Personally, I'd have expected "round to nearest" to be the standard pretty much everywhere, as that's what's typically meant when speaking about rounding. Rather than a single issue/decision, I'd suggest that this is really the interplay of two issues:
- Lack of clarity in the "decimal places" setting(s) - is it rounding or truncating?
- Order of operations - are letter grades calculated before or after the rounding (or truncation)?
Simply changing the rounding/truncation strategy still seems to leave room for confusion, as letter grade boundaries can themselves contain decimal places (which may not have been the case when MDL-21746 was first opened, based on the early comments). Consider the following scenario:
- The lower boundary for an A is 89.5%
- Student 1 scores 89.4%
- Student 2 scores 89.6%
If we simply round down (truncate) to 0 decimal places, both will see a score of 89%, but Student 1 gets a B while Student 2 gets an A.
Perhaps we should add a new "letter grade strategy" setting (subject to better naming), with options such as:
- Round, then apply letter grade
- Round, but apply letter grade to raw score
- Truncate, then apply letter grade
- Truncate, but apply letter grade to raw score
As I understand it, the second option is the current Moodle behaviour, so should probably be the default (so that there are no unexpected changes upon upgrade).
There's also the completely non-technical solution: rather than actually changing how it works, simply make the current behaviour clearer by flagging grades which have been rounded to/over letter grade boundaries in some way. This could either mean flagging them to teachers, so that they can choose to either adjust the affected grades as desired, or flagging them to everyone who can see the grade in question (so that students can also see that their grade was just below the boundary). Flagging to students needn't reveal the raw grade (though it probably should for teachers, so they can see just how close the student came) - simply pointing out that the displayed grade has been rounded up should hopefully make it clear why there's a discrepancy, though. This could even be combined with the technical solution, so that if a "letter grade strategy" which applies the letters based on raw grade is chosen, there's still an indication as to why there's a discrepancy.