If you're looking for a non-scientific but warm fuzzy approach, I would generate about, oh, 100 files with n runs, with each character representing 8 runs. (So you get an ascii character). Then, I would make a copy of the file and gzip it. Compare the sizes of the orignal and new file, and the better job the compression does, the less randomness it is... (This approach was sort of discussed a long time ago in perl monks about checking for randomness of passwords...)