Grande Acedrex

[hgm]

In the year 1283 a manuscript about games was pubished, and it contains a very elaborate description of the rules of a Chess variant called 'Grande Acedrex'. Better translation of the old Spanish now has led to a revision of the rules compared to what was originally believed. And what emerges is a very interesting game, that was far ahead of its time:

Grande Acedrex was played on a 12x12 board, with 8 piece types. Amongst those were King, Rooks and Bishops (although the latter were called 'Crocodiles'). There was no modern Queen yet, but insted the King was accompanied by a piece nearly as strong, the 'Griffon', which is sort of a 'double-barrel' Rook: it moves by first stepping one diagonally, and then continues sliding outwards like a Rook. Similarly, there was a double-barrel Bishop, (the 'Rhino') which first jumped like a Knight, and then continued the move with a diagonal slide in roughly the same direction. This piece is also stronger than a Rook. Then there was a Lion, which jumped 3 orthogonally (3,0), or (3,1) in any direction. Finally there was a 'Giraffe', which jumped (2,3). The Pawn was an old Shatranj Pawn, i.e. no double move or e.p., but the 12 Pawns started on the 4th rank. There was no castling, but a virgin King was allowed to jump 2 squares orthogonally or diagonally, which can be seen as a first step towards castling.

There currently doesn't seem to be any WB engine that could play this.

[Ajedrecista]

Hello:

In the year 1283 a manuscript about games was pubished, and it contains a very elaborate description of the rules of a Chess variant called 'Grande Acedrex'. Better translation of the old Spanish now has led to a revision of the rules compared to what was originally believed. And what emerges is a very interesting game, that was far ahead of its time:

Grande Acedrex was played on a 12x12 board, with 8 piece types. Amongst those were King, Rooks and Bishops (although the latter were called 'Crocodiles'). There was no modern Queen yet, but insted the King was accompanied by a piece nearly as strong, the 'Griffon', which is sort of a 'double-barrel' Rook: it moves by first stepping one diagonally, and then continues sliding outwards like a Rook. Similarly, there was a double-barrel Bishop, (the 'Rhino') which first jumped like a Knight, and then continued the move with a diagonal slide in roughly the same direction. This piece is also stronger than a Rook. Then there was a Lion, which jumped 3 orthogonally (3,0), or (3,1) in any direction. Finally there was a 'Giraffe', which jumped (2,3). The Pawn was an old Shatranj Pawn, i.e. no double move or e.p., but the 12 Pawns started on the 4th rank. There was no castling, but a virgin King was allowed to jump 2 squares orthogonally or diagonally, which can be seen as a first step towards castling.

There currently doesn't seem to be any WB engine that could play this.

An incomplete translation of the book can be found here:

http://checkersusa.com/books/History_Bi ... _1283).pdf

'Great chess' starts at page 41/65 of the PDF, just under [f.81] note.

To get a PDF version from the original book looks a more difficult task. Some pictures of the 'Great Game of Chess' taken from the book are here:

F81R
F81V
F82R
F82V

Good luck if you want to write an engine for this 12x12 variation!

Regards from Spain.

Ajedrecista.

[Evert]

Isn't this something that HaChu could be made to play?

The Rhino move looks insane, but could be fun. I'm not a big fan of bifurcating pieces in general because I find their moves hard to visualise.

Sadly, SjaakII cannot handle the bifurcating pieces (leaving aside the board size, which would also be a problem). I may be tempted to write something basic for this though (which will then probably also play Tamerlane Chess; I don't know if that's an actually playable game but I like watching computers play historic chess variants).

[hgm]

The problem is that HaChu's move generation is really Shogi oriented: radial moves and Knight jumps. The Griffon and Rhino moves would not fit in at all. Fairy-Max can do Griffon and Rhino without problems. But it can only handle 8 ranks. Spartacus is limited to 12x10.

Darn... So many engines, and not one that can do it!

Easiest would be to patch Fairy-Max to do other than 8 ranks. One of the things that has always prevented me from doing that is the extremely tricky Pawn-bonus code (for 6th and 7th rank) that it inherited from micro-Max. It derives an extra push bonus from the bit pattern of the to-square rank number, and hides a given bonus in the piece code so that it can be deducted again when the Pawn gets captured. But of course I can simply take that out, and switch to using a real PST, which for Pawns could be initialized to give the same N-1 and N-2 rank bonuses. I still would have to solve the problem that the 128 bit of the to-square number is used as a flag to indicate the move is valid as hash move.

Hm, perhaps it is worth it. Then Fairy-Max could also do Los Alamos Chess.

[Evert]

The problem is that HaChu's move generation is really Shogi oriented: radial moves and Knight jumps. The Griffon and Rhino moves would not fit in at all. Fairy-Max can do Griffon and Rhino without problems. But it can only handle 8 ranks. Spartacus is limited to 12x10.

Darn... So many engines, and not one that can do it!

I misremembered then, I thought some of the large Shogi variants had funky bifurcating movers as well. I guess not.

I did some brainstorming to see how much effort it would be to write a simple engine (he says) that could play this. It shouldn't be too bad, but I'm not stubborn enough to base the move generation on bitboards (bit-lists might still be a useful data structure for piece-lists though). I haven't done anything mailbox-like in about twenty years though so I need to rewire some of my brain.

Tamerlane chess is insane, by the way: 11 different pawn types and three different royals that all move like a king but have subtly different rules with respect to the fortresses and so need to be told apart. It'd be hard representing it all using the standard pictographs. I'd still like to include that while I'm at it. I guess throwing in Courier is free at that stage.

Easiest would be to patch Fairy-Max to do other than 8 ranks. One of the things that has always prevented me from doing that is the extremely tricky Pawn-bonus code (for 6th and 7th rank) that it inherited from micro-Max. It derives an extra push bonus from the bit pattern of the to-square rank number, and hides a given bonus in the piece code so that it can be deducted again when the Pawn gets captured. But of course I can simply take that out, and switch to using a real PST, which for Pawns could be initialized to give the same N-1 and N-2 rank bonuses. I still would have to solve the problem that the 128 bit of the to-square number is used as a flag to indicate the move is valid as hash move.

Hm, perhaps it is worth it. Then Fairy-Max could also do Los Alamos Chess.

Sounds like win/win, but a bit tricky...

[hgm]

The large Shogi variants have hook movers, which are even more disastrous, because the point of bifurcation is not fixed. But HaChu does not implement any of that yet. HaChu was supposed to do things incrementally, which for large boards should be enormously competitive. So it generates captures from an attack map. (Currently thi sis still generated from scratch in every node, which for Chu dos not entirely kill you.)

But the attack map is not set up to handle around-the-corner attacks. So the hook movers will need very special treatment, and perhaps I will just keep generating their moves and captures from scratch in every node.

I never studied Tamelane Chess much.

[hgm]

I seem to have succeeded in letting Fairy-Max play Los-Alamos Chess. This mainly involved replacing a lot of hard-coded eights in the initialization code that placed the pieces and calculated centralization bonus. And the move <-> text conversions, as unfortunately black plays towards higher square numbers in Fairy-Max. But it all seems to work now.

Of course using just a sub-set of the existing board is easier than making it larger, but it is a good first step. And it already has given me Los-Alamos Chess.

[Ferdy]

In the year 1283 a manuscript about games was pubished, and it contains a very elaborate description of the rules of a Chess variant called 'Grande Acedrex'. Better translation of the old Spanish now has led to a revision of the rules compared to what was originally believed. And what emerges is a very interesting game, that was far ahead of its time:

Grande Acedrex was played on a 12x12 board, with 8 piece types. Amongst those were King, Rooks and Bishops (although the latter were called 'Crocodiles'). There was no modern Queen yet, but insted the King was accompanied by a piece nearly as strong, the 'Griffon', which is sort of a 'double-barrel' Rook: it moves by first stepping one diagonally, and then continues sliding outwards like a Rook. Similarly, there was a double-barrel Bishop, (the 'Rhino') which first jumped like a Knight, and then continued the move with a diagonal slide in roughly the same direction. This piece is also stronger than a Rook. Then there was a Lion, which jumped 3 orthogonally (3,0), or (3,1) in any direction. Finally there was a 'Giraffe', which jumped (2,3). The Pawn was an old Shatranj Pawn, i.e. no double move or e.p., but the 12 Pawns started on the 4th rank. There was no castling, but a virgin King was allowed to jump 2 squares orthogonally or diagonally, which can be seen as a first step towards castling.

There currently doesn't seem to be any WB engine that could play this.

I have seen a site describing the piece movements with grahics.
http://history.chess.free.fr/acedrex.htm

Are these the rules you want to implement to winboard someday?

On same link, there seems to be other rule for Lion.

Code: Select all

The Lion (b1,k1) moves 3 steps orthogonally or 2 steps orthogonally followed by 1 diagonal step (jump to the opposed square of a 2x4 rectangle). It leaps over occupied squares. For Murray, the Lion could only steps 3 squares orthogonally on 4 directions, forward, backward and on both sides. For Sonja Musser, the Lion could only steps 3 squares orthogonally on 3 directions only, forward and both sides, never backward.

I plan to create an engine to play this variant at some point in time, but I need a clear rule to do this.

[Evert]

In the year 1283 a manuscript about games was pubished, and it contains a very elaborate description of the rules of a Chess variant called 'Grande Acedrex'. Better translation of the old Spanish now has led to a revision of the rules compared to what was originally believed. And what emerges is a very interesting game, that was far ahead of its time:

Grande Acedrex was played on a 12x12 board, with 8 piece types. Amongst those were King, Rooks and Bishops (although the latter were called 'Crocodiles'). There was no modern Queen yet, but insted the King was accompanied by a piece nearly as strong, the 'Griffon', which is sort of a 'double-barrel' Rook: it moves by first stepping one diagonally, and then continues sliding outwards like a Rook. Similarly, there was a double-barrel Bishop, (the 'Rhino') which first jumped like a Knight, and then continued the move with a diagonal slide in roughly the same direction. This piece is also stronger than a Rook. Then there was a Lion, which jumped 3 orthogonally (3,0), or (3,1) in any direction. Finally there was a 'Giraffe', which jumped (2,3). The Pawn was an old Shatranj Pawn, i.e. no double move or e.p., but the 12 Pawns started on the 4th rank. There was no castling, but a virgin King was allowed to jump 2 squares orthogonally or diagonally, which can be seen as a first step towards castling.

There currently doesn't seem to be any WB engine that could play this.

I have seen a site describing the piece movements with grahics.
http://history.chess.free.fr/acedrex.htm

That's the one I'm going with. I currently have the following piece movements and letters:

Code: Select all

Name:       Symbol:  Move:
--------------------------
King        K        KiDA
Aanca       A        FtR?
Crocodile   B        B
Giraffe     Z        Z
Rhino       U        NtfB?
Lion        L        HC
Rook        R        R
Pawn        P        fmWfcF

I went for "Aanca/A" rather than "Gryphon/G" to keep G free for the Giraffe in Tamerlane Chess (which is sortof the same piece as the Aanca except it cannot stop on the first two squares it slides over). For a similar reason, but also because it describes the move, I went for Z (Zebra) as the symbol for the Giraffe. I went with Unicorn/U for the Rhino and Bishop/B for the Crocodile, although there is something to be said for keeping with the jungle theme of the variant.

By the way, I'm not entirely sure how to do the Betza notation for the Aanca and the Rhino. HG?

Are these the rules you want to implement to winboard someday?

There's really no need for Win/XBoard to know the rules though, is there? If the engine can play it, XBoard could probably be configured for it already.

On same link, there seems to be other rule for Lion.

Code: Select all

The Lion (b1,k1) moves 3 steps orthogonally or 2 steps orthogonally followed by 1 diagonal step (jump to the opposed square of a 2x4 rectangle). It leaps over occupied squares. For Murray, the Lion could only steps 3 squares orthogonally on 4 directions, forward, backward and on both sides. For Sonja Musser, the Lion could only steps 3 squares orthogonally on 3 directions only, forward and both sides, never backward.

Well, there seems to be quite a bit of confusion on the translation, but my understanding is that what is presented on the page is the "current"/"best" translation.
Another point against descriptive notation for chess moves I suppose...

I plan to create an engine to play this variant at some point in time, but I need a clear rule to do this.

Give the rest of us a sporting chance, will you?

[hgm]

Please beware that the website history.chess.free.fr is not completely up to date. I exchanged a few e-mails with Jaen-Louis Cazaux (the website owner), and it turned out that his current opinion on the Rhino move is different from what that website states (although he still does not consider it completely impossible that the historic rules were as the website states). The website shows the Rhino having a diagonal slide that goes absolutely forward after the initial Knight leg. The move that he currently considers most likely (and that I also prefer) would make this diagonal slide relatively forward, i.e. in the direction that most closely resemble that of the initial Knight jump. For a white Rhino on e6 this would be f4-g3-h2 or c7-b6-a5.

The XBetza notation for Griffon is yafsF, which describes a two-leg move (a) that starts as F, and then changes to slider (y = m + range toggle). The 'fs' on the second leg of a diagonal atom indicates an orthogonal continuation direction.

The Rhino is even more complex, because in XBetza multi-leg moves have to describe all their legs with the same atom, so that you would need to use the 'smallest common divisor' for all legs. Which usually means describing the path square for square with the aid of King steps even for squares it leaps over. So Rhino is mpafsyafW, which adds the W squares as 'ghost squares' to the trajectory (with mp rights, so that the move can continue irrespective of the occupancy without any side effect on it), then continues at 45-degrees (fs) to finish the Knight jump, and then toggles to slider (y) continuing in the forward direction.

Note that King here is KiAiD: all modifiers, including 'i', group with only a single atom.

Also note that there is a naming confusion here: the common English name for the piece presented here as Aanca is 'Griffon/Gryphon', which apparently is not even a correct translation. But unfortunately Ralph Betza introduced another piece (moving very much like the Rhino here, except that it actually visits the W squares, but at the time this was not known) as diagonal counterpart to the Griffon, which he did call Aanca. So a Griffon is a sort of double-barrel Rook, while Betza's Aanca is a double-barrel Bishop. So on Spanish 'Aanca' now means something different then in English. This is not unprecedented, though: 'Alfil' in English refers to the old Shatranj 'Elephant', jumping 2 diagonal, while in Spanish it is the name of the modern Bishop.

As to what XBoard pieces to use for this variant: I always prefer to represent pieces by how they move, rather than by how they are called. (If only because names can differ between languages.) In this policy the 'Giraffe' shoudl be represented by the Zebra pictogram, which is one of the new pieces (the 25th in the pieceToCharTable) that will be added to XBoard 4.9. The Rhino was actually called 'Vnicorno' in the old Spanish manuscript, so representation by XBoard's Unicorn symbol seems logical. (Also because the Rhino is an enhanced Knight, and the Unicorn (the 21st piece) looks like one.) I would always use a Bishop symbol for a piece that moves as a Bishop, no matter how it is called. Unfortunately the XBoard Lion cannot be used for the piece called Lion here, because XBoard subjects it to the Chu-Shogi anti-trading rules. I would recommend usage of the Camel pictogram here (the new 24th piece), because it is an enhanced Camel. XBoard has no standard piece associated with the move of the Griffon, but the latter at least has some bird-like aspects (and Aanca does seem to refer to just a monster bird more than to the mythological eagle-lion chimera anyway) the the Falcon symbol (the 18th piece) suggests itself.