On chess variants

Jan 21, 2008, 8:28 PM |

Note: I wrote this for an audience not terribly familiar with chess. I don't mean to insult anyone's intelligence by explaining who Capablanca was, or other things that might seem unnecessary in this venue.

In modern times some have begun to worry that chess will soon be played out, and that at the top levels virtually all matches will end in draws. One of the earliest notable voices to issue this warning was the legendary José Raúl Capablanca, who at the time was the reigning world chess champion. (This was in the 1920s, so depending on one's perspective his fears were either premature or prescient.)

In order to address the coming crisis, Sr. Capablanca proposed a more complicated version of the game called (not too modestly) Capablanca chess. Played on a 10x8 board, Capablanca chess introduces two powerful new pieces: the archbishop1 and the chancellor2. In nearly all other respects, Capablanca chess is played according to the same rules as orthodox chess.3

Capablanca chess never developed a wide following, partly because it turned out that there was more to discover in orthodox chess and partly because the requisite equipment — four extra pieces per set and an unusual board — made the whole thing terribly inconvenient.4 Another turnoff to players was that widening the board changed many aspects of the opening game, altering relationships between pieces and significantly weakening the influence of minor pieces.

Last year grandmaster Yasser Seirawan presented his own solution to the problem, which seems more pressing now that computers are advancing the development of new chess theory at a rapid pace. His invention, which he calls — wait for it — Seirawan chess, is similar in some ways to Capablanca chess. It contains a bishop-knight, which Seirawan calls a hawk, and a rook-knight, which he calls an elephant.5 Oddly enough, what makes Seirawan chess distinctive is that the game is played on an 8x8 board with a conventional starting position. Upon the first movement of any piece,6 the player may place the hawk or the elephant on the square vacated, with a limit of one hawk and one elephant (though pawns may be promoted to either, without limit). Seirawan is currently working with a game manufacturer to produce and market sets, and in a break from tradition the hawk and elephant actually look quite a bit like a hawk and an elephant.7

Seirawan chess is a pretty cool idea if you ask me, which we'll pretend you did, but I'm not sure it's an ultimate solution to the problem. Once someone modifies a bit of good chess software to account for the new rules and hooks it up to a fresh cup of really hot tea, the best openings will soon be determined. The game will still be more complex, and there will certainly be fewer draws when there's a piece (the hawk) that can mate singlehandedly, but it still feels like a temporary fix.

Enter the late, great, brilliant, sad, tortured, insane Bobby Fischer. Eleven years before Seirawan made his idea public, Fischer proposed an entirely different sort of variant that involved randomizing the starting position for each game. And what did he call it? Please, let's not always see the same hands. He called it Fischerandom chess.8

Fischer's idea wasn't entirely new. Games with randomized back rows have been recorded at least since the mid-19th Century, sometimes with black mirroring white's position and sometimes not. What Fischer did was introduce a set of rules to maintain all common aspects of chess play on a shuffled board, while ensuring a certain degree of fairness. In Fischerandom chess, also called chess960 for its 960 possible starting positions, pieces are positioned randomly at the beginning of the game, either by computer or some other random number generating system.9 Certain restrictions apply: each player's two bishops must be on opposite colored squares, the king must lie somewhere between the two rooks, and the black and white back rows must be mirror images of each other.

Fischer's castling method is the only part of chess960 (as it is more popularly called in my experience) that I dislike. Regardless of the starting position of the king and the appropriate rook, castling results in a position identical to that found after castling in orthodox chess, with the same asymmetry between the A and H sides (called queen and king sides in orthodox chess, but not in 960 since the king and queen are only on their usual sides half the time). Rather than devote too much space to the issue, I'll recommend that anyone interested read John Kipling Lewis's proposal for a better system. The one catch is that since Lewis's system does not preserve the board's traditional asymmetry, it effectively halves the number of starting positions by making horizontally mirrored positions identical with regard to game play. Kipling's modification therefore creates a variant that might be called chess480.

Fischer's goal in creating his variant, call it whatever you will, was to end players' dependence on opening game memorization, which he felt had made the game less interesting and inhibited the development of principled strategy and tactics. By presenting nearly a thousand different starting positions, each with its unique interrelations of pieces, his proposal elegantly solves the problem of rote openings by putting them beyond the capacity of even a highly gifted human mind. A player can spend years memorizing the contents of Modern Chess Openings, but swap the queen with the king's bishop and all bets are off. Fischerandom/960/480 defies useful deep analysis without being any less rich, complex or challenging. Some people on Chess.com are now asking for it to be added as a feature.

Of course, Seirawan was aware of Fischer's proposal when he came up with his own. As far as I can tell he dislikes the idea of randomized starting positions on an intellectual-aesthetic level. I can respect that, but I still see Fischer's idea as the one that will save chess from people much better at it than I.

(On a personal note, I spent over an hour looking for a bug in my HTML that was making this article half unreadable.10 I hope it was worth it.)

1 The archbishop begins between the queens's knight and bishop, and may move as a knight or a bishop on any given turn.
2 The chancellor begins between the king's knight and bishop, and may move as a knight or a rook on any given turn.
3 Castling is slightly different. In order to accommodate the greater distance to the edge from the center, the king moves three spaces rather than two when castling. In order to fill out the board, each player begins with ten pawns rather than the usual eight.
4 Modern players of Capablanca chess and its subvariants typically create chancellors and archbishops by acting as latter day Drs. Moreau, cutting and gluing parts of orthodox pieces to form distinctive shapes.
5 Mr. Seirawan was born in Damascus to an Arab father and an English mother. I wonder if he's considered the difficulty of introducing Seirawan chess to the Arab world, where the piece called "bishop" by English speakers is referred to as fil, "elephant."
6 In chess only the game pieces beginning in the back row are called "pieces," pawns being a separate category. The collective term for all movable units in a chess game is "chessmen."
7 Thus making life even harder for the Arabs. Seriously, shouldn't someone mention this to him? He didn't learn to play chess until five years after his family moved to Seattle, so it's possible that he just doesn't know.
8 In recorded conversations he can be heard always emphasizing the first syllable of the portmanteau, and correcting anyone who would call it merely "random chess."
9 One simple method uses a six-sided die, averaging 6.7 die rolls per setup. If one had access to a four sided die and a five sided die in addition to the usual cube, one could set up the pieces with exactly five rolls each time.
10 There was a quotation mark missing in the worst place possible.