"The final result is a normal distribution with a peak at 21 hits aand a maximum of 40 hits. Since you got 33 hits, you're doing significantly better than chance."

I think Roger means this: assigning Don's predictions random dates instead of the dates Don actually specified gives 21 hits, on average, although once in 10,000 tries, the random choice yielded 40 hits. Don's score of 33 is three sigma better than the random choice's average performance, that is, the random score only achieved a score of 33 a few dozen times out of 10,000.

However, since Don probably chose his times based on when unusual levels of seismic activity were happening, it remains possible that Don's score were above average because he was essentially predicting aftershocks, as well as the more obvious possibilities that he was just lucky or he has discovered a new phenomenon.