Hamming distance computation
I am thinking that Hamming Distance is a good starting point, but may have to be modified by issues of kinesiology such as "it's harder two swap two fingers (i.e. pick one up and put the other down) that are next to each other vs. on different hands". The condition where two fingers always move together (I think it's called Focal Hand Dystonia) may be an extreme example of this issue.
I wrote a quick YAML (Yet Another Markup Language?) parser in perl to check for errors and coverage, but wound up manipulating my experimental mappings in Excel and (gasp) pencil and paper.
One decision I am thinking for a designer of a finger mapping is:
"Should the finger mapping cover all the possible fingerings, or leave some NOP (no-operation) fingerings that would be ignored (i.e. play the most recent pitch)".
I chose to leave some NOP fingerings because
(a) I suspected that additional fingerings might take up storage (does not see to be the case, based on the description of @join), and
(b) the 15 NOP fingerings in my YAML spec (out of 64 non-pinky fingerings) do seem to me to be "mistakes" and are probably best left at the most recent pitch.