i know i was suggesting an alternative, in which the probability of being correct is actually 2/3 rather than 1/2

edit: course we'd have to repeat the experiment if we wanted to do that

edit2: with our current experimental results, as well as looking at LMNO's "is magic 25% better" hypothesis, we could also calculate whether there is a significant statistical difference using a normal distribution (as the mean correct answers would be 50 p=0.5, n<10 so its basically a perfect set-up for that test... i think... nigel?)