Earlier quoted context omitted.
What was the result?
If I remember correctly, somewhere in the range 17--23. The way one should have thought of it, with hindsight, is that there are going to be several groups of people: - those who don't understand the problem and just guess a number between 1 and 100, - those who only reason to one degree, and guess 33, - those who only reason to two degrees (ah but if everyone else guesses 33 I should aim for two thirds of that!) - t…
But then you probably still have to account in your strategy for estimating the proportion of people who will reason to infinite degree and estimate 1, as well as the proportion who will reason to an infinite degree and misread the rules, and guess 0...