When learning a strategy in a game, we barely focus on what’s possible (that’d be done outside of the game), but rather learn gradually context manifolds given (more common) game interactions.