Fabrice Liardet's site (mostly in French), including three pages of exercices about tactics, openings and endgames, but also featuring this and this games with comments in English, and
John Beasley's comprehensive study of three-man pawnless endings).
Losing Chess, also known as Suicide Chess, Giveaway Chess, Killer Chess or Take-all Chess is called Vinciperdi in Italian and Qui perd gagne in French, both of which mean the loser is the winner or vice versa. The idea of switching loser and winner is recurrent (you can play Losing Backgammon, Losing Othello, Losing Checkers... not to mention the card games which, like the Australian 500 or the Queen of Spades, offer the possibility of playing for reverse goals). But contrary to Chess, where captures are optional, Losing Chess had to borrow a mandatory capture rule from Checkers.
Rules
The rules below are those commonly used for Losing Chess.
The opening setup is as in normal chess. All pieces move as in normal chess (but see below for the King).
Capturing is compulsory. When a player can capture, but has different choices to capture a piece, he may choose which piece to capture.
There is no check or checkmate. The king plays no special role in the game, and can be taken as any other piece.
Pawns may also promote to Kings.
Castling is not allowed.
There is some controversy about stalemate. It is an important question because the richness of Losing Chess stems from the possible endgames. The following three rules are the most popular.
Stalemate is a win for the stalemated player. (International rules)
Stalemate is a draw. (AISE rules)
Stalemate is a win for the player who has fewer pieces left. If both have the same number,
the game is drawn. The type of pieces makes no difference. (FICS rules)
Other Variants
Historic Variants
In 1844, the "Codrus game" (published in the Brede Almanach) is a win for the player whose King is captured.
In 1874, Walter Campbell published the rules of Take me: a player whose piece is to be captured may choose the capturing piece.
Modern Variants
Some variants try to reintroduce checkmate, and decree either that checkmate is a win (on the Internet Chess Club) or a loss for the checkmating player.
More interesting seems Anders Ebenfelt's variant:
When
stalemate, the stalemated player does not move but the opponent can if he wish to play for win go on moving
and do as many moves he wants to do. If the stalemate then disappear, both players move again as usual. So,
if white for example has a pawn on h2 and nothing more and black a pawn on h3, pawn on a7 and rooks on a8
and h8 black can win by moving : 1.-,Rh4 2.-, a5. 3.-, a4 4.- Ra5 5.-,a3 6.-,a2 7.-,a1=R 8.-,Rg1 9.-,Rg3
10.hxg3,h2 11.gxh4,Rg5 12.hxg5,h1=Q 13.g6,Qh7 14.gxh7 and black has won.
This variant has the
advantage that there is often dangerous to let the opponent take all pieces but one if the one left is a
pawn which can be blocked. The game will then be more complicated, or sophisticated (the strategy is not
only to throw away pieces).
Play
General principles
Quite often, a player will suffer an ignominious defeat. One of his pieces will be lured inside the enemy camp and forced to capture the bulk of the enemy army, like a zombie. (The Bishop is particularly at risk, but the Knight and the Rook are also frequent vectors of the disease, and the other pieces aren't completely immune.) Hence, Losing Chess is first and foremost about control (perhaps not unlike Peter Aronson's Anti-King Chess). The players need to maximize their number of options. Not only should they refrain from tossing away their pieces as fast as they can (more material means more opportunities), but they should also develop them a bit (more reachable squares means more opportunities).
Watch Peter Minear translating his material and space superiority and read Richard Lewis' instructive comments.
Another control-related issue is that when a player may capture with two pieces, one of which (and one only) would be immediately recaptured, he should not exchange the second piece.
Fabrice Liardet offers the following example on his mostly French-language site.
White to play
The Knight isn't bothersome, because Black is unable to unprotect the Pawn on a7, but there is a real threat that the Bishop will wipe out the entire Black army.
The natural, but wrong, move is 1.Nxa7??, which could lead to 1...Rxa7 2.Bxc1 Ra3! 3.Bxa3 and an easy Black win.
The right move is 1.Bxc1, keeping the Knight for a better time.
Over 1...g5, 2.Bxg5 Kg8 3.Nxa7! Rxa7 4.Bxe7 wins.
(If 1...g6, 2.Nxa7 Rxa7 3.Bxg5 leads to the same result.)
Piece value
The value of the pieces is too fluctuating to be expressed mathematically.
The late Stanislav Goldovski (1975-1999) offered this recapitulating table:
Pieces in Losing
Chess
Piece
+
-
King
The most important piece. Its safe moves are often needed to avoid
zugzwang.
Too slow for other tasks.
Queen
Very useful for middle-game tactics, for attacking the king, weak
squares, etc.
Dangerous in open positions, in the endgame.
Rook
Best endgame piece, quick and powerful.
Can easily turn into a 'loose cannon'.
Bishop
Good for endgame. Also for draws (due to 'oppi-colored'
bishops).
See for Rook. Much worse even.
Knight
Very good for destroying pawn formations, for
'forking' weak squares.
Too immobile, very bad in the endgame.
Pawns
Very useful for restriction of the opponent's pieces, and also
destroying pawn formations.
Slow, immobile. Often dangerous when it comes to
promotion.
If material as a whole is usually good, the King seems particularly useful for control, as shown in Lewis' lecture.
Opening Theory
Mark Watkins finally proved that White can win with optimal play, starting 1.e3. His report is available
as a pdf.
Any serious play should therefore forbid this starting move (or perhaps take some further variant).
There are those who will like it and those who will hate it, but opening theory is *MORE* important
at Losing Chess than it is at International Chess. One mistake, and your Bishop is lured into the enemy camp
and fed one unprotected piece after another. Thus it has been demonstrated that several first moves, including
1.e4??, 1.d4?? or 1.d3?? for White are automatic seventeen-move losses against best attack.
(You can find the other losing first moves there.)
We present Candidate David Bronstein's solution for Black against 1.e4??.
(You can find the solutions for 1.d4?? or 1.d3?? either on David Pritchard's Popular Chess Variants, on Fabrice Liardet's site or simply through the use of Zillions of Games.)
For a while, the most common opening for White was 1.e3, with the idea 2.Ba6 b7xa6
(2...Nxa6 loses on 3.b4). Again, some answers, like 1...d6??, had been proved to lose.
Liardet hints that the answer 1...b5 is misguided (particularly when followed by the beginner's
2.Bxb5 Bb7?, which loses the King and gives several recapturing options to White after
3.Bxc7 Bxg2 4.Bxd8 Bxh1 5.Bxe7 Qxd2, some of which, not surprisingly, are very good), because it
gives away too soon the important b Pawn (the b and g Pawns, which control the enemy Bishops and Knights,
have a non neglectible value in the opening.) and suggests 1...e6 (the solid modern defence, which
had stood the test of time), 1...b6 (the Liardet defence, which allows 2...Ba6 and keeps
the aforementioned b Pawn) or 1...c5 (the Polish defence, which allows 2...Nxa6 over 2.Ba6
followed by 3...c5xb4 over 3.b4.)
According to David Pritchard in the 1994 edition of his Encyclopedia of Chess Variants, "The endgame is the most appealing stage of the game, a garden of surprises".
Excluding as "trivial" positions where the side to move must make an immediate capture or can win by making an immediate sacrifice, the general results for piece against piece are as follows: Q/R/B against Q/R/B is trivial; Q or R wins against N or K; B or K wins against N, draws against K; N against N is a win for whoever is to move when the knights are on squares of the same colour. (...)Knowledge of the piece-against-piece results enables the results for pawn-against piece and pawn-against-pawn to be worked out.
But even two-piece endgames feature a handful of non-trivial exceptions, and Beasley doesn't shy away from adding "Every pre-computer writer who attempted to give a complete exposition, myself included, appears to have overlooked at least one of these exceptions."
Beasley wrote a comprehensine study of three-man pawnless endings, but Liardet's array with links is perhaps more accessible.
It should be noted that the side with two pieces has usually the better hand (again, more control), unless these pieces are Knights.
Three-man endings featuring Pawns and four-piece endings defy human systemization. Of the former, Liardet treats King + Pawn vs. King, Rook + Pawn vs. King and Rook + Pawn vs. Pawn. Of the
latter, he treats 3 Kings vs. King, 3 Kings vs. Rook, 2 Kings + Rook vs. King and 2 Rooks + King vs. King.)
Sample Games
On this 1997 FICS game, Fabrice Liardet translates a space advantage into a cleanup with a King.
Three games from the 1998 Geneva International Tournament.
Tossing away one's pieces may backfire.
Even a meek Pawn advance to its third rank may result in fatal attraction.
This game type is oddly characterized as a duel.
A lively AISE game.
Another sharp endgame, with two uncommon features (but uncommon is common at Losing Chess!): a Pawn-only late stage and a promotion to Queen, from David Pritchard's Popular Chess Variants.
Another game where a promotion to King doesn't save Black's last man.
There is a seven-page entry for Losing Chess in David Pritchard's Popular Chess Variants.
Computer play
The mandatory capture rule translates in a very low branching factor. Hence, computers are very good at Losing Chess. Note, though, that the version implemented in the release of Zillions of Games allows castling.
Written by Hans Bodlaender, using texts of Andrea Mori, Alessandro Castelli, and Anders Ebenfelt.
Rules, as used in AISE, were described above. Thanks to Michael Fischer for
drawing my attention to this popular variant. Thanks to Tim Mann for information on the variant played on
the Free Internet Chess Server. Ross Crawform wrote me about Suicide Chess. WWW page created: 1995 or
1996.
Last modified: December 15, 2005.
April 9, 2000: David Howe modified the rule groupings to match the more commonly played suicide chess
rules. Based on input from Lenny Taelman.
Rewritten and expanded by Antoine Fourrière, with many borrowings from Fabrice Liardet, Stanislav Goldovski, John Beasley, Fabio Forzoni, David Pritchard...
May 8, 2018: updated with the information that the game is weakly solved.