Comments by HGMuller

'Tascbase (now discontinued) had a TrueType chess figurine font for diagrams, which is freeware I'm sure'

Yes, but it seems to contain only symbols for orthodox Chess pieces. There are many freely available fonts that do that, (e.g. see http://www.enpassant.dk/chess/fonteng.htm ), but virtually none those support fairy pieces. And the few that do usually support them as rotated ordinary symbols (upside-down Queen for Grasshopper etc.), which I positively hate.

There is George Tsavdaris' WinboardF font, but it is far more limited than Alfaerie. And there is the Superchess font that supports the orthodox plus 9 non-orthodox Chess pieces, but it has a licene agreement that allows it only to be used for Superchess. So besides the fact that it is even more limited than the WinboardF font, it cannot b distributed with WinBoard.

M. Winther: 
| Why don't you download a font editor and complement the Tasc figurine 
| font with fairy pieces.
| /Mats

Would this be before or after I wrote the 7-men tablebase builder, a Shogi engine, back-ported WinBoard to Linux, fixed up Joker80 for the upcoming World Championhip of the unspeakable variant, and implemented on-the-fly tablebase generation in my normal-Chess engine Joker?

Problem is that I am aready severely overloaded with Chess-programming projects, so I try to avoid doing things that others might do as well or better as I could do them, or might even have done already. So that I can concentrate on those that people are waiting for most, and will not be done unless I do them.

Sorry to post off-topic here, but I have been trying to send e-mail to
Fergus, but it seems that the POP-server at chessvariants.org is no
longer
working. So as Fergus was posting here, I hope to catch his attention. I
wanted to submit an entry for the piececlopedia.

A second question that I am not sure where to direct:
I am looking for the e-mail address of Bill Angell, the author of the
Capablanca version of GNU Chess (of which the executable is available
from
the CV website). This because I wanted to ask him for the source code.
But
the CVpage says 'contact us' for Bill's e-mail address. Does
CVpages still have a valid e-mail address for Bill Angell? The
cais.cais.com address was not working.

Well, as most varints are backed by zero-person organizations, one person
seems already a huge leap forward. I would say it doesn't really matter
very much how many persons you throw at this: if people do not like the
game, it will not catch on even if you would throw 100 people at promoting
it. If they do like it, it shoul propagate itself whereever you plant the
seeds, and one person could do that easily.

| 2. The cost of designing and manufacturing, not to mention maintaining a
| stockpile for sale, of all the potential piece types is nearly prohibitive
| if applied to the Staunton-style appearance. Thus the use of simple
| colored discs with either letters or symbols embossed or painted.

Yet this is exactly what some people do:

OK, €89,95 for this entire set of 32 pieces (16 white, 16 black) is not extremely
cheap. It amounts to €2,81 a piece. (For more pictures, look here.)

Well, the person selling these does not do it for profit (he is retired),
so my guess is this is close to manufacturing cost.

I am not sure what you mean by 'cover all piece types'. What is 'all'?
The number of possible piece types is unlimited, but except for the
orthodox Chess pieces no Staunton shape is defined for any of them. People
do not even agree about how they should be called, so taking a shape that
could be considered a logical representation for that name is already a
hopeless task.

So any unorthodox piece can represent what you want it to represent. If
there are enough different models, you won't have too much trouble
picking a subset for playing any given variant.

Btw, about the Superchess software: I have instructions available in
English for that, on my website:
http://home.hccnet.nl/h.g.muller/superE.html .

Reinhard:

When I rigged WinBoard and Fairy-Max to play Superchess, I interpreted the
rules that are given on the superchess website rather liberally, in order to
have a quick result. (The Dutch Championship takes place Oct 12!)

I did not feel like implementing the prelude as it is described in the
rules. The prelude in my eyes is merely a way to randomize the opening
array without the aid of any external equipment, so that it can be used in
Human over-the-board play. No strategical advantage can be derived from it,
as one is forced to symmetrize after every choice by the opponent, undoing
every advantage the choice could have had. It is just a clever way of
making sure both players have an effect on the opening position, so that
neither of them can be sure what he gets to play, and derive opening
theory for it.

When a computer participates, there is no need for this. Even if the
engine cannot be trusted not to cheat, one can have the GUI set up a
shuffled array, and transmit it as a FEN to the engine(s). So this is what
I do. The GUI creates a random setup according to the rules, by starting
with the FIDE array, and then randomly deleting 2 pieces from a1-d1, and
two pieces from f1-h1, and then randomly filling the holes with the four
'exo-pieces'. If people object that this gives them less influence on
the opening array than with te prelude method, they are in fact wrong:
when playing a computer, they can click 'new game' until they get a
position that they like, and if they are patient enough, they could even
get exactly what they want. (There are 6 x 3 x 24 = 432 arays possible.)

This avoids the problem of having to design a protocol for exchanging the
pieces. (These are basically drop moves to occupied squares, so I could
have used the WinBoard crazyhouse syntax for drop moves, e.g. A@d1. But it just did not seem worth it.)

You are right about the FEN complication. In implementing Superchess, I
leaned strongly on the WinBoard Crazyhouse capabilities: in Crazyhouse
captured pieces are also put next to the board. And of course there you
also have the problem that these 'holdings' are part of the game state.

WinBoard uses for Crazyhouse, Bughouse and Shogi (the three variants with
holdings it supports) a FEN format that contains between board and stm field an optional holdings field. This field contains all pieces in the holdings (indicated by the same letter as they woud be on the board), enclosed in brackets []. So a FEN for an opening position could look like this:

rnavkser/pppppppp/8/8/8/8/PPPPPPPP/RNAVKSER[NBBQnbbq] w KQkq - 0 1

(I used A=Amazon=QN, E=Empress=RN, S=Princess=BN and V=Veteran=KN, but in
WinBoard this is user-adjustable.) On input WinBoard also understands the
b-FEN standard, which encodes the holdings as an extra rank of the board.

Castling in Superchess is possible only with a Rook; the fact that all
Rooks are still on the board in the FEN I gave is accidental. You might
have no Rooks at all, and then there is no castling.

It would be great if you could make SMIRF play Superchess. If it would
play through Smirfoglot uder WinBoard, it would just be a matter of
implementing the two new pieces, and adapting the promotion choice.
(Something I haven't fixed in Fairy-Max yet. It is a bit hard to fix
this, as Fairy-Max / micro-Max used to be an 'always-promote-to-Queen
engine'.) WinBoard already takes care of shuffling the opening array.

Well, I am not very happy either about the renaming of quite common pieces
in Superchess. But when in Rome, one does like the Romans...

The use of the [] came more or less automatic, as I started defining
Superchess in WinBoard as a form of Crazyhouse. I am not entirely happy
with that either, as the 'holdings' here have a different meaning than
in Crazyhouse (promotion pieces in stead of drop pieces).

OTOH, and this we discussed before, I am skeptical about your desire to be
able to encode the rules of the game in the FEN. Even requiring that each
piece has a unique letter, which is universally valid over all variants,
is doomed: as John remarked below, there are more pieces than letters. So
whatever system you devised, it would necessarily be limited to a subset
of the variants. While other variants would still need FENs.

For the sub-variant of Superchess played at the Dutch Championship, it
would be feasible to unify it with Capablanca-type variants.

If you want a FEN format which uniquely specifies the variant, which is
usable over a wide range of variants, I think you should build in a way to
specify exotic pieces (for which no standard letter exists). What I would
do is to allow replacement of a single piece letter by a description of
the piece in parentheses. E.g. in stead of G for Giant/Amazon you could
use (QN). 

So you would get FENs like:

3k4/6(QN)1/8/8/8/8/8/3KN w - - 0 1

Then you would only need a fairly limited set of move-descriptor
letters for use within the parentheses. You could use B,R,Q to indicate
sliders (where Q would be shorthand for RB), and the Betza system ( http://www.chessvariants.org/piececlopedia.dir/betzanot.html ) built on F, W, A, D, N, H, L, J, G for leapers. Repetition of a leaper symbol would indicate a slider with that step (i.e. (NN) would be Nightrider, B would be shorthand for (FF)).

For compactness, you could allow definition of shorthand letters within
the FEN: (NN=H) would mean that subsequent H or h (without parentheses)
would indicate white and black Nightriders, respectively. This would be
especially important for Pawns, of which there usually are a lot. And
promotion rules are a property of a Pawn, so Pawns with non-standard
promotion rules would need to be described. A Pawn that could promote only
to Ferz (like in Shatranj) could be designated as (P:F). I would use an
explicit negation character for excluding pieces, like (P:!C) in Janus.
And of course define a shorthand letter for it, as there are likely to be
many Pawns in most Janus positions, (P:!C=P), so all subsequent Pawns
would be simply P or p. (P:*) could mean promotable to every captured
piece. Superchess positions would allow promotion to some pieces that were
replaced in the prelude as well as to captured pieces, which could be
written as (P:*qbbn=P), where lower-case indicates the piece is in finite
supply.

Reinhard, I am not sure what you are trying to say. How can you separate
variant-dependent naming from the FEN standard? The FEN is one of the few
places where pieces are named in the first place.

In PGN the problems are far smaller, as these have a variant tag. So a PGN
game always unambiguously specifies the variant it is for. And indeed I
exploit that in WinBoard: if you paste a OGN game into WinBoard, it
automatically switches to the variant the PGN is for. FENs encountered in
this context can benifit from the fact that the variant is known as well.

The problem is isolated FENs, in particular isolated FENs for non-starting
positions. I have not found a way to deduce the vriant from looking at the
FEN string. So FENs that obviously must belong to a different variant as
the current one, because they use non-valid piece letters or wrong board
size, are simply rejected when ou paste them into WinBoard.

It seems to me you want the variant (and by inference the rules) to be
recognizable from the FEN, without prefixing the FEN with an explicit
variant name. Otherwise there would be no reason, for instance, to specify
the type of castling in the FEN.

I think predefining many pieces in a standard is self-defeating, as you
would be forced to pick letters for pieces that are unacceptable to those
playing the particular variant, even long before you would run out of
letters. So the only thing universal in such a 'standard' would be that
it is universally not used...

The major variants (Xiangqi, Chess, Shogi, Capablanca) are fortunately
recognizable from their board size, and this could be used to define a
default piece encoding acceptable to that variant. You will never get
(Western) Shogi players to have the Gold general represented by anything
like G, or Xiangqi players to represent the Cannon by anything else but
C...

Indeed, unifying Chess960 and Chess this way is a nice concept. But it only
works because these games nearly are the same variant. You acknowledge
yourself that you already run into problems with Janus vs CRC, which,seen
from the viewpoint of a Xiangqi or Shogi player are practically the same
game. But the promotion rules are slightly different, as are the castling
rules.

The promotion rules could be attributed as a property of the Pawn, and in
this view a CRC Pawn and a Janus Pawn are different pieces. This becomes
more obtuse in Chaturanga, where the promotion is determined by the board
square you promote on, and thus can no longer be considered a property of
the Pawn.

And how about Losers Chess vs normal Chess? How could you recognize that a
FEN represents a poition from Loser Chess rather than normal Chess. How
would you see it from the PGN if the variant tag was merely a comment? The
game might end with a resign, so the absence of checkmate might not be
apparent.

Your unified approach simply does not work when the variants differ more
than a trifle, or becomes exceedingly cumbersome. WinBoard aims at
supporting a wide variety of variants. A really universal FEN standard
should be able to handle variants the designer of the standard did not
even know. This is why X-FEN is unacceptable for use in WinBoard, both in
communication with the user and with the engine. It does a lousy job
representing Xiangqi and Shogi positons....

Reinhard:
It is of course OK to design X-FEN with a limited scope. But that makes it
unsuitable for applications tht require a wide scope, such as WinBoard (or
Game Courier). And when X-FEN includes features that are incompatible with
the needs for the variants in the wider scope, it makes it unacceptable for
use even for the variants it was designed for in that scope.

Fergus:
Is the FEN format you use in Game Courier described somewhere? IMO a FEN
is a device for describing game state, not for identifying the variant, so
that the variant cannot be deduced from the FEN is not really a problem. I
never thought about hexagonal boards and such, but now that I do the
logical way to implement those would be to use another charater than '/'
for separating the ranks of the FEN. E.g. '\' could mean 'start a new
rank, offsetted half a cell left w.r.t. the rank it terminates'.