| Author |
Message |
No One (Singularity)
| | Posted on Sunday, March 19, 2006 - 11:30 pm: | |
A few quick questions:
1. I am making a chess variant with over 50 types of pieces so far. When right-clicking to select pieces to drop, it is becoming increasingly tedious to scroll through all the choices because they are arranged in a single column. Is there some way of splitting the list into multiple columns or sub-menus?
2. I am making original graphics but have been inspired by the "giant" graphics set that often seems to be used by Marek14. Who is the creator of this graphics set? Do I have permission to create a derivative set?
I believe one of Zillions' main advantages over other programs such as Game Courier is it's ability to provide an AI opponent.
3a. I am wondering what can be done to maximise AI strength for custom chess variants (something involving WinBoard and chess engines perhaps?). The following are some of the strategies I am using. Any criticisms (these ideas could be misguided) or other ideas are welcome.
3b. Zillions doesn't detect "overlaps" when calculating the values of a piece. To try to maximise AI efficiency, I have adopted the policy of "one 'add' per target square" (so I combine Zillions' pawn-capture and en-passant-capture moves into one).
3c. Similarly, I don't think Zillions recognises that some promotion choices are clearly redundant vs others. I am assuming that Zillions will grossly overestimate pawn values if there are many promotion choices, so I try to mitigate this by only allowing promotion to your own pieces that have been captured.
3d. AIs tend to be good at 'brute force' tactics and weak at 'principled' strategy. I am thus under the impression that often, Zillions will have difficulty checkmating powerful King pieces [eg, Pope = King + Cardinal (modified Bishop)], so I intend to make bared King a loss.
3e. However, I feel this advantages the first player, so I would like to make this loss condition negated if the bared player can also bare the opponent on the next move (this motif seems to be reflected in some historical variants?). I am wondering how I could implement this? |
Mats W (Kålroten)
| | Posted on Monday, March 20, 2006 - 1:40 am: | |
Hi No One!
When dropping pieces one doesn't right-click. One points at the square and left-clicks. Right-click is only for editing the board.
En-passant captures are very rare, so it doesn't help to fiddle with them. Moreover, if you have more than one add per target square then a window will pop-up that asks whether you want to do this is that move. This never happens in regular chess, so it's a rare feature. In my latest chess variant, Burmese Chess (Sittuyin) this double add sometimes occur, because, while leaving a square, one can choose whether to promote a pawn or not. But more than one add per square is a rare feature, I suppose.
It doesn't seem like pawns are overestimated in Zillions chess, so I don't think it's a big problem. I don't think it's the number of possible pieces that matters, but only the power of the most powerful piece. But I'm not sure.
A brute force method is ideal for mate-searches. To my knowledge, this "check-mated King" option seems to work fine. I don't see how "bare King" can replace it.
These historical "bare king" rules are redundant. The games end long before this. But the simplest way of "implementing" this is simply to recount the "bare king" rules. If there are two bare kings or only one bare king on the board, the player(s) can decide themselves whether the position is a win or a draw. I don't think you can have the win-condition negated if the next move reduces the opponent to a bare king. It's not necessary to implement this rule. What you can do, of course, is to set two bare kings to draw. This is good enough.
Mats |
No One (Singularity)
| | Posted on Monday, March 20, 2006 - 10:55 am: | |
Ah, sorry, I meant editing the board, rather than dropping pieces.
Unfortunately, even adding the same piece twice to the list of possible promotions increases the estimated value of the pawn. At this rate, the AI soon values pawns more than bishops, etc.
The bare king rules in redundant in most forms of chess, but in fairy chess, where the king may be replaced by pieces such as maharajahs and scorpions, mating becomes much more difficult (without gross numerical superiority), even for king vs king + several powerful pieces. Other fairy pieces try to deal with this problem in other ways (from memory, I think a 'king battler', moves as a queen, but only as a king when checked). |
Mats W (Kålroten)
| | Posted on Tuesday, March 21, 2006 - 12:12 am: | |
To restore a correct pawn evaluation one could punish promotion by capturing, or flipping, an invisible friendly piece beside the board.
Pawn-promotion could be solved in other ways, too. For instance, if the pawn-reaches the last rank. then one makes an "add-partial". This implies that the player must move again. This time he must drop a piece on his own pawn. I'm not certain that this helps.
Anyway, "bare king" is no solution to the problem, because generally the game is lost long before the king is bare. If mating is such an unlikely event, then an alternative solution is to give victory to the party who has acquired a much stronger piece majority. In many historical games the game ends when one party acquires a certain piece majority. In this case one must instead reckon with the power of the pieces. When there is a certain piece power superiority, then the game is won. Generally, one cannot play until the king is bare. That's tedious. |
|