I thought the same thing, but it turns out the test is not all multiple choice---in the 2006 exam, 28 out of 58 problems were open answer, and the rest were 4-option multiple choice. So the expected value of a totally random selection would be 7.5 questions out of 58---meaning this guy was at least not doing worse than chance.
Getting 10 questions right may not be statistically more significant than just answering randomly. I would calculate this myself, but I'm going to bed, however, just eyeballing it makes me think this is quite possible.
