Check out Glinski's Hexagonal Chess, our featured variant for May, 2024.

This page is written by the game's inventor, Benoit Dauphin.

Dervish Chess

By Benoit Dauphin



Playing and designing chess variants has been one of my many hobbies for some time now. This is the first time I bother to publish a chess variant, because I didn't think the previous ones were bringing something original or really interesting. I would play them with a semi-blindfold fashion (I don't even have a chessboard at home, so I make a drawing every five or ten moves) and get bored after a few games. Now this would mean that I believe Dervish Chess to bring something new, or at least to be really enjoyable. Maybe. Although I'm not sure that this variant has a great 'chess-appeal'. The proof is in the playing... I have written a ZRF file, which implements almost every rule. Hopefully some people will get interested in this variant.

What was the main idea? A large chess variant, yet another. With a great variety of pieces, not too strong, nor too weak. If I wanted many different pieces, I needed to keep a high piece density, which would lead to a long game with many pieces to maneuver, or I could stop duplicating the pieces. That's the solution I adopted. It leads to a game where the board is nearly two times bigger than a chessboard, although there aren't so many more pieces. The mobility of the pieces being adjusted to the size of the board, the games shouldn't be much longer than chess.

The other point was to deliberately pick pieces of equivalent strength. In most large chess variants I have played, it's quite the opposite. Game designers tend to stretch the scale to the maximum, having Zebras and Amazons in the same game. This increases the possibilities of having sacrificial or complex exchanges instead of the boring trading of equal pieces. The point is that I usually abandon the weakest piece (of course I shouldn't...) and take most of my time looking at the threats created by the strongest pieces. This doesn't mean the game is bad, rather that I don't play it the right way. But it explains why I wanted to design a game based on the opposite principle. I didn't try to make a large Shatranj variant either. Dervish chess has a violent opening, is barely longer than Chess, and to some extent it has a chess feeling. By the way, I made a Shatranj variant out of it (see below).

I'm positively convinced that a good chess variant can be made out of these principles. Having pieces of equivalent strength doesn't simplify the trading part. It is believed that a Bishop and a Knight have roughly the same value, but you'll be happy to find yourself with a Bishop rather than a Knight in the endgame. Trading pieces that have a close value may lead to those tiny little positional differences that will make the final decision. So I don't think this is a real drawback. Besides the game is quite tactical, and tactics will often make the difference. Finally, in this game, the strongest piece is about two times stronger than the weakest (not counting Pawns).

The last point I will evoke is the choice of the pieces. I was looking for ten different pieces with roughly the value of a chess minor piece. I wanted them to be sufficiently different from one another, and the whole set to be well-balanced (which by no way means perfect symmetry). I wanted original pieces, but not weird pieces that have a very complicated move (maybe another variant). I also needed pieces that wouldn't make the board feel to big for them. I'm quite happy with the final set of pieces (of course, if I hadn't published this page, it would be changing over and again).

Now, to the game.


The game is played on a checkered 11x11 board.


White's first rank: Mandarin (Ma1), Bishop (Bb1), Pao (Pc1), Tiger (Td1), Wolf (We1), King (Kf1), camel (Cg1), Elephant (Eh1), Vao(Vi1), Dervish (Dj1), Rook (Rk1).

Black's first rank: Mandarin (Ma11), Bishop (Bb11), Pao (Pc11), Tiger (Td11), Wolf (We11), King (Kf11), Camel (Cg11), Elephant (Eh11), Vao (Vi11), Dervish (Dj11), Rook (Rk11).

On his second rank, each player has eleven Pawns.

This opening setup was chosen upon the following characteristics:
- Every Pawn is protected.
- The colourbound leapers are on two different colours, as well as the two colourbound riders.
- The two wings are well-balanced (a straight rider, a crooked rider, a cannon rider and two leapers).
- The leapers are in the center, the riders outwards.
- It doesn't seem to be too dangerous for Black (after play testing many initial arrays, I'm not sure any more any opening gives a decisive advantage to White, as long as all Pawns are protected; the random variant should be playable under these provisions).

These characteristics were partly chosen for aesthetical reasons- reintroducing some kind of symmetry...



The Pawns are much like chess Pawns. They are quicker and may move up to three squares right in front of them at any time of the game. They capture diagonally forward. The 'en-passant' rule has been extended; a Pawn can capture 'en-passant' every time an opponent's Pawn has passed on its left or right with a double or triple move. It works even if the Pawn that passed by has been promoted (this is not implemented in the Zillions file). The capture 'en-passant' is not, in this game, a rare rule; it's about the whole balance of the game (see the wild variant below).


The King moves one square in any direction. It is identical to the chess piece, except that it cannot castle (after play testing, this move didn't seem really necessary).


The Bishop moves like a chess Bishop, any number of squares in a diagonal direction.


The Rook moves like a chess Rook, any number of squares in an orthogonal direction.


The Pao is the Chinese chess Cannon. It moves like a rook, but needs to leap other exactly one piece (of either side) to capture the enemy piece it lands on. It is a very tricky piece in the opening.


The Vao is the diagonal counterpart of the Pao. It moves like a Bishop but needs to leap other exactly one piece (of either side) to capture the enemy piece it lands on. I kept the meaningless name of Vao mainly because it was meaningless... The Vao too is tricky in the opening.


The Elephant is better known as the FAD. It leaps one or two squares diagonally, or two squares orthogonally. It has the shortest range in the game and is rather a defensive piece.


The Camel is better known as the Wizard of Omega chess. This enhanced form of the traditional Camel moves one square diagonally or leaps three squares, two orthogonally and one diagonally. I kept the name so that all the leapers had animal names.


The Wolf moves one square orthogonally, or leaps two squares diagonally, or three squares orthogonally. I chose it for its pattern of moves, each being one knight-leap away from the others. Can this help its defensive task? I have no idea.


The Tiger combines the moves of a Knight and a Zebra. It may two squares, one orthogonally and one diagonally, or three squares, two diagonally and one orthogonally. The Tiger is slightly more powerful than the other leapers and is a terrible forking piece.


The Dervish is not a new piece. It has been described by Ralph Betza in his 'Chess on a really big board'. But I'm not sure anyone has ever played the variant and the piece didn't drive much attention on it (as far as I know). It is a circular rider, having alternately a diagonal and an orthogonal move. Its complete move pictures an octagon. I reduced its power by forcing the Dervish to begin its move with a diagonal step. When you know its pattern, it's an easy piece to use. It is a difficult piece to develop, but when the game opens up, it becomes a powerful piece, probably the strongest in the game. For this reason and because it is the most original piece, I named the game after it.

Here's an ugly little diagram indicating how many steps the Dervish needs to get to a given point. The higher the number, the bigger the chances that the path is obstructed somewhere.


(Knights or Camel moves are easy, Alfil or Wazir moves much more difficult, except in the endgame; and it can never make a Dabbaba or a Zebra move; note that the squares '2' can always be joined by two paths, the other one requiring six steps).


The Mandarin is probably the trickier piece, and it has a complicated move. It moves like a Maorider, making alternatively orthogonal and diagonal steps, but stopping only after a diagonal step is completed. It captures like a Rhino, making the same steps, in the same order, capturing either on an orthogonal or on a diagonal step. In any case, the first step is orthogonal. This piece changed its move many times so that it wouldn't become too strong or too weak. The Mandarin name comes from the recognition of its Chinese ancestor, the Mao.


The game is won by checkmate, stalemate, or when the opponent's King is bared.

These rules were implemented to diminish the number of draws. It is difficult to checkmate when the game is balanced and most pieces have been traded; promoting a Pawn might not be enough. Thus the player who has the best chances is given the choice. Stalemate should be very rare (I haven't seen any so far).


I have play tested the game quite a while now (with ever-changing rules), and it seems to work as intended. But I am not a good chess player, neither is Zillions. So any feedback will be appreciated.

I made an attempt to determine the value of the pieces. It is more of an intuition than anything else. Here are my premises. A Bishop ought to be worth the two thirds of a Rook, and a Pao slightly more than that due to its opening abilities. The relative values of the Bishop and Vao should be the same. The Elephant, the Camel and the Wolf have roughly the same value (they all leap to 12 different squares), and on an 11x11 board it is close to that of a Bishop. The Tiger is slightly more powerful (it may leap to 16 different squares). It's hard to get a correct evaluation of the Dervish- it can move to more squares than any other, but its paths are easily obstructed. The piece density being low, it becomes more powerful than the other pieces when it is developed. I rate it slightly stronger than a Rook. The Mandarin's value is really a guess made after some play testing. It stands right in the middle of all these pieces. Pawns have an easier promotion than in chess, but then it is less decisive. Their value should be close to a chess pawn, maybe slightly superior.

Here's a summary:
Pawn 1
Vao 2.25
Bishop, Elephant, Camel, Wolf 3
Pao, Mandarin, Tiger 3.5
Rook 4.5
Dervish 5

Many openings are possible, both with Pawns or leapers. My favorite is the c-Pawn advance. It opens the way to the Pao, a dangerous opening weapon, to the Bishop and to the Mandarin (although you may need to move the a-Pawn to mobilize efficiently the Mandarin). The c-Pawn advance is less decisive than in the previous versions of the game, because it became easier to defend against the Pao's attacks.

Wild Dervish

To get a wilder, much more tactical game, you simply need to change the Pawn movement. In this variant, the Pawns can move any number of squares forward. The 'en-passant' capture doesn't exist. When there is a hole in the Pawn chain, promotion isn't far away.

I have played a few games of this variant against Zillions. Games are usually shorter (40-45 moves) and violent. A material advantage may be less decisive than having more Pawns. This makes a game full of sacrifices and reversals. Strangely, when Zillions plays against itself, Black wins most of the time...

Dervish Shatranj

In this variant, I changed the movements of the pieces to have a less tactical game with weaker pieces. All the other rules are unchanged.

The Bishop, Rook, Pao, Vao and Mandarin are replaced by their halflings. A halfling piece is a piece that moves half as far as its normal counterpart rounded up (see Ralph Betza's pages on halflings). A halfling Rook in the center of the board can go three squares in each direction.

The leapers are replaced by less powerful ones. The Elephant may leap one or two squares in a diagonal direction; the three others are the Knight, moving with a (2,1) leap, the Camel, moving with a (3,1) leap, and the Zebra, moving with a (3,2) leap. These are traditional pieces.

The Shatranj Dervish moves like the regular Dervish, but cannot go farther than three steps away. An ugly diagram will show the move:


All the pieces that have been described are meant to be worth roughly between a half and two thirds of their value in regular Dervish. The hierarchy would be somewhat different. The leapers tend to lose less mobility than the riders. I'd expect the (Shatranj) Dervish, the (halfing) Rook and the Knight to be on the top spots.

What about the Pawns? In most (if not all) large Shatranj variants, the Shatranj Pawn doesn't move more than one square per turn. This is really slow when you contemplate the big 11x11 board. But the fourth variation of Turkish Great Chess described by Gollon has the same rule on a 14x14 board. So the Pawns won't be allowed to make a double move. Shatranj is a game of strategy and patience. On the eleventh rank, the Pawns will be promoted to any starting piece.

With no long-range piece left and seven empty ranks between the two armies, openings should be really quiet. This looks as a great game (for those who like Shatranj games), but it hasn't been play tested yet.

Dervish Shatranji

With weaker pieces, a game with Shogi-like drops looks natural. In Dervish Shatranji, the pieces are the same as in Shatranj. But captured pieces may be dropped back anywhere on the board (you can checkmate with a drop). There are usually precise rules organizing the Pawn drops. But the Shatranji Pawns being (mostly) chess Pawns, they protect themselves by making a chain rather than a column, so it didn't seem so necessary to keep these rules in Dervish Shatranji.

The Shatranji variant has some advantages over the Shatranj variant- it is more dynamic and probably less drawish. On the other hand, it is more tactical. But this could be said of any Shatranj/Shatranji couple. Question of taste...

I hope you'll enjoy playing these games.

Computer Play

Last note. I wrote a ZRF including Dervish Chess and Wild Dervish. It is far from perfect, and the 'en-passant' move still has to be completed. If someone capable in computer programming and interested in the game wants to take a look at the file and try to find a solution, it would be appreciated (I have never done any programming in my life and it shows).