It might be useful to have some method on the UniversalDetector class that populates the confidence of every encoding for a given string. This can maybe help out figuring why we are failing on those 25 other test cases and also help if any encoding is to be added / deleted. Thoughts?