# Random Card Selection Algorithm Comparison

I'm working on a programming puzzle/challenge for fun, and had a question about whether or not two approaches for doing random selection are equivalent.

Lets say I have a class Deck which contains initially a collection of 52 cards. Then I have a class Shoe which holds a collection of Deck.

Would the following approaches be statistically equivalent?

Approach 1: Select a random Deck in Shoe. Then from selected deck, select a random card.

Approach 2: Pull all the cards from all the Decks in Shoe into a single collection, and randomly select a Card from that collection?

In use, the selection would be done many times, with the selected card being removed from the Deck / Shoe.

• In thinking about this, I'm fairly certain they are not equivalent. If I somehow had a Shoe with 10 Decks, and one of those Deck had exactly one Card, but the rest had say 50 Cards, approach 1 would yield a 1/10 probability of selecting the single card in the small Deck vs approach 2's probability of 1/501. Commented May 7, 2014 at 17:33

As an example consider two $52$-card decks and the probability that the first two cards drawn are both the Ace of Spades:
1. The probability the first card drawn is an Ace of Spades is $\frac1{52}$. The conditional probability that the second card is also an Ace of Spades is $\frac{1}{2} \times 0 +\frac{1}{2} \times \frac{1}{52} = \frac{1}{104}$. So the probability of a pair of Aces of Spades is $\frac{1}{5408}$.
2. The probability the first card drawn is an Ace of Spades is $\frac1{52}$. The conditional probability that the second card is also an Ace of Spades is $\frac{1}{103}$. So the probability of a pair of Aces of Spades is $\frac{1}{5356}$.