It looks like it's the same idea as static Huffman without the variable length part. Comparing it to UTF-8 is pointless, of course - UTF-8 is an universal text representation (whose major selling point is it's also ASCII-compatible and human-readable), and this is pretty much a compression algorithm. It is obvious natural language based text is redundant, so no wonder it's compressible and I am sure a lot of literature exists on which compression works best on which kinds of texts.
