Comment by jokoon

Isn't possible to achieve a great compression ratio when you recognize grammar forms?

Statistical compression seems a little random and unpredictable.