Check out Chess with Different Armies, our featured variant for July, 2024.

Alice Chess

Invented by V.R. Parton in 1953

Alice Chess was invented by V. R. Parton in 1953. It is one of the top 20 played chess variants in NOST, and David Pritchard in his Encyclopedia of Chess Variants recommends it as one of the best variants. Personally, I think it is one of the few variants that has that 'neato' quality to it; one of the most fun to play variants.

The standard game of Alice Chess is played using two boards, A and B. All pieces move as in standard chess. The normal array is on board A; board B starts empty.

The rules are very simple. In turn, each player makes a single move on either board following these three rules:

  1. A move must be legal on the board where it is played.

  2. A piece can only move or capture if the corresponding destination square on the other board is vacant.

  3. After moving, the piece is transfered to the corresponding square on the other board.

Since a square and it's corresponding square on the other board are never both occupied, Alice Chess can actually be played on a single board. Checkers are placed under pieces that are on board B. I don't know if this makes moves easier or harder to visualise.

The starting setup:

There are less played variations of Alice Chess. One variation is to simply have all Black's pieces start on board B. Another, Ms. Alice Chess, allows 'zero' moves. To make a zero move, simply transfer a piece to the same square on the other board. Kings may not zero while in check.

Parton also suggested a smaller version, played on two 4x8 halves of a single board:

Clarifying the Rules

Written by Fergus Duniho

Parton describes Alice Chess in Curiouser And Curiouser and in Chessery For Duffer And Master. Both descriptions are the same, and in neither one does he mention castling or en passant. In the Zillions Rules Files for Alice Chess by Jens Markmann and by Pierre Tourigny, each has stipulated that castling and en passant are not part of the game. Furthermore, on the Alice Chess page of the British Chess Variant Society, George Jellis has written, "En passant capture is abolished. The inventor of Alice Chess, V. R. Parton, said nothing about the rule for en passant capture, but since the rule for orthodox chess can be interpreted in at least two different ways (does the capturing pawn have to be on the first or second board?) and is subverted by the fact that the square passed over may be occupied on the other board, it is usual to forgo it." He doesn't mention anything about castling. In one of the Alice Chess games listed by David Pritchard in his Encyclopedia of Chess Variants, one of the players castles.

Given that Parton himself didn't state that castling and en passant are not part of the game, I think it is fair to assume they are. They should be discarded only if the rules of Alice Chess somehow make them impossible, but in fact they do not. Let me begin by describing how castling would work in Alice Chess. All the usual conditions for castling will apply, plus Alice Chess introduces a new condition. When a King and Rook castle, the spaces they move to should both correspond to empty spaces on the opposite board. This is because a piece may move only to a space whose corresponding space is unoccupied. Also, once the castling is complete, both pieces transfer to the corresponding spaces on the opposite board.

Jellis mentions some details about en passant that I also thought of while working on my own Zillions Rules File for Alice Chess. First is the question of whether the capturing Pawn has to be on the first or second board. As I understand en passant, it allows a Pawn to capture a Pawn it could have captured if that Pawn had made a one-step move instead of a double move. Thus, the Pawn that can take another by en passant must be the one that could have taken the Pawn if it had moved only one space. This means a Pawn on the second board. When a Pawn makes a double move, it will switch boards, and if it lands beside an enemy Pawn on the other board, that Pawn will normally be able to take it by en passant. But Alice Chess does introduce one situation in which the rule of en passant becomes ambiguous. When a Pawn makes a double move, it may pass over a space whose corresponding space on the other board is occupied. Thus, the space the enemy Pawn would have to go to for an en passant capture will be occupied.

There are three options here. One is to allow both a regular capture and an en passant capture when it's a piece the Pawn can capture. Another is to allow only a regular capture. And the third is to abolish en passant altogether. Since double captures never happen in Chess, I don't favor the first option. The usual assumption in Chess is that a Pawn making an en passant capture moves to an empty space. If this is a rule of Chess, it never has to be enforced, because the available space always is empty. Besides this, if the piece occupying the space is the same color as the Pawn that could otherwise move there by en passant, en passant will still be impossible. The option of abolishing en passant seems too extreme. As in Chess, Pawns can still block each other in Alice Chess, and the original purpose behind en passant was to help blocked Pawns get unblocked. So it makes sense to include en passant as far as it is possible to do so. I favor the middle option, in which a Pawn may capture by en passant any enemy Pawn that lands beside it while making a double move -- so long as the space corresponding to the one it passed over is also empty. It does this by moving to the empty space behind the Pawn that just moved. If that space is not empty, then en passant is impossible, but so long as it is empty, en passant works.

Another ambiguity in the rules concerns whether a piece may block a check by moving from the board the King is not on. The first rule states, "A move must be legal on the board where it is played." Since this move would leave the King in check before the transfer, it might be thought that it would be illegal. After all, it is illegal to leave the King in check. But since the King is in check on a different board, it doesn't matter. The clause "on the board where it is played" narrows the domain of legality to a single board. Leaving a King in check would count toward this move being illegal only if the checked King was on the same board as the piece was moving from. Logically, this interpretation of the rules makes more sense, because it is non-recursive. It is defining the legality of moves in Alice Chess in terms of what would be legal in Chess, as played on a single board. The other interpretation, that a piece may not move from the other board to block a check because that leaves the King in check until the piece transfers to the King's board, is circular. It is defining legality in Alice Chess in terms of what is legal in Alice Chess. But you can't define something in terms of itself. It makes more sense to define something in a non-recursive, non-circular way.

Furthermore, this interpretation is the consensus among those who have programmed Alice Chess. This interpretation is accepted by Fergus Duniho, Jens Markmann, Pierre Tourigny, Ed Friedlander, and H. G. Muller. It is also accepted by Antoine Fourrière, writing on Alice Chess strategy down below. And it is reflected by a comment that shows up in a sample game published in David Pritchard's Encyclopedia of Chess Variants. Here is the game transcribed for Game Courier. The relevant comment is included after Black's 22nd move. It mentions a threat that could win the White Queen, and the only way this would be possible is if the White Queen could block a check by transferring to the King's board after it moves.


Written by Antoine Fourrière.

Although two corresponding squares are never occupied, Alice Chess looks more like a 128-square, eight-color game than a 64-square, two-color game. Kings are much more at risk than in Orthochess. Theoretically, they still have up to eight moves, but Queens, Rooks and Bishops enjoy a freer ride on the Board and have more opportunities to cut their retreat. It is also more difficult to protect a King by interposition: the interposed piece must come from the other chessboard, and is never protected by the King. Thus, Queens are juggernauts which may keep the number of moves into the single digits (see the sample games below). On the other hand, Knights and Pawns are rather weak. Like Bishops, Knights are restricted to half the field (from a 128-square perspective), and would be colorbound if one chessboard had the colors of its squares reversed. (Let's say Knights are N-colorbound and Bishops are B-colorbound.) Non-capturing Pawns also commute from a dark square to a light square and vice versa exactly as they commute from one chessboard to the other (that is, Pawns are N-colorbound as long as they do not capture), but if we should color the even-numbered rows of one chessboard like the odd-numbered rows of the other and vice-versa, the Pawns would remain colorbound (let's say P-colorbound) after completing their initial move, regardless of their initial position, which is of the same P-color for a given side. Anyway, Zillions, which values the Knight at 3.2 Pawns, the Bishop at 5.4 Pawns, the Rook at 8.2 Pawns and the Queen at 13.2 Pawns probably offers a conservative estimate.

Since each move starts on one chessboard and ends on the other, a piece can be guarded only by a piece standing on the other chessboard (except against the King). It follows that the guarding piece no longer provides protection to the guarded piece whenever it has to move to escape capture. After the exchange, the guarding piece will take the place of the captured piece. Hence, two Bishops cannot defend each other, which is no big news, but neither can two Knights, nor two Pawns which haven't undergone the same initial one-square or two-square advance. Maintaining a chain of adjacent Pawns of the same B-color on different chessboards, though even more desirable than in Orthochess in view of the aggravated rider threat, requires only one-move advances or only two-move advances, the latter of which feels impractical. As a result, Pawn promotion is usually longer (six moves), and Pawn-by-Pawn capture, which is permissible only between Pawns of different P-color, remains frequent in the endgame. (On the other hand, advancing a Pawn by two squares could ensure the absence of capture from enemy Pawns.) If Bishop-by-Bishop capture is permissible only between Bishops of the same B-color (B-color is unaffected by the change of chessboards), Knight-by-Knight capture is permissible only between Knights of different N-color, that is, Knights which started on the same file. As for Knight vs. Pawn, if we restrict our analysis to Pawns which haven't undergone an initial two-square move, a b Knight (that is, a Knight which started on b1 or b8) is able to defend a,c,e,g Pawns (that is, Pawns which stand currently on one of these files) and to capture enemy a,c,e,g Pawns, and is threatened by b,d,f,h Pawns. (Then again, it may be reason enough for advancing a Pawn by two squares.)

Regarding the opening, 1.d3/B seems a reasonably aggressive first move for White. Other good openings are 1.e3/B, 1.Nc3/B and 1.Nf3/B more or less as in Orthochess (c3/B and f3/B are the positions of choice for the Knights as long as most riders haven't been exchanged), but also 1.a3/B and 1.h3/B which open lines for a Rook and prevent the symmetrical Pawn from advancing. Like other two-square advances, 1.d4/B and 1.e4/B are too speculative, while 1.b3/B and 1.g3/B permanently weaken the Knights. 1.c3/B followed by a quick c4/A is unclear and 1.f3/B looks awful. Black, though unable to answer 1...d6/B to 1.d3/B, 1...a6/B to 1.a3/B or 1...h6/B to 1.h3/B, has the other corresponding choices at his disposal. It should be noted that after 1.d3/B 1...e6/B, 2.Qxd7/B loses the Queen on 2...Nc6/B 3.d4/A 3...Bd6/B, followed by a later Nf6/B. (Capturing at once a Rook's Pawn with the Rook begets similar punishment.)

In the endgame, there is less difference from Orthochess. The most noticeable fact seems to be that when a King wants to stop a Pawn on its own from promoting, the square rule holds true only if the King and the Pawn aren't on the same chessboard. Apart from that, since a King may never reach a square that is threatened on either chessboard, King and Rook, King and two Bishops, or King, Bishop and Knight still give mate to a bare King, and King and two Knights still result in stalemate before mating occurs. However, there is a difference of flavor in the executions. For instance, in King and Rook vs. King, White starts as usual by repelling the Black King on a row, which has the effect of N-colorbinding the King. White then places his Rook in such a way that the King isn't allowed to capture it, and advances the Rook until the King is restricted to two squares, say Rf7/A on one chessboard and the enemy King commuting between g8/B and h8/A. White then has plenty of time to place his King on g6 or h6 on either chessboard when Black's King is on g8/B, followed by Kh8/A Rf5/B Kg8/B Kh6(g6) Kh8/A Rf8/A mate. Likewise, the other mates require some twists and turns to avoid stalemate.

Sample Games

Alessandro Castelli sent in three short games of Alice Chess.


1.e4/B e5/B 2.Qh5/B g6/B 3.Q:e5/A mate (Castelli-Pugnali, Grand Prix 1993, 1-0)


1.d4/B e6/B 2.Bg5/B Be7/B 3.Bf4/A Bb4/A mate (Kustrin-Castelli, Grand Prix 1995, 0-1)


1.e4/B d6/B 2.Bc4/B Q:d2/B 3.Bb5/A mate (Castelli-Jelliss, Grand Prix 1995, 1-0)

Play It!

You can play Alice Chess online with Game Courier:

There are three different Zillions Rules Files for playing Alice Chess with Zillions of Games. Each is by a different author, and each handles the mechanics of game play very differently.

The Jens Markmann ZRF is available from Jens Markmann's Art of Games website. This one lets you play on two boards, moving a piece directly from one board to the other. It does not implement castling or en passant. It includes both Alice Chess and Mini Alice Chess with both the usual setup and the mirror setup. It includes two ZRFs, each using a different piece set. One uses the default Zillions set, and the other uses pictures of Alice characters.

The Pierre Tourigny ZRF uses the method of playing Alice Chess on a single board with two sets of pieces. It includes a modified version of the default Zillions set for this purpose. It does not implement en passant or castling. It includes the variants Alice Chess, Alice Mirror, Alice Zero (aka Ms. Alice Chess), Alice Grand, and Alice Extinction.

The Fergus Duniho ZRF is the most recent, and it is the only one that requires Zillions 2.0. Like Markmann's ZRF, this one lets you play Alice Chess on two boards, but unlike that one, it lets you move pieces to spaces on the same board. It will then automatically move the piece to the other board. I find this more user-friendly than the methods used by the other two ZRFs. It includes multiple piece sets, including the special set used by Jens Markmann. It includes both Alice Chess and Ms. Alice Chess with both the usual setup and the mirror setup. Unlike earlier ZRFs for Alice Chess, it implements both castling and en passant.

Version without graphics.
Written by Edward Jackman and Fergus Duniho.
WWW created: 1995 or 1996.