Comments by HGMuller

'Plase endow the elephant with prominent tusks so that it can be taken for
a young mastodon or mammoth, too.'

This is not so easy: in the current design, the direction of the tusks is
perpendicular to the grain of the wood. This means they would almost
certainly break off very quickly if I would try to make them longer. They
can only survive by being fused to the lower part of the trunk.

I considered maing the grain of the wood run front to back, so that it is
in the correct orientation for supporting free-standing tusks. This,
owever, causes the problem that the trunk is now perpendicular to the
grain, and would become extremely fragile. (Although it would not be so
bad as the tusks, wich are thinner.) This could be solved by not making a
completely free-standing trunk, as in the current design, but one curving
back towards the front of the head, touching it before bending forward
again. (And the end would then point forward, in the same directio as the
tusks.) The separation between lower art of the trunk and head would then
have to be a driled hole.

I guess an illusion of longer tusks witthin th fraework of the current
design could be made by cutting more space between turnk and front of the
head. Together with making the tips of the tusks stick out just a few
millimeter more forward, this might be good enough.

As to the baby-step approach: I am currently looking at the possibilities
to modify state-of-the-art Internet Chess Server software (in particular
the 'chessd' ICS, used at FICS, where many thousands of people visit
everyday to play on-line Chess), to have it support variants. Supposing I
succeed, it would be great if we could convince FICS to make use of this
extension, and offer some more variants beside those they already do.
(They currently offer variants like Crazyhouse, Suicide, Atomic.)

This would create a great opportunity for offering the variant-unaware
Chess player a possibility to make the baby steps. We could for instance
offer variants like 8x8 Chancellor Chess or Cardinal Chess, where a single
un-orthodox piece replaces the Queen, just to get them acquainted with the
piece in question, and prepare them for other Capablanca variants.

Similarly, we could run games on bigger boards (10x8, 10x10) with only
normal pieces, but perhaps more of them. (e.g. 4 Bishops, such that you
also have Bishops on like colored squares, with which you can form
batteries.) This to get them acquainted with different boards.

By offering both, we would discover atomtically which deters ortho-Chess
players most: othe roards, or other pieces. But whatever they refer, it
would make the next step smaller.

What is wrong with PGN (= Portable Game Notation)?

This format contains the name of the variant (or game) in its header tags,
so the exact move syntax could be dependent to the requirements of the game.
For Chess-like games it uses SAN (Standard Algebrac Notation)format to
describe the moves. This is a pretty universal format. (It could also be
used for games like Go, Draughts or Othello, where moves have unambigously
defined side effects, once the move or drop of a single piece is specified.
If a game needs multiple moves per turn, SAN could be extended with a
concatenation operator, say '&', to string together all moves that make
up one turn, and you would already capture an enormously larger number of
games.)

Why re-invent the wheel, if we already have bicycles?

OK, great that you are contributing to the effort! I have not tried to
compile yet. I am just trying to find my way around through the code,
learning which tasks are done in which source files. And I am already
converting the game-specific code where I am sure it has to be converted,
making board width and height variables.

I would not say that the Falcon is difficult to handle for a computer. Just
that 

(1) It is a piece with comparatively many (potential) moves, so naturaly
it requires more effort to generate its move than of pieces with fewer
moves. But it is not worse than, say, a Queen.
(2) It behaves differently than other pieces w.r.t. pins, due to its
multi-path nature. This means it need separate code from the other pieces
to handle it. But this could just as easily be blamed on the other pieces
as on the Falcon. In a game with only Facons it would not be more
difficult than in a game with only other pieces.

It is true that for divergent sliders and lame leapers you have to examine
several board squares to know if you can do a single move to one. This is
also true for distant moves of normal sliders, but there every square you
pass gives you a valid non-capture, to which you could assign the effort,
so you examine only one board square per generated move. But,
unfortunately, almost all search effort of a computer goes into its
capture search, and generating non-captures is a waste of time there.

Well, WinBoard usually saves games in PGN. There is an obsolete dedicated
format, but no one hardly ever uses it, and it is there only for backward
compatibility.

Dear mr. Duke,

As I mentioned here before, my policy is to respect the intellectual
property and wishes of variant inventors. Future releases of my software
will therefore not contain explicit support for any Falcon pieces.

What I already did cannot be undone. Neither do I see any moral obligation
to do so: I asked you for permission to do what I did, out of courtesy, and
you gave it in public on these pages. If you now want to take the stance of
a treacherous and backstabbing business partner... Well, it be so noted,
but I will of course ignore it in my actions.

I look forward to chatting with your attorney. The longer the better. I
hope you pay him well! :-) As far as I am aware of, your patent is not
valid in the part of the world where I live. And even if it were, there is
the tiny matter that the permission you gave will be considered legally
binding. It is immaterial if I spent 3 lines of code or 300,000 on it.
Perhaps my code is so good that it is worth $100,000 a line. ;-)

Well, you can be pretty sure that they will not agree to it if it is not
either upward compatible with, or highly superior to what they are already
using. What you propose on the IAGO site is neither.

For Chess variants, the existing PGN standard, with a few obvious
generalizations (e.g. to allow larger number of ranks and files, and other
letters for piece indicators), seems entirely satisfactory. Even games with
large player bases and a long tradition, like Xiangqi and Shogi, which have
their own notation systems, are starting to recognize he advantage of
algebraic nottion, and are embracing variations on Standard Algebraic
Notation. Now it exists next to th traditional notation, but Chess also
has alternative notations, such as the infamous descriptive notation
('N-KB3'), which are slowly but surely losing ground against SAN.

Where a player base is virtually non-existing, traditonal notation methods
(or in fact any notation method) might not exist, and any system will be
easily adapted.

So in short: to get wide-spread acceptance of a standard, first look to
the main-stream games, how they prefer to do it, and design something that
accomodates their wishes. Otherwise, failure is a guarantee. To a Chess
player, it will not seem an advntage that his saved game can be read by
software that plays Go, so his willingness to compromise to make that
easier will be zero.

If mr Duke's memory span is so short, it might be a good idea to remind him
of his earlier words:
| Muller was asking about Falcon Chess in Joker or Smirf; I forget 
| the exact question and am more interested in relooking at Jeliss 
| and Trenholme articles this week. Just as Greg Strong was about to 
| finish Falcon Chess for ChessV, it is fine to put Falcon in engine
| free of charge throughout years 2008, 2009 and 2010 to play, so long|
| as strictly not commercial (unlike standards-degrading Zillions).
| Please inform what is going on, and put the patent #5690334 two or
| more times about the Rules or Board, since ultimately we would like
| to market Falcon material too. It may be coincidence, but Strong took
| offense at our critique of his grotesque game ''Catalysm'' and never
| returns since. [Muller rates 'Poor' over at Falcon Chess and says he
| has not decided whether it is good. If it is deemed not good, please
| do not include FC, but something not commercial like that can be up
| to you during 2008-2010.]

Note especially the incoherence of his opinions: first he would never
consider the 'standards-degrading' Zillions for permitting it to play
Falcon, and now that he apparently takes a dislike to the imagined acts of
the author of Fairy-Max, he thinks he can reverse the roles at a whim.

Well, George, let me tell you a secret: if people see how you reserve the
right to revoke your agreements, no programmer in the world would ever be
prepared to implement Falcon Chess. Knowing that you would be likely to
cause them having done their work in vain, from the first day you happen 
to step out of bed with your wrong leg first. They would not touch anything 
that depended on permission of such a deceitful business partner.

The bottom line is this: you gave permission for me to put up a
non-commercial Chess program capable of playing Falcon Chess on my website
upto and including 2010, and I intend to do so. I don't think that
violates any rules of the CV website, in which case I will maintain links
to that page from here.

It seems you badly need some lessons in manners and civilized behavior, 
and I will be more than happy to teach you! In court or otherwise!

George Duke:

| He never responded at all. That and inveterate criticism revoked
| permission. Chess Variant Page I am sure will not permit links to
| Muller's by mid- to late-2009 and 2010 indicating licensing
| unapproved. At Falcon Chess we want nothing to do with him for
| months now since when he started his rants. However, Muller is
| welcome to enter other topics on CVs here intellectually and look
| forward to debate. 

You seem to have a big attidude problem. In particular, you seem to suffer
from an affliction known as 'delusions of godhood'. So you think you can
make an agreement, and then simply revoke it because the person you made
it with criticizes something you say, or uses a standard term you don't
like, or perhaps simply because you are in a foul mood and want to take it
out on someone? You must be out of your mind to think you can treat people
that way, and even more out of your mind to think they would let you get away with it.

Too bad for you that you have picked the wrong person to act out your
omnipotence fantasy. There was no legal requirement for me to respond to
your granting of permission, but in fact I did, by several posts, and, of
course, by actually making the engine. There is no way you can revoke
that, I will not allow you to weazle out of our deal. Fairy-Max will
remain on my website including the Falcon Chess preset until the end of
2010, as you agreed to. And I am sure CV pages will have no objection that
the links to it remain here for as long as I wish, and if they don't, I
will hear it from them, not you. If you want to renegociate, fine, but I
will only do it through your attorney, as I am fed up with you.

I won't be coerced in what to say and what not to say by you, or anyone
else!

Well, I am not seeking conflict on this. The engine I wrote has helped me to conclude that the Falcon is a very interesting piece, in a value range that makes it combine very well with orthodox Chess pieces. It would be a a pity if it would go to waste due to personality conflicts.

I just want to point out that I have acted in good faith from the very start, to help creating a situation that is a win/win/win for everyone, and that I don't like to be accused of stealing intellectual property, and even less to get covert threats of lawsuits.

Censorship on what I could say on this site was never a part of the deal.

OK, I claered up a lot of the confusion I was subject to. It seems that I had
been downloading the chesd C sources from a location that described an
entiely different, PHP-based server also called chessd...

I have found a chessd version now which is called Lasker-2.2.3, which does
include a correct description to install it. Indeed I had problems
compiling it: apprently the Ubunt compiler is more strict than the one
Lasker was developed with. So you get compiler errors.

All errors are basically the same: some global vaiable is delared as
'satic ...', while it is in the included header file as 'extern ...'.
Apparently, this is no longer allowed. I cold solve it by simply deleting
all the offending 'static' keywords (in 5 different source files occurs
one, IIRC). After that it compiled.

You have to be superuser to do the 'make install', something the README
fails to mention. But then there were no more poblem, and I could run the
server according to the instructions, and login on it from another
computer on my intrnal network. Looks just like FICS, except the place is
deserted, so you have no opponents! ;-)

So I guess I will start hacking this one, to expand the board size, and
add some more piece types.

David Paulowich:
| When the King and Rook are widely separated, the winning move in this
| endgame involves moving the Queen 'x' squares, checking the King ('y'
| squares away) and attacking the Rook ('z' squares away). Usually at
| least one of x, y, z is greater than four, which suggests that the
| Short Queen would not win against the Rook in an endgame with only 
| four pieces on the board.

Your conclusion is correct: King + halfling Queen vs King + Rook is in
general a draw. Only 3.8% of the btm positions are won to white, and 40% of
those because black is already checkmated. There are a few lengthy mates,
though (where the black King starts trapped in a corner), the longest
taking 44 moves to mate or Rook capture.

I am not sure if this suppresses the value of the halfling Queen as much
as you think, though. KQKR end games are not that common. I play-tested
the 'Quareter Queen' (FWAD, which I believe is also known as Mastodon),
and in end-games with only King and Pawns as other pieces, it seemed te
Mastodon  was very close to being exactly halfway between R and Q. I would
expect the halfling Q to be better than halfway between Mastodon and
Queen.

Commoner was only slightly stronger than Knight, perhaps a quarter Pawn,
in endgames of two Knights against two Commoners in the presence of King
and (many) Pawns. So I got the end-game values

Knight   = 325 (Kaufman)
Commoner = 350
Rook     = 500 (Kaufman)
Mastodon = 750
Queen    = 975 (Kaufman)

and, based on this, would estimate the Halfling Queen somewhere around
875-900.

It is not that bad. Depending on where exactly you start the Falcons on the back rank, it is possible to develop them laterally. E.g. on a 10x8 board a Falcon can move from i1-h1-g1-f2 or h1-g1-f2-e2/3 to the squares in front of K or Q after you evacuated the B and N from the back rank. This was the solution usually preferred by Fairy-Max in the Falcon Chess demo-match games it played. It is true in general that 'castling early' is in general difficult in 10x8 variants, and that many games went without castling.

Note that a single Falcon does have mating potential against a bare King, and is a close match for a Rook both in opening and end-game.

I could not find a description of your Falcons anywhere, so I am not sure
if they could be implemented in Fairy-Max. If they could be, it would be
easy to evaluate them by self-play of Fairy-Max from positions with
material imbalance.

About R vs C=(RN) and A=(BN) you are again correct. KQKR is generally won
(99.4% wtm, 46% btm), but KCKR is in general a draw (84% won with wtm,
1.6% btm), and so is KAKR (63% won wtm, 3% btm). It might sound strange to
descibe an endgame where 84% of the positions wins with wtm as generally
drawn, but the won positions are almost all positions where you either
immediately capture the black King (36%) and a very similar number of
positions where you capture an undefended Rook. That leaves only about 15%
of the positions (there is some double counting), and most of those are
positions where white can give an immediate fork or skewer. Normally, you
would not consider any of those positions as KCKR, as it is obvious that
the Rook is tactically doomed from the very start. If you can't gain the
Rook in on or 2 moves, it is almost always draw.

A Bison loses against any of Q, C and A. KAKBi is 98.3% won with wtm, and
52% with btm. The others do even better.

All this on 8x8.

David Paulowich:
| There may be some forced checkmates that involve a Bison leaping 
| over one of the Kings, but I assume that a Falcon can still force
| checkmate, possibly taking a few moves longer.

The patented Falcon is actually 100% equivalent to the Bison in the KBiK end-game, because of its multi-path character. No Falcon move can be blocked on a single square; at least 2 blockers are needed. But the blockers wil then always have to be on adjacent squares, and two Kings a never do that. So KFK can be played aexactly as KBiK.

In end-games with more than 3 pieces this is no longer true, and a Falcon might be weaker than a Bison. In KQKBi, KCKBi and KAKBi this is of no importance, as they are aready lost for the Bison, and can only be more lost for the Falcon. KRKBi is a dead draw, and downgrading the Bison to a Falcon does not change that: KRK is only marginally won, and only by virtue of zugzwang. Any extra piece for the defender spoils the zugzwang, and thus make the K+R attack on K unsuccessful. And a Bison is far too agile to be hunted down by K+R. Even KRKN is almost always draw because K+R cannot get a Knight. The other way around, a Rook can defend quite easily against Bison by cutting off the opponent King from your own.

This piece is in the Piececlopedia under the name 'Buffalo'.

8x10 = 8 files and 10 ranks, and you start counting ranks at 1? This indeed is a mate in 28, with white to move.

But I don't think you can derive the value of a piece from how long it needs to checkmate a bare King. Indeed, there are pieces that cannot mte  bare King at all, and yet are worth more than a Rook. A Nightrider, for instance. A more obvious case would be a piece that moves like a King, but captures to any square of the color it is on. I am pretty sure this must be worth more than a Queen.

The Bison must be stronger than the Falcon, but I did not develop a etup to play-test it yet. If you start with a conventional array (pawns on 2nd and 7th rank), Bisons make it into a highly tactical position, as they can fork lots of trapped or undefended pieces on the back rank, before you have tme to develop them. Grasshopper play-testing also suffered from this, but after some experimenting, it turned out that starting all Pawns on 3rd and 6th rank was a very good remedy against it. I guess that for Bisons this should work too, but I haven't tried it yet.

My guess would be that the Bison would gain 100-200 centiPawn compared to the Falcon, as the latter is not that easy to block. This would put it at 6-7 pawns.