Rarest I think is pretty easy. It's just board state transition.
You'd probably need 15 orders more magnitude of data to get there, but the definition itself is pretty straight forward.
You'd probably need 15 orders more magnitude of data to get there, but the definition itself is pretty straight forward.
EDIT: so you could define "rare" moves as the biggest difference of occurrences between state N and state N+1.
Computing the number of permutations is thoroughly unconvincing.
For instance, there's 20 possible first moves and of those only probably 2 are played 95% of the time. You can certainly compute what the rates open is and the rarest response that's actually played or the rarest response to the most common open
This would include not only the board state, but all the moves made up to reaching it.
It has a useful meaning for openings.
But it is so sparsely populated by actual games that "rarest" becomes too easy - many moves have been made only once under this definition.
https://en.m.wikipedia.org/wiki/Perfect_information
It might be interesting, but the previous board state should not be a relevant factor when considering rarest move