But the actual predictive accuracy depends on very few factors - the main significant factor is probably the word-lexing technique used, and this matters mostly for eliminating spurious random strings.