The optimized-out game which resulted after my last rumination on three-person games wound up being quite complicated and non-intuitive. Since that game has two distinct things which can happen in each round, it makes sense to remove the weirder one of them, which results in the following game:
A spinner picks one of the three players to be the selecter. That player picks one of the two other players to get a point. The selecter obviously knows who did the selecting and who got the point. The non-selected player is told this information as well. However, the player who got the point isn't told which other player selected them. Play is repeated for a number of rounds.
The scoring can then be restated in a way which leaves play mostly unchanged as follows:
A spinner picks one of the three players to be the selecter. The selecter picks one of the other two players to be the victim, and the victim loses a point (scores in this game go down, but they never go up). The selecter and victim know who did the selecting, but the other player is left in the dark.
Come to think of it, this randomness and being left in the dark doesn't add much to the game. We can get things back to being zero sum as follows:
Each round each player gets one point, then each player points to one of the other two players. Each player then loses one point for every player pointing at them.
This game has has the interesting situation in which two players are pointing at each other and a question of how to get out of that. But it also has the unfortunate simple equilibrium where each player points to the player on their right, then they each point to the player on their left, then right, etc. Perhaps there is something to the slightly randomized version.
March 24 2005, 18:00:34 UTC 7 years ago
March 28 2005, 02:15:40 UTC 7 years ago
March 25 2005, 03:38:31 UTC 7 years ago
Isn't this then isomorphic to one of the versions you mentioned earlier, before introducing the spinner to make it so there would be only one distinct class of outcome? I guess maybe not, since the new two outcomes are (0,0,0) and (-1,0,1), whereas before they were (0,0,0) and either (-2,1,1) or (2,-1,-1), I forget which. I guess those three pretty much cover all the possible variants of this game, though you can also arbitrarily vary the assignment of the different scores to the players (i.e. for the outcome (a,b,c) the person pointed at by {0,1,2} people can get {a,b,c} points). Then again, perhaps (3,-1,-2) would lead to a distinct game from the above. Hm.
March 28 2005, 02:21:16 UTC 7 years ago
If you get rid of the who knows what weirdness, which I think is a good idea, then it's a lot like a game I gave earlier but with a spinner rather than simultaneous play. I think the spinner actually simplifies things and gets rid of a lot of the weirdnesses though.
The frustrating thing about this game is the 'hurt the other player who most recently hurt me' strategy tends to work very well. It's unclear whether something more sophisticated could win in a field of different strategies.