That is not IMHO what he is trying to say, you don't shift the distribution, you measure if somebody passed a test. I the test is "passable" then one side of "distribution" is at least cut off. E.g. it's normal (and sometimes expected) that the whole class will pass without issues.