In some sense betrayal is required to win, but that doesn’t necessarily mean deception is required to win. For example, you could make deals of the form “If you do A, I’ll betray you with probability X”, and simply outplay everyone by making good bilateral trades. I don’t know anything about the game, but maybe with sophisticated enough players, deception would actually be limited as players figure out that the optimal strategy is to never lie in the early-midgame and never believe other players in the endgame, or something.
I’d like to see the game played by better agents, but it would be unclear whether any change in deception rates or type is about the propensities of agents vs the metagame differing by capability level.
In some sense betrayal is required to win, but that doesn’t necessarily mean deception is required to win. For example, you could make deals of the form “If you do A, I’ll betray you with probability X”, and simply outplay everyone by making good bilateral trades. I don’t know anything about the game, but maybe with sophisticated enough players, deception would actually be limited as players figure out that the optimal strategy is to never lie in the early-midgame and never believe other players in the endgame, or something.
I’d like to see the game played by better agents, but it would be unclear whether any change in deception rates or type is about the propensities of agents vs the metagame differing by capability level.