Hint – Paper, Scissors, Rock Home -> Paper, Scissors, Rock

We know that optimal play should be a random sequence of Paper, Scissors, or Rock. We know that that people are bad at being random number generators. How can we measure how bad people are?

If PSR are really chosen with equal frequency, what is the probability that you would see the sequence RRRRR, i.e. 5 choices of Rock in a row, compared with some other sequence like RSPSR, where the string is more balanced?

Well, it may be surprising but those should be equally likely occurrences. People are not likely to properly balance those long strings. They know that they should keep the number of Paper, Rock, and Scissors about equal, but they tend to be too balanced and don’t make sure that regular clumps like RRRRR show up just as often as ones that seem more mixed up.

It is fairly easy to keep track of all the history that the user has typed. It is likewise easy to look at the frequency with which the user enters all possible di-grams, or tri-grams, or higher order chains. The question is: how do you exploit the non-even patterns that you see the person generating, to make the computer more likely to win a hand against the human?