lichess.org
Donate

Game lost when the rules say it's a draw

@Jackstraw67 said in #10:
> you had the pawn , just premove the pawn down the board.

I don't know the game, but maybe it was blocked? And even if it made it to queen, it would still be a loss.
That's a completely different case than what was discussed here. This seems to be one of the extremely rare examples where the software is not smart enough to recognize that the game should be a draw according to the rules. Actually, if my understanding of the rules is correct, the game should have been in fact declared draw right after the previous move (apparently exd5+) even without a timeout.
I believe that it's actually impossible for software to be written which *always* detects fortress positions like the one in #12 above. You can write it to detect it in a specific type of position, such as a pawn endgame with no possibility for either king to penetrate as in that example, and maybe to expand that by detecting that a bishop on the board is unable to capture any of the pawns either due to all the pawns being on the other colour or due to the bishop being locked in, but covering all possible fortress possibilities is a big problem.

Simple forward searching analysis is obviously not sufficient because you cannot conclude that just because no way through has yet been detected it necessarily cannot be found with yet more forward searching. It's a bit like the so-called Halting Problem, the problem of detecting in all general cases whether an input program terminates, which is well known to be non-computable.