A question about program logic... program provided

[Tim Rentsch]

[discussing shuffling using a PRNG with only 30 bits of state]

Using that algorithm with a PRNG that has only 30 bits of state
can generate no more than 2**30 distinct shuffles, which is smaller
than the set of all possible orderings by a factor of about 7.5e58.
[snip]

I remember reading an article on just this point recently (though
forget where!)

If your random number generator has only a 32-bit seed, then the
number of possible shuffles is very low compared to the actual
possibilities, such that it is theoretically possible to deduce
the the whole pack after seeing only the first six or seven
cards. [snip elaboration]

Sure, if the algorithm is known. If the algorithm is not known,
and all that is known is the shuffles are produced using 32
bits of internal state, it's easy to get tens of millions
of shuffles without getting any useful information about
which permutation will come up next, even half way through
one.