verbal, and focusing their best efforts on that...) To get an A, a kid will have to correctly cover at least some ground in each modality (so, it connects up to the Rule of 4...)

In all, I'm pretty intrigued by the happy accident-- and I'm definitely looking forward to designing my next test in a more straightforward way, to work on the same principles.