Oh wow. Oh wow oh wow oh wow! This is exactly the kind of thing that I love doing. So, what you have stumbled upon here is a type of "Markov Chain", where you take percentages for each letter or number or whatever you are generating and than generate new stuff. @codinghorror, the co creator of discourse, has a great blog post on them:
Now, here is something cool to think about:
You have the percentages for each letter, and you will generate random words or at least things that look like words with this, but what if instead of taking the percentage that each letter appears in the English language, you take the percentage that each letter appears in Shakespeare's plays? Would you be able to generate Shakespearean words and even sentences/paragraphs?
If you look up 'Markov chain percentages' or 'Markov Chain transitional matrices' I'm sure you can find some more percentages for letters that you can experiment with. The only thing is, in most chains, the percentages are how often a follow b or c follow d, not just how often they appear. But that's besides the point. Experiment with this. Have fun with this. These types of experiments are exactly the kind of stuff I love and will take you to awesome places.