Abstract: Based on the analysis of LZ77 algorithm, this paper discusses its major improvements. A new algorithm adopts a circular history list, leaving the look-aheadbuffer of LZ77 discarded. That makes it possible to let a matched string be as long asthe size of the history list and results in the reduction of code length. Another main improvement is that the new algorithm takes variable-length code instead of LZ77’s fixedlength one. Besides, a new data structure called matching position list is introduced,which contributes much to the high efficiency of the encoder. How and why the improvements affect the algorithm’s performance are shown in detail.