Reputation: 3023
The riffle shuffle algorithm is too mechanical and predictable. How would we add some randomness as if a real human was shuffling the cards but still using the riffle method?
Upvotes: 2
Views: 1252
Reputation: 17346
In the paper Shuffling Cards and Stopping Times by Aldous and Diaconis (from the American Mathematical Monthly, 93:5, pp 333-348) the authors looked at how many shuffles are required to randomise a deck of cards. Section 4 of that paper is An Analysis of Riffle Shuffles and they discuss some equivalent formulations for (non-perfect) riffling of cards.
In particular, they quote a model for random riffling shuffling that simulates how "real" people riffle shuffle. Essentially, the deck is split in to two pieces according to a binomial distribution, leaving c
cards in one hand and n-c
in the other. The cards are then "dropped from a given hand with probability proportional to packet size".
So a simple application will be to assume that your deck is originally split in to two equal pieces and then repeatedly select from which piece the next card comes, with a / (a+b)
probability of it coming from the piece of size a
and b / (a+b)
probability of the piece of current size b
.
Upvotes: 4