The "arcane magic" impression likely comes from the heavy use of information theory concepts and maths that are encountered in the traditional literature. Probably necessary to analyse in detail, but quite daunting for the beginner. I find LZ to be particularly intuitive since it embodies the natural notion of "abbreviation" by referring to something we've seen before, and writing compressors/decompressors for the algorithm is relatively easy (especially decompressors.)
On the other hand Huffman is somewhat less easy to grasp intuitively, but some aspects of its principle of reducing bits/symbol based on their frequency shows up in natural languages: the more common a word is, the shorter it tends to be. It's also a bit more difficult to implement.
On the other hand Huffman is somewhat less easy to grasp intuitively, but some aspects of its principle of reducing bits/symbol based on their frequency shows up in natural languages: the more common a word is, the shorter it tends to be. It's also a bit more difficult to implement.