>I was assuming that calibration of difficulty implied some sort of curve. No, it just means "last year's exam seems to have been a bit too difficult, so I'll

>I was assuming that calibration of difficulty implied some sort of curve. No, it just means "last year's exam seems to have been a bit too difficult, so I'll make this year's a bit easier" (or vice versa). Curving means that you assign grades based on a student's performance relative to the others in the class rather than on their absolute score on the test. >but I don't really understand why difficulty would be a useful metric, If you really don't understand how it's possible for an exam to be too easy or too difficult, then I'm not sure what to say. The exam is intended to differentiate between students who've mastered the material, students who understand the basics, and students who learned nothing. To do that it has to be difficult enough that people who skipped half the classes† can't get full marks, but also easy enough that not everyone fails it. Even these basic boundary conditions on a good exam can be quite tricky to satisfy in some cases. †Not that I care about people skipping classes per se. I just mean that typically they'll have significant gaps in their knowledge of the material.