It would be interesting to know the distribution of incorrect answers for the problems. Another interesting thing would be how many incorrect guesses people had for each puzzle (that is, how many people had >5 guesses and did not solve it and such up to how many solved it firs time).