Our biggest improvement came when we adapted CryptoSolver to seed its initial population with solutions it found to big words. For instance, a word like "statistics", is pretty much unique in the English language and only one permutation can match this. So we added the capability when the population is initialized to search if there are any big words with a small number of solutions, and then to seed our initial population with these individuals. The results helped dramatically and allowed us to solve quotes that before were too hard. Raising our solution set from 5 to 8. See Figure 1.
Figure 1 |