Often, a bigger issue in practice than "

Is my card shuffling algorithm statistically random?" is simply "

Is my deck space big enough?", which is basically a test of your random seeding.

*You only have as many shufflings as you have random seeds.* There are 52! unique deck shufflings, which is around 2^{220}. Many random number generators accept only a 32-bit seed, and we note that 2^{220} is a **much** bigger number than 2^{32} (2^{188} times bigger, in fact).

The real world implications of this is that in cases where you need a random shufflin, you may be vulnerable to brute force attacks. In late 1999 it was discovered that three online (live money) poker sites had exactly this problem, and a sample attack program was created that would take the known cards and quickly narrow down the possible decks until the player had perfect information about the cards the other players held and the cards to come.

Obviously, not every program needs 220 bits of randomness in their shuffler. A Texas Hold-Em program with a statistically perfect shuffling algorithm that supported 10 players would need about 120 bits of randomness (the exact number is log base 2 of (52!/27!)).

Anyway, the pont being that when you are evaluating a card shuffling algorithm, the right choice depends on the application, and when you have a deck space that is 52! points long, typical seed limitations are pretty puny.