[Retros] A single integer N = a complete chess game

[Seth]

"Otto Janko" <otto at janko.at> wrote:


> An asymptotic optimal algorithm therefore counts the frequency of

> all needed representations in a first pass, calculates the optimal

> representation in a second pass and then encodes the game in a third

> pass.


But then you have to include the encoding dictionary in the
representation, so it isn't clear that this method is that good (since
games may be shorter than dictionaries).

Or you could do that over all games (defined how?  All possible?  Or
somehow weight by probability?) and use that.

Or, in any position, order the possible moves, and give the index of
the move to be represented.


> For example, if the pawn a2 never moves in the game, then it is not

> neceesare to reserve encoding space for pawn a1 moves. (And in many

> games many pieces never move!)


But it's necessary to, at least, list the pieces that don't move (32
bits).

Seth