When you curve a test, how do the scores actually change? You might decide that the top scorer (with 89/100, say) should get 100%, but do you add 11 points to everyone’s score, or change the denominator to be 89? Maybe you want to drop a particularly hard question, but doesn’t that hurt people who got it right?

I was all geared up to explore these questions, but as it happens someone beat me to it!

Rather than reiterate what that article already says very nicely, I made a little toy example. Using made up data, see what happens to the score distribution when you apply different curves.