In each step the LZSS algorithm sends either a character or a <position, length> pair. Among these, perhaps character \"e\" appears more frequently than \"x\", and a <position, length> pair of length 3 might be commoner than one of length