I am an instructor that teaches math problem-solving courses where raw exam scores are naturally low (e.g. a 50% might be a "C" performance and 75% might be an "A"). There is a common grade-curving policy of rescaling the grades linearly, so if a student's raw score is n, then their scaled score is f(n)=an+b, where a and b are chosen parameters. In other words, the raw score is multiplied by one number, then multiplied by another.

I use this to map scores on challenging assignments into a range of numbers that is appropriate for the percentage-based letter grade system. For example, if raw scores range from 20/40 (50%) to 36/40 (90%), but the 20/40 is really a "C-" performance not an "F", I would choose a and b to map the raw scores to the range [72%, 100%].

The existing bell-curve feature does not suffice for this. The bell curve puts students in competition with one another, and I cannot control how the minimum/maximum score map to percentages.