Abstract: A system, method and algorithm for performing statistical analysis on an input data source to determine an optimal token size resulting in an increased ratio value is described. Changes to the values of the tokens prior to compression may result in an overall compression increase even if the data source has been previously compressed. Specific techniques for modifying the tokens include varying the base size, rearrangement of the tokens, and modifications by matrix multiplication. After modification of the tokens, compression of the data using one or more lossless compression methods may be performed. The technique may be repeated by using the output of previous compression processes as input to additional entropy and compression optimization.