Menu
▾
▴
Re: [B1cc-contact] B1cc-contact post from anatha@cs.technion.ac.il requires approval
|
From: FloF <bla...@gm...> - 2009-01-05 12:54:34
|
Well, I guess you found a bug... This shouldn't happen normally, but since this project was only a master's project, we stopped working on it and did only a few tests on the different algorithms/heuristics combinations. So it may be that for certain configurations, there are some endless loops. Notice also that minimax and alphabeta are strictly equivalent in terms of game power, minimax is just far slower than alphabeta, so minimax is only useful for theoretical complexity considerations (ie show that it's slower ;) ). The important things for having better AI's are the search depth and the heuristic used to evaluate game positions. A statistical comparison has been made in the downloadable report pdf. best regards, Florent foucaud. 2009/1/5 Anat Hashavit <an...@cs...> > Hello, > Thanks for the useful information. I have one more question. > I run one test of minimax against alphabeta and the game did not finish. > Both players converged to two rows outside their camp but did not enter. > Does this problem sounds familiar to you? > Thanks, > Anat > > > > FloF wrote: > >> Hello, >> >> From: Anat Hashavit <an...@cs... >> <mailto:an...@cs...>> >> To: b1c...@li... >> <mailto:b1c...@li...> >> >> Date: Thu, 01 Jan 2009 16:19:01 +0200 >> Subject: Chinese checkers questions >> Hello, >> >> I am using you're Chinese checkers program and I have a few >> questions: >> >> 1 - Is there support for N players? >> >> >> Yes : you can choose the number of players in the "new game" dialog : 2, >> 3, 4 or 6 players. The players may be human or artificial, as you want. >> >> >> 2 - About the simple distance heuristic , it is a bit unclear , do >> you have some kind of formula you can send me? >> >> >> This is detailed in the report, which is unfortunately only in french. >> >> The simple distance heuristic is an evaluation function for the >> alpha-beta/minimax algorithm. >> It simply computes the sum of the distances of the opponent's pawns to his >> goal, minus the sum of the distances of the current player's pawns to his >> goal. >> >> Hence, the more the current player's pawns have moved towards their goal, >> the higher the computed value will be. And the more the opponent's pawns >> have moved towards their goal, the smaller the value will be. >> >> This expresses quite well the idea of moving your own pawns while blocking >> the opponent (remembering that the higher the value computed by an >> heuristic, the better should be the game situation). >> >> We called that heuristic "simple" because one can then add some more >> strategies, like favoring grouped pawn moving, covering the middle of the >> board, etc. >> >> Thanks, >> >> Anat >> >> >> You're welcome, I hope this is more clear now. >> >> Best regards, >> Florent Foucaud >> >> > |
