I think this would be more interesting if there was a way to explicitly declare the marks post-curve. In the event the distribution of marks is smoothed to any

I think this would be more interesting if there was a way to explicitly declare the marks post-curve. In the event the distribution of marks is smoothed to any sort of typically college distribution it's entirely rational to just select 6 points and not care whether 10% of the others do or not. Selecting 2 leaves you open to a worse result on a curve, but selecting 6 is nearly risk free if the distribution of marks is precisely controlled anyway. 2 marks when the average pre-curve goes up more than 2 and the average needs to be fixed post-curve is a really crappy proposition.