I am assuming a list of numbers (1 - 64) has already been generated in sequential order. The task then is shuffle them randomly. According to the Wikipedia entry I linked to previously, the Fisher-Yates shuffle does this in linear time consuming N Log N bits of entropy. My proposal would have been to perform that same shuffle some number of times where sometimes groups were treated as single elements and sometimes as individual elements. The process was obviously flawed because I wasn't changing the selection criteria for each group.

Setting the flaw aside, my question was how many times you would need to shuffle before you had an acceptable fake and how you would go about telling the difference between a list done with a single Fisher-Yates shuffle having sufficient entropy against the same list down with many Fisher-Yates shuffles (on different scales) without sufficient entropy.

If I have understood the replies from you, blokhead, and others - using a single list wouldn't be statistically sufficient to determine an anomoly. For a single list, leaving the sequential order is just as valid as any other permutation of this list. Again, I don't mind being wrong for the sake of a good discussion.