Comments by HGMuller

The idea is to support them all. But I have to develop a way to configure
the server for which variants to support. Perhaps the logical way to do it
would be to group them all under the category 'capablanca', and then have
sub-categories ('boards') for 'capablanca', 'embassy', 'carrera',
etc. The server currently triggers on the category name for deciding what
board size to use, becaue that ws the easiest way to do it for a quick
pilot experiment. I want to change that by something in the files that
describes the position, e.g. board=10x8. The server code could then be
completely oblivious of which variant uses which board. The same for other
rules.

Last night I did another trial run with the server. This time I found out
how to suppress the auto-logout feature (which logs out users after a
certain idle time), so that the Fairy-Max bot could stay on-line for the
entire night. It played one game of Gothic Chess against a Human, two
games of Capablanca Chess against a computer opponent, and a lightning and
standard game of normal Chess against another computer opponent. On my
internal network I was able to play some Carrera, Embassy and Bird Chess
games against it, as well as Knightmate.

So indeed, Embassy, Carrera, Bird, Capablanca are now all supported, as
different setups of the variant Capablanca.

As for the variants to offer in the future, I do think that this is an
area where a greater framework similar to IAGO Chess could actually be
useful. To get orthodox Chess players to play the variants, I think it
would be useful to educate them in small steps. So I actually would like
to offer a number of modest variants, introducing only a single unorthodox
piece. For instance:

8x8 Janus         ( (BN) replaces Q )
8x8 Chancellor    ( (RN) replaces Q )
8x8 Amazon        ( (QN) replaces Q )
8x8 Centaur       ( (KN) replaces Q )
Synode            ( 10x8 board with 4 Bishops per side )
Cannon Chess      ( Pao replaces Pawns on b2/7 and g2/7 of 8x8, no castling )
Janus Chess
Chancellor Chess  (with 2 (RN) on 10x8 )
Capablanca Chess  (with many sub-variants differing in opening array)
Xiangqi
Shatranj
8x8 Heian Shogi   (Silver replaces Bishop, Gold replaces Queen, no drops)
9x8 Heian Shogi   (same, but with 2 Gold Generals in symmetric setup)
Crazyhouse
Shogi
Superchess        (with pieces from the set (KN),(BN),(RN),(QN) randomly
                   replacing some of the pieces of a normal FIDE array)
Superchess II     (similar, but with a larger variety of unorthodox pieces)

Of course there could be training bots as well: 'wild' games where you have to checkmate a bare King with a (BN) or a Commoner.

The way I read it, Minister and High Priestess are regulare FDW and FAN, which can jump to all their distant destinations...

I do like Great Shatranj for the consistent application of a single idea, mnamely to replace all distant slider moves by a single 2-jump. But I don't think it stands any chance of replacing the Mad Queen game. Slider moves are simply too interesting to discard completely.

I might include Great Shatranj and its pieces as standard variant in the Variant Server.

I put the source of the lasker-2.2.3 server as I hacked it so far at:

http://home.hccnet.nl/h.g.muller/capablanca.tar.gz

You might have to remove the autoconfig.h before you configure for your
system as described in the README file in the lasker-2.2.3 directory.

It sounds a lot like what I am already doing. Except that I usually do not
bother with ZoG, but configure Fairy-Max to play the game, or make a
dedicated engine by adapting Joker. For on-line play I am developing the
Internet Chess Server now. Using the material-imbalance-self-play method I
have, for instance, determined the piece values for Capablanca, Knightmate,
Great Shatranj, Falcon Chess, and Superchess. I have tens of thousands of
comp-comp games on file for these variants, which could be filtered for
interesting checkmates (e.g. early in the game). Only opening theory is of
no interest to me; even my normal Chess engine plays without opening book.

I do think that pieces like the Perier Cannon could be very interesting pieces to create a modest variant. But I am still skeptical about the gating mechanism you propose in this variant to introduce them. It introduces a second unusual complication. You mention yourself that it is a difficult strategic decision when to introduce the Cannon.

Wouldn't it be much simpler to start the Cannon, say, on b1, moving the Knight to a2, and the a-Pawn to a3?

It seems your use of the word zigzag is at odds with the definition in the dictionary, quoted below. The latter clearly specifies the need for multiple angles in the trajectory, while your pieces have only one:

zig⋅zag 
noun, adjective, adverb, verb, -zagged, -zag⋅ging. 
–noun 
1. a line, course, or progression characterized by sharp turns first to one side and then to the other. 
2. one of a series of such turns, as in a line or path. 
–adjective 
3. proceeding or formed in a zigzag: zigzag stitches.  
–adverb 
4. with frequent sharp turns from side to side; in a zigzag manner: The child ran zigzag along the beach.  
–verb (used with object) 
5. to make (something) zigzag, as in form or course; move or maneuver (something) in a zigzag direction: They zigzagged their course to confuse the enemy.  
–verb (used without object) 
6. to proceed in a zigzag line or course.

I know that for the unspeakable variant a lot of opening theory already
exists, but I don't trust it, as those who have been playing the Capa
variants seem to have been consistently underestimating the value of the
Archbishop. The Capablanca position, for instance, is often mentioned as
unplayable for black, due to the thereat Axi7, where the A-fork on Ch8 and
Rj8 gains white the 'Unspeakable exchange' , A vs C (often in exchange
for a Pawn elsewhere on the board). Joker80, however, often allows this as
black, as it apparently feels the half-open i-file for its Rook is
sufficient compensation for this 'exchange', which it values close to
zero. IMO this puts any existing opening theory on very shaky grounds.

I added crazyhouse to the server, and made it such that the initial
position is printed together with a move list, in games that were not
starting from the default opening position. This latter feature now allows
correct implementation of shuffle variants, and I am working on
implementing Chess960. (For this, I have to add more liberal castling
styles to the move generator and checker.)

My current plan is to make the server configurable for a wide range of
variants by supplying game-definition files. The original design was
already using this system, but the files could only describe an (8x8)
board setup, while the rules were immutable. I want to make it possible to
write some extra lines in those files, defining other game parameters. So
far I already did this for the board format, and some flags I implemented
in the server to tell it if captured pieces should go into the holdings
(and if their color should be flipped before doing so), and if drops from
the holdings are allowed. I also want to add options for setting the
castling style (none, normal, wild, Fischer, free). 

Currently I also have a flag which indicates if the Capablanca pieces are
allowed as promotion choice, represented by the hard-coded letters A and
C. But in the future I would like to handle piece types in a different
way. Each variant must be able to define its own piece set, and the
letters by which they will be indicated in the board and move
representation. In particular, I am looking for ideas how to describe a
nonstandard piece, in a way that is easy to understand and apply by the
move generator. Betza notation could be one way to do it, although I am
not sure how to describe lame leapers in it. Pieces whose move depends on
their position on the board (the Xiangqi palace and river contsrains being
a simple example of that) are also problematic. Pieces with side effects,
such as catapult pieces, pose an even larger challenge.

Are there any suggestions on how this could be done?

You seem to define left and right in an unusual manner, though, comparing
the Betza notation with your board drawing.

The Betza notation is a bit cumbersome. It would be better to introduce
lower-case modifiers for handedness, that could be applied to the 8-fold
moves (N, J and L), to split, sa, the Knight moves into left and
right-handed Chiral Knights.

At the risk of making myself very impopular on this site:

To me, inventing Chess variants is like 'inventing' integer numbers.
Make a string of some 100 digits, and the odds are overwhelming that you
are the first ever in this universe to have mentioned this number. OK, so
you can marvel at your own private number, but who cares? Pritchard was
quoted to say: Ït takes about 10 seconds to invent a Chess variant, and,
unfortunately, some people do'.

It is just like with the numbers, it had better be very special in some
respect that you point out, or it cannot be considered an invention at
all. The axioms of number theory already imply the existence of all
integers, and states that there is an infinity of them, so the fact that
you can name a few that no one ever mentioned before adds absolutely zero
to what was already known. AFAIK, there is no website where people can
post large numbers they invented. Prime numbers are already a bit more
interesting, but still so common that it makes little sense to post
everyone prime you discover. Unless it is the largest prime ever
discovered so far. (Did you know that about 0.45% of all 100-digit numbers
is prime?) Some numbers are very interesting, though, and entire books
could be written about their deep mathematical properties. This applies to
numbers like pi, Euler's constant gamma, the base of natural logarithms e.
(They are not integers, though, but the analogy would work just as well for
real numbers.) 

IMO, it is much the same with Chess variants. The 'axioms' of a royal
piece, translation-invariant piece moves and replacement capture imply an
infinite set of Chess variants, and the fact you can mention one (or a
hundred) explicitly is as meaningless as designing a hundred huge
integers. A Chess variant is only worth mentioning if it it has some very
special properties not found in most other variants, or solve some
problems found in existing popular variants.

With Chess pieces the situation is similar. A Chess variant can be
worthwile as a vehicle to exercise a novel piece, but only if the piece is
interesting. But also novel pieces can easily be uninteresting
run-of-the-mill constructs. Merely bringing up novel combinations of the
Betza atoms does not make a worthwile piece. Breaking the eightfold
symmetry gives even more pieces that could be useful on boards of limited
size, but so what? It woulkd only be of interest if it creates some
interesting irreversibility in play (such as with the Pawn), or a weird
color-boundedness not seen in other pieces. Or some intersting end-games,
where it is difficult, but nevertheless possible, to mate a bare King. New
capture modes or other side effects of piece moves could be interesting,
but have the disadvantage to make the piece less 'Chess-like'.

To demonstrate that a variat you designed has any such properties that
could make it worthwile does require a lot of analysis effort.

A piece with a constant (psition-independent) gait that has only two
targets, located in an inversion-symmetric way, will be able to move back
and forth along a line. Moves in one direction will exactly cancel moves
in the other direction, so that only the difference determines where the
piece is. This means all possible long-term destinations can be reached by
moving only in one direction. If in this process the piece skips over a
square, this square is unreachable.

If a piece has 4 moves in an inversion-symmetric pattern, such as narrow
or chiral Knights, but of course also Alfil and Dababba, the moves can be
grouped in pairs of opposing moves. For each pair the same situation as
above exists. All long-term targets can be reached through N moves in one
direction of the first pair, and then M moves in one direction of the
other pair. The targets can thus be mapped onto a two-dimensional grid,
which in general will be a subset of the board. The Wazir is the only
inversion-symmetric piece with 4 destinations that can access the entire
board.

With 6 or 8 moves and inversion symmetry, the destinations logically map
onto 3- or 4-dimensional grids, but as the board is two-dimensional, you
will see a projection of such grids on the board. Such a projection can
quite easily acces every square, as the number of grid-points in a
three-or more- dimensional grid is so much larger than the number of
squares on a to-dimensional board. So color-boundedness is the exception,
rather than the norm, in inversion-symmetric pieces with more than 4
destinations.

For pieces that do not have inversion symmetry the situation is different.
On a two-dimensional board you need at least 3 moves to lift
color-boundedness. With 2 moves, the piece is either restricted to (a
subset of) a line, or is irreversible an cannot return to its original
position after it moves. The 'Y-piece' (fFbW) is an example of a piece
with 3 moves that can acess every square of the board reversibly.

I agree: I have always disliked such 'cut and paste' representations,
both as symbols in diagrams as in 3d pieces. If I would have to design a
Capablanca set from scratch, I would use a piece with a wide-rimmed hat as
Marshall. Much like the one shown below in the picture with King and
Cannon, from the ExChess piece set (where it is supposed to represent a
Centaur (K+N), called 'Veteran' in Superchess).

The Archbishop design, as a Bishop with a V-cut mitre, does seem OK to me.
Close resemblance to the Bishop is no disadvantage there, as the name
already suggests that.

Of course I know that Archbishop and Marshall were not the original names
given to those pieces by Pietro Carrera, but I do not care much for the
name Centaur as a B+N compound.

A fortress draw originally meant a position from which you can prove that the weak side can hold out forever. This in contrast to drawn positions where the weak side draws by gaining a piece. The latter occur a lot in end-games like KFFWK, when the bare King can chase an F or W cut off from their allies into an edge or corner, after which the remaining KFFK or KFWK is a draw.

An example of a fortress in KQKBN is this:

. . . . . K Q .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . n . . . .
. . . . . . . .
. b . . . . . .
k . . . . . . .

All black pieces are defended, and the white King cannot approach the bishop to attack it a second time. This fortress holds out even against an Amazon.