Is there a general way to find zugzwang positions in a game (not chess) or does it require deep and specific game knowledge? I think it should be relatively easy when the game allows draw states in repeated moves - since I don't need to reconstruct the whole game, just the zugzwang position, I can arbitrarily make many positions a draw state. Do you have any experience with this?

You could create an end game database for the game. From there finding zugzwang positions should be straightforward. You're looking for positions where all legal moves end up as a loss if it's your turn to move, but don't if it's the opponents turn.

Quote: Original post by Adam_42
You could create an end game database for the game. From there finding zugzwang positions should be straightforward. You're looking for positions where all legal moves end up as a loss if it's your turn to move, but don't if it's the opponents turn.

Unfortunately the game is a divergent one, so the number of endgame positions is enormous.

Zugzwang can be revealed by looking farther ahead using any of the minimax variants, for if its really a zugzwang then the immenent loss is forced.

Some games may have obvious situations that a static eval can pick up.

In chess, a qsearch will certainly show (but not confidently) the position as bad for the zugzwanged.

I am not sure what the formal definition of zugzwang is, but naive attempts like "a position where the score if it is my turn is lower than if it is not" won't work. For instance, see this example from Wikipedia. The only definition I can think of that would detect the position after 85. Bf5. as zugzwang is "a position where the score improves if we extend the rules to allow passing". One should be careful about what we do when two passes are performed in sequence, but one could just declare that a draw, or something simple like that.