Sometimes a submission manages to pass tests by means of a cheat. For example, just returning true for a Boolean function, scoring 50% of the test cases (or more if the tests are skewed). One could say that the tests are bad in that case, but one would still want to override the grades from testing since the submission does not reflect a correct solution of the assignment. For such cases there should be a grade override as part of the regular grading procedure. The override grade feature achieves the same effect, but is not integrated in the checklist grading process and is intended for exceptional cases.

Submitted by Eelco Visser on 16 March 2015 at 17:32

This should be implemented as additional field in the checklist UI while grading. It’s a penalty field, which requires motivating the penalty. The penalty gets subtracted from the overall grade, not from the checklist part. The UI should explain this.

Currently, a solution which fails many tests but is done in the right style gets full points from the checklist. Here is an example from the A&D exam. Students should implement a queue with two stacks. When they implement the queue differently, the grade should be 1.0. This could be achieved with a penalty. But there is also an efficient and a less efficient solution. Could the inefficient solution result in another penalty? Will it be a fixed deduction, or proportional to the number of tests the submission passes, e.g. a 8.0 for an inefficient solution passing all tests, but a 4.0 for an inefficient solution passing half of the tests?

Note that this is intended to allow applying an ad hoc penalty to a submission that does something that is not foreseen in either the tests or the rubric.

I would expect that you can express the scenarios you describe either with tests (timing) or with the rubric; only check the box for the efficient implementation if it is indeed the efficient implementation.