Not strictly curve grading, but when I was studying electronics one of the professors had a 9-step algorithm for processing raw scores into grades. That involved normalizing the lab scores between different reviewers (so that harshness of a particular reviewer would be compensated for). He also assigned `cos(percentage_from_other_test as radians)` instead of resits if you missed one of the tests for valid reason but had shown up on the other.

On one hand I really like the former approach and could understand the latter (organizing resits is likely a huge PITA). On the other that's the only professor I heard of that was attacked with an axe by one of his students.