| Author |
Message |
Karl Scherer (Karl)
| | Posted on Wednesday, March 21, 2001 - 8:25 pm: | |
Hi Jeff,
I understand that the 3-move-draw property
is turned off for puzzles.
This often leads to nonsense moves and games when I want to test the game by letting the computer take the players's role.
I suggest that you introduce a menu item
where the user can switch the 3-draw-move rule
off himself.
(Maybe (?) then there is not need anymore
to have it permanently turned off for puzzles.)
----------------------------
Furthermore, for all games (not ony puzzles)
it would be often of advantage to have a
draw-avoidance factor, as in chess
(as another menu item).
This would help to find a solution to a puzzle
to a win where without it the computer would
always be happy with a draw.
This would also avoid the tendency to
finish a game early with a 3-move draw,
hence not going for a win, especially in those
games/puzzles where any move is better
and more fun for the opponent than running into
a repetition that stops the game.
If it works for chess-programs, why not have it for Zillions? |
Dan Troyka (Dtroyka)
| | Posted on Sunday, March 25, 2001 - 9:39 am: | |
In two-player games you can define repetition as a loss-condition, which should approximate the behavior you are looking for. Such coding is ignored in one-player games. It seems that at a minimum Zillions should enable the code (win-condition (Self) repetition) in a puzzle so that developers can choose whether repetition is permitted.
It would also be helpful if repetition were addressed in the context of partial moves. In principle Zillions evaluates repetition at the conclusion of a player's turn, so that repetitive partial moves within the turn can loop eternally. A human player in fact can do this, but in practice I have found that the AI, when repetition is defined as a loss-condition, is less likely to make endless partial moves even when the opponent has a winning move on the next turn (but it is inconsistent). I would recommend evaluating repetition after each partial move so that repetition within a turn can be defined as a loss.
Also, a (repetition true/false) statement would be helpful with the default being true for two-player games and false for puzzles. Currently two-player games must end upon three-move repetition (either as a draw, win, or loss) and there is no way around this. Some flexibility would be nice. |
Jeff Mallett (Jeffm)
| | Posted on Wednesday, June 13, 2001 - 7:47 pm: | |
There seems to be a slight misconception that 3-move-repetition detection is totally turned off for puzzles. The truth is that it is turned off in the user-interface for puzzles, but is still counted in the search. This makes it so that a human player is allowed repeat positions as much as they want while trying to find a solution to the puzzle, while a computer opponent will avoid repetitions as pointless. |
|