Classes of chess pieces for chess variants

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
hgm
Posts: 23881
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Classes of chess pieces for chess variants

Post by hgm » Mon Jan 13, 2020 3:01 pm

I uploaded a zip file with the converted Fairy-Max as I have it now, plus WinBoard, configured for playing on the cirsquare board, to http://hgm.nubati.net/ThroneChess.zip . When you launch the winboard.exe in the folder, it should automatically switch to 'variant throne' (as I called it), because this is the only variant this Fairy-Max derivative says it can play.

Many things do not work yet; if I let Fairy-Max think it sometimes crashes, or makes a false mate claim. Promotion also doesn't work yet (although I specified a zone depth of 4?!?). But it should be good enough to check if the moves are implemented correctly, by switching WinBoard to Edit Game mode, so that you can move pieces of both sides around, and the engine will never start thinking. When you 'grab' a piece, WinBoard should indicate where it can go to (per specs received from Fairy-Max). Perhaps you can check if you see any foul-ups there.

Castling is not implemented yet. To be frank, I wonder whether castling as you described it has any use in this game; the Rooks will not be trapped when you move the King towards a corner on its own steam, and the King can forthermore take a short-cut by stepping diagonally. And directly behind the Pawn wall doesn't seem the safest place; it would probably be much better to just step it one square backwards or behind the Queen.

User avatar
hgm
Posts: 23881
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Classes of chess pieces for chess variants

Post by hgm » Mon Jan 13, 2020 10:16 pm

The tables are corrupted for diagonal sliders on square 34 (and equivalents). It thinks it can move from there in 1430 different diagonal directions! (And these cover the entire board, causing spurious checkmates.) Apparently because some list of trajectory starts overflows into the counter. I still have to figure out how this comes about.

RichV
Posts: 43
Joined: Sun Apr 14, 2019 3:33 pm
Full name: Rich VanDeventer

Re: Classes of chess pieces for chess variants

Post by RichV » Tue Jan 14, 2020 5:16 am

hgm wrote:
Mon Jan 13, 2020 3:01 pm
I uploaded a zip file with the converted Fairy-Max as I have it now, plus WinBoard, configured for playing on the cirsquare board, to http://hgm.nubati.net/ThroneChess.zip . When you launch the winboard.exe in the folder, it should automatically switch to 'variant throne' (as I called it), because this is the only variant this Fairy-Max derivative says it can play.

Many things do not work yet; if I let Fairy-Max think it sometimes crashes, or makes a false mate claim. Promotion also doesn't work yet (although I specified a zone depth of 4?!?). But it should be good enough to check if the moves are implemented correctly, by switching WinBoard to Edit Game mode, so that you can move pieces of both sides around, and the engine will never start thinking. When you 'grab' a piece, WinBoard should indicate where it can go to (per specs received from Fairy-Max). Perhaps you can check if you see any foul-ups there.

Castling is not implemented yet. To be frank, I wonder whether castling as you described it has any use in this game; the Rooks will not be trapped when you move the King towards a corner on its own steam, and the King can forthermore take a short-cut by stepping diagonally. And directly behind the Pawn wall doesn't seem the safest place; it would probably be much better to just step it one square backwards or behind the Queen.
I opened Throne Chess in Winboard in Edit mode. The setup of the pieces was arranged for an 8x8 board on top of the 12x12 Thrones Chess board image. All the pieces were setup on ranks 1 & 2 and 7 & 8 for an 8x8 board. . When moving the pieces, they seemed to be aligning with an 8x8 grid. When I moved the black Knight, it allowed me to move it all the way across the board on one move. Could not really try to check for correct moves.

User avatar
hgm
Posts: 23881
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Classes of chess pieces for chess variants

Post by hgm » Tue Jan 14, 2020 7:10 am

Ah, I understand what goes wrong: you should not tick "Just view or edit game files" in the Startup Dialog, but click "OK" immediately. And then use the Mode -> Edit Game menu when you want to play for both sides without the engine thinking. The difference is that in "Just view or edit games" mode WinBoard is started without any engine, and then it always starts in normal Chess mode. The board will look a bit strange, like a mosaic, because it then cuts out squares from the (now too large) background drawing in wrong places, but it is a normal 8x8 board.

The Fairy-Max entity is the only one that knows anything about Throne Chess, so it is vital that WinBoard is started with that engine. The engine will then tell WinBoard which variants it can play, which in this case is only 'throne', so that WinBoard cannot stay in its normal Chess mode (which you will shortly see first), but switches to throne, and tells Fairy-Max it must play throne. In response to that Fairy-Max will tell WinBoard that throne is a 12x12 game, and what the initial position is, and WinBoard will adapt its display accordingly. By default you will then be in 'Machine Black' mode. So to prevent that the engine will start thinking and play its own move after you play one, you still have to switch to Edit Game mode through the menu.

While you are playing in Edit Game mode, Fairy-Max will be consulted by WinBoard for any piece you 'pick up' to know where it can move to. So even in this mode the engine will be used all the time; it will just not do deep searches.

User avatar
hgm
Posts: 23881
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Classes of chess pieces for chess variants

Post by hgm » Tue Jan 14, 2020 6:37 pm

I uploaded a new version of the Throne Chess package (same link), which fixes some of the mentioned problems. Promotion now works; in Fairy-Max it always worked, but this was not properly communicated to WinBoard, so that Fairy-Max' Queen in WinBoard stayed looking like a Pawn. When Fairy-Max now highlights the legal moves, it will clearly indicate which of those moves are promotions, so WinBoard will know.

The problem with diagonals from square 34 (and mirror images) is now also solved; I had specified one of the possible trajectories starting one square too early, so that it wrongly started from 34 as well as a 5th diagonal, causing a buffer overrun, as the table was only designed to hold 4.

Fairy-Max still crashes if I set it thinking, though.

RichV
Posts: 43
Joined: Sun Apr 14, 2019 3:33 pm
Full name: Rich VanDeventer

Re: Classes of chess pieces for chess variants

Post by RichV » Fri Jan 17, 2020 12:17 am

HGM, Excellent! Following your latest directions for opening WinBoard for Thrones Chess works perfectly. Now the board looks great and moving the pieces works great. I have only moved a Pawn and a Knight. I will move all the pieces and check for proper moves against the CirSquare 96 board with notation for the 12x12.

When I first tried to move a Knight, I could not visualize the moves properly and thought they were incorrect. But when I checked them against the CirSquare 96 notated board, they were perfect. Then I went back to WinBoard and tried visualizing the moves again and then I could see them correctly. The moves through the four 4x4 corners of the board are hard to visualize at first. But I think I will soon get used to it.

Have you had a chance to look at or download my CirSquare 96 notated board yet? http://castlestrife.com/08_csg_story_08_23_p23.html
The images display on the page now and the links to open the image files work.

THANKS!!!

RichV
Posts: 43
Joined: Sun Apr 14, 2019 3:33 pm
Full name: Rich VanDeventer

Re: Classes of chess pieces for chess variants

Post by RichV » Fri Jan 17, 2020 2:04 am

HGM, I just played one game all the way through making sure I moved all the pieces. I promoted 2 Pawns for each side. Performed numerous checks. I captured everything except the 2 kings. I forgot to check the checkmate condition. All the moves worked properly, except for a king in check next to a wall--it could cross when in check. But I think you have not programmed that yet.

It is much easier to visualize the "circular" moves on the CirSquare board, but that was expected. Still, it was exciting to play. Even though the circular moves are easier to see on the CirSquare board, a first time player of Castle Siege Chess (Thrones Chess II) still does not usually start looking for the circular moves until well into the first game or even into the second game. They tend to initially mainly look at moves across the middle of the board--typical classic chess moves. It takes layers new to CSC usually 2 or 3 games to get used to it, depending upon their chess skill level. The more skilled they are, the quicker they catch on.

When I tried to apply for a trademark, I checked with a lawyer first, and he said that I would most likely get turned down because similarity to Game of Thrones Chess Set. So I canceled the application and will go back to Castle Siege Chess for now.

Great work!!

User avatar
hgm
Posts: 23881
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Classes of chess pieces for chess variants

Post by hgm » Mon Jan 20, 2020 5:20 pm

Indeed, I did not program the king-in-check rule yet. What is the use of this rule anyway? I tried a few end-games (K+R vs K, K+2B vs K, K+B+N vs K), and Fairy-Max never had much trouble checkmating the bare King in any of those cases.

RichV
Posts: 43
Joined: Sun Apr 14, 2019 3:33 pm
Full name: Rich VanDeventer

Re: Classes of chess pieces for chess variants

Post by RichV » Mon Jan 20, 2020 10:04 pm

HGM, A circular board has no corners to push a king into where it has a maximum of only 3 escape spaces. Spaces next to a sidewall provide a maximum of 5 escape spaces. With the board being 50% larger than a classic 8x8 board, it takes much longer to push a king against a sidewall or to push a king a corner, even if it did exist. The castle walls provide a means to temporarily create another sidewall both inside and outside the castles. And since the castle walls are perpendicular to the outer circular wall, that means there are 8 locations where temporary corners can be created. I could easily envision a medieval battle in which a King gets pushed up against the inside of a castle wall or on the outside of the castle wall on the battlefield, so it sort of carries a certain level or realism with it.

The rule that prevents a king from crossing a castle wall while in check helps to shorten a Castle Siege Chess game so that it is much closer to the length of a classic chess game. Without it, there can be too much chasing of the king all over the board.

When I tried having a king by itself against a king and a knight, it was hard for me to create checkmate on a classic chess board and very hard on the CirSquare board. But then I am not an excellent chess player. I was able to "fabricate" several setups of those pieces that would create checkmate, but when I backed off those positions to the middle of the board, it was virtually impossible to move them back to the checkmate condition using legal moves without the king always having to make an illogical/nonsensical move at some point near the side or corner. I did manage to get it to occur once after many tries.

The castle wall rule allows the attacking pieces to attack from both sides of the wall, while the king next to the wall can only escape away from the wall, which helps the attacker force checkmate more easily and sooner. After playing 100s of games over 5 years of testing with many different skill levels of players at chess clubs, I have found that this rule adds more excitement to the game and helps keep it from lasting too long without it.

Since the players are having to stay aware of how many times their B, R, and Q cross the castle walls on each turn, it is easy for them to be aware of the castle walls when they are in check. I have found that it rarely takes more than one game for them to remember the rule. The addendum that was added -- the king can only cross the wall when in check to capture an unprotected attacking piece prevents a Queen from too easily placing the King in check directly across the wall from the king. (I was not able to program the addendum in ZoG.) The addendum was strongly encouraged by two chess masters that played me and I tried to checkmate them that way.

User avatar
hgm
Posts: 23881
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Classes of chess pieces for chess variants

Post by hgm » Tue Jan 21, 2020 4:28 pm

I have had Fairy-Max play games against itself now without using this rule, but they don't last particularly long. I guess this is because the pieces are so much stronger, especially rooks and queens. KR-K can be won very much faster than in normal chess, because the rook creates its own corner, where its move along the circumference of the board intersects that across the board. So instead of having to drive the king into a corner in many steps, you just bring corner-confinement to the king with a single rook move.

And bishops standing next to a triangle get two of their sliding moves bent in the same direction, effectively making them double-barrel bishops; two of those can cover 4 adjacent diagonals, so that the winning king only has to guard a single square to create a death trap in the center of the board. This can also be done quite fast; no need to drive the bare king to the edge at all.

A queen seems horrendously strong on this board, and most of the initial games I tried ended in draws through perpetual checking by a queen thar sneaked behind enemy lines. I did play some games for piece-value determination without queens, to reduce the draw rate. Tentative conclusions are that a rook is worth almost exactly two knights plus 1 pawn. (But that is a rather weak 'opening pawn', not an end-game passer that is the usual standard. The bishop pair is worth two knights plus two such pawns.

Post Reply