Suppose one has a deck of cards. If one reshuffles it using a discrete probability model (based on combinatorics) one will still have the same probability of getting a particular card on each draw.

This seems to be a little bit naïve for real cards, since reshuffling is somewhat non-random. Suppose you know the ordering of the cards in the deck before reshuffling.

Is it possible to tackle non-randomness of reshuffling to get conditional (on original ordering) probabilities of getting a particular card from the deck?


1 Answer 1


There is an extensive mathematical theory of shuffling. For example, one well-studied shuffle is the random riffle shuffle, which is the inverse of partitioning the cards randomly into two piles, then stacking those piles.

From one perspective, this is a random walk on the symmetric group to get a random permutation $\pi$. $\pi^{-1}(1)$ is a distribution on $\lbrace 1, 2, ..., n \rbrace$, the probability that each card ends up first. If you apply $\pi$ $k$ times then the distribution of the first card is $\pi^{-k}(1)$.

If you start with a random permutation $\nu$ and then apply $\pi^k$, this produces the random permutation $\pi^k \nu$, and the distribution on the first card is $(\pi^k \nu)^{-1}(1) = \nu^{-1} \pi^{-k}(1)$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.