Unorthodox chess
Nov. 26th, 2012 09:49 pm![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
davvYesterday, I thought about how to apply minimax - the algorithm used in chess AIs - to realtime or near-realtime games, and I came up with this chess variant that can only realistically be played on a computer.
First of all, before the game starts, define a speed by setting one time unit equal to some period of time. It might be a second, a minute, more, or less. Then the rules are:
After moving a piece, it can not be moved for a certain number of time units equal to the time value of that piece. The time value of a pawn is 1 TU, of a knight and bishop, 3 TUs, of a rook, 5 TUs, a queen, 9 TUs, and a king, 4 TUs. These are basically the material values of the pieces.
If a piece is moved, it can't be moved again until enough time has passed, but other pieces of the same sort are not affected. Moving one bishop does not freeze the other bishop.
At the beginning of the game, each piece is given an initial random time penalty of between 0 and its time value[1], to deter unfair wins by people with the quickest fingers[2]. Alternately, one might decide in advance upon the timeouts. In any event, all black pieces get an additional TU to wait before they can be moved the first time[3].
Kings cannot be captured. Check and checkmate rules apply even when the king or the opponent pieces are out of turn: a player loses at the immediate point his king is checkmated, and can't move any other pieces while his king is in check.
If white and black each makes a move within 1/20 TU of each other, and the moves can't both be carried out, the tie is broken in favor of white. For instance, if both players are trying to move a knight to the same square within that time, then white gets the move. Black's knight is not frozen after this - because it never actually moved - but can immediately move elsewhere.
All other rules are as in ordinary chess.
[1] either fractionally or rounded to nearest TU. Fractionally means the random value for a pawn before its first turn might be, for instance, 0.35 TUs. If a TU is a long time, that may make sense - if it's not, it doesn't make much of a difference.
[2] Or in other terms, the person with the greatest "APM" shouldn't win just because he's got the greatest APM.
[3] Although with the time-based nature of this variant, one could also imagine a variant upon the variant where black is not penalized in this manner.
Is such a variant interesting at all? What would the strategy look like? And how would one make an AI to play this variant? I think I know the answer to the last question, but chess is really a complex game and so actually implementing it... well, that's another matter. And the theorist in me says: "the rest is merely an engineering problem"... even as the hacker is really curious what the game would look like.
First of all, before the game starts, define a speed by setting one time unit equal to some period of time. It might be a second, a minute, more, or less. Then the rules are:
After moving a piece, it can not be moved for a certain number of time units equal to the time value of that piece. The time value of a pawn is 1 TU, of a knight and bishop, 3 TUs, of a rook, 5 TUs, a queen, 9 TUs, and a king, 4 TUs. These are basically the material values of the pieces.
If a piece is moved, it can't be moved again until enough time has passed, but other pieces of the same sort are not affected. Moving one bishop does not freeze the other bishop.
At the beginning of the game, each piece is given an initial random time penalty of between 0 and its time value[1], to deter unfair wins by people with the quickest fingers[2]. Alternately, one might decide in advance upon the timeouts. In any event, all black pieces get an additional TU to wait before they can be moved the first time[3].
Kings cannot be captured. Check and checkmate rules apply even when the king or the opponent pieces are out of turn: a player loses at the immediate point his king is checkmated, and can't move any other pieces while his king is in check.
If white and black each makes a move within 1/20 TU of each other, and the moves can't both be carried out, the tie is broken in favor of white. For instance, if both players are trying to move a knight to the same square within that time, then white gets the move. Black's knight is not frozen after this - because it never actually moved - but can immediately move elsewhere.
All other rules are as in ordinary chess.
[1] either fractionally or rounded to nearest TU. Fractionally means the random value for a pawn before its first turn might be, for instance, 0.35 TUs. If a TU is a long time, that may make sense - if it's not, it doesn't make much of a difference.
[2] Or in other terms, the person with the greatest "APM" shouldn't win just because he's got the greatest APM.
[3] Although with the time-based nature of this variant, one could also imagine a variant upon the variant where black is not penalized in this manner.
Is such a variant interesting at all? What would the strategy look like? And how would one make an AI to play this variant? I think I know the answer to the last question, but chess is really a complex game and so actually implementing it... well, that's another matter. And the theorist in me says: "the rest is merely an engineering problem"... even as the hacker is really curious what the game would look like.