| Author |
Message |
Ronald W Brierley (Ronaldwb)
| | Posted on Thursday, February 27, 2003 - 10:19 am: | |
I am a new owner of ZOG2 but I have read various notices and articles about the facilities provided by the programming system. I hope to implement a true multi-player game. That is a game where three or more players will take turns to make a move with their own pieces. So far in my search I have only found ZOG games where in effect the players combine to make a pair of combined player groups. For example players numbered one, two, three and four, play in turn but in effect player number one plays followed by player number two. Then player number one plays the pieces of player number three and then player number two plays the pieces of player number four. Hence the program knows of only two actual players but it knows of four sets of pieces. This is not what I want. I want four players playing with their own four sets of pieces. Does anyone know of a ZOG implementation of a game of this type or how it could be implemented? |
David Eugene Whitcher (Dralius)
| | Posted on Thursday, February 27, 2003 - 11:51 am: | |
Dear Ronald
You missed something somewhere. Lets say you have four players red, black, white and blue.
(players Red Black White Blue)
(turn-order Red Black White Blue)
This would have red move once then black move once then white etc...
I hope this is what you are looking for |
Keith Carter (Keithc)
| | Posted on Thursday, February 27, 2003 - 3:26 pm: | |
Hobbes is a game that uses four players.
The problem with games that have more than two actual players is that Zillions plays most of them poorly. In a four player game Zillions, as the moving player, must, as a matter of practicality, treat all of the three players as an enemy. It goes further than Zillions thinking it is outnumbered 3 to 1 by three players with separate and conflicting interests. Zillions seems to treat all enemy pieces as being a single large team. Again this may be a matter of AI practicality.
A multiplayer game like Chinese Checkers works fairly well because there are no captures and enemy pieces may be jumped so that being outnumbered is not as much of an issue.
My assessment of how the Zillions AI handles multiplayer games is limited to my deductions from trying to develop my own four player games. I did not publish them because Zillions played them so poorly. Perhaps someone at Zillions will note this thread and jump in.
treat |
Jeff Mallett (Jeffm)
| | Posted on Thursday, February 27, 2003 - 3:48 pm: | |
I'll just jump in to say that a player's strategy becomes very unobvious when more than two interests are involved. Alliances, backstabbing, psychology,... If it is an optimal strategy to attack the strongest player, then it might also be an optimal strategy to not be the strongest player. You might want to help the weakest player, since it would be hard to overcome the strongest player without his help. If, in looking ahead, you see that a move you can play is bad if two other players cooperate in their play, do you avoid playing it? Or do assume that the the weakest player won't do it because you are not the strongest player? If a player has not harmed you in the past, do you avoid being the first to strike? I don't think there are easy answers. It would be interesting to spend a lot of time implementing these decisions in the AI though. |
Ronald W Brierley (Ronaldwb)
| | Posted on Saturday, June 28, 2003 - 9:01 am: | |
There is a possibility for humans to treat all situations as though they knew that the behaviour of their opponents was dictated by what is assumed to be human motivations. To make good moves in such situations may be much more difficult than it appears. That, I thought was one of the great merits of that popular game Poker. Even if the player is to accept the suspicion that one of his/her opponents holds a certain type of hand he/she does not know how the opponent will react. He/she can only guess, since the opponent may vary his/her techniques, within reason. When playing a game a computer does not need to compute assuming human tendencies but rather to compute with consideration of all the possibilities. If there are too many possibilities to allow for a cleverly optimised tree search technique it may be acceptable to reduce options by using a human guided order of evaluation, known as a heuristic.
It, to a machine, would not matter whether a situation was reached due to collaboration by some of its opponents. Merely the possibility of reaching that situation would be given some heuristic priority and evaluated when the heuristic priority showed the necessity for evaluation. |
Patrick S. Duff (Pduff)
| | Posted on Monday, May 01, 2006 - 2:12 am: | |
I'll offer my go-moku-9 rules file as a test case for investigating how to improve multi-player logic. It supports up to 9 players, but the regular Zillions AI plays poorly with more than two players. Some of its move choices are really odd! |
|