There is no algorithm that, on average, will successfully compress ANY data set, even if it leaves uncompressible data sets uncompressed. You can find a beginning of a proof in my article "How much is big data compressible? An interesting theorem" at