I wonder if by using the statistically most used words one could automate at least a big part of the guessing process