Comments by Joost Brugh

I played it a few times. I think I figured out the algorithm.

After placing the tiles, there are 12 ranks with total number of squares
64. Define: Area(n) = Number of squares on ranks 1..n. As there are 12
files, Area(1) = 0..12, Area(2) = 0..24 etc. Area(12) = 64. White may
place pieces on rank r if Area(r) is 32 or less. For Black, it works the
same, except that rank 12 is now rank 1 etc.

For example, if White and Black construct a Chessboard on ranks 2..9 (with
eight squares on each rank. Then, for White: Area(1) = 0, Area(2) = 8,
Area(3) = 16, Area(4) = 24, Area(5) = 32, Area(6) = 40 Area(n) > 40 for
n>6б, so White can place pieces on ranks up to 5. For Black the same
results in ranks from 12 down to 6.

If Area(n)=33, you just can't place pieces on the n'th rank. The maximum
number of squares on the n'th rank is 12, so Area(n-1) must be at least
21. This is enough space to drop the 16 pieces. The game, however gets
stuck if you have to drop your King into check.

I don't know. But I got my ass kicked by Zillions several times. But I
don't know any DVONN tactics, so probably it is just a sign that I'm
still not good at it.

My ZRF is built with the idea that the only things that matter for a stack
are its size, its owner and whether or not a DVONN-piece is in there. I
didn't bother about stacks larger than 25 (because they and larger stacks
are equally immobile and equally winning when surviving). Rules like the
'No move with enclosed pieces' are trivial to implement. After each
move, an administrator (?-player) must remove all disconnected pieces.

I used a pass-detector that detects when players pass. Then I create dummy
pieces to make high stacks count for that many pieces and then carefully
trigger the count-condition.

But I think that the maximum height of a stack is more than 25. Take the
leftmost positions as building position. Then try to get stacks with
heights 1, 2, 3, 4, etc. on positions on the center row. From four other
positions, stacks can directly be moved to the target. Still, I don't
think the answer is 49 (or 46 for a DVONN-less stack), but probably they
are close.

The Harpy is not just a piece, but a whole idea. Like in Shamanic Chess,
the piece can go in move-mode or in fight-mode. In move-mode, they are
more mobile and in fight mode, they can capture. You can make a whole
chess variant (or a 'Chess with Different Armies'-army. For example,
strong fighting pieces that can't move to any square in move-mode, but
for example just like a queen. Or pieces that have an effect when
deployed, but can't move then (fight-mode), for example the
Ultima/Rococo/Maxima-Immobilier (that does not immobilize in move-mode and
does not move in fight (immobilize)-mode. There are many possibilities.

A problem for the real mathematician about DVONN: How high can a stack
maximally grow in the game when:
a. The stack contains one or more DVONN-pieces (so it will survive by
definition)
b. The stack contains no DVONN-pieces
bi. The stack dies later in the game by becomming disconnected
bii. The stack gets a DVONN-piece later in the game (and thus survives)
biii. The stack survives without getting a DVONN-piece

Some one-dimensional examples. Assume these lines as isolated islands: A
number is a stack without DVONN-piece, X is a high, immobile stack without
DVONN-piece, D is a single DVONN-piece.
bi: X - 1 - D. The singleton has to move and stack X dies. It can only
become X+1 upon dying. The X+1-stack never really lived.
bii: X - 1 - D - 1. The stack can only be saved by using the rightmost
singleton to put the DVONN-piece on the stack and it can grow to X+3 with
a DVONN-piece, but had highest size X without DVONN-piece.
biii: X - 2 - D. The stack stays connected and survives.

Note that White and Black play together to get the high stack. But the
rules must be obeyed. The problem can be simplified by disregarding one or
movre rules.

Could there be a systematic way to solve this problem.

This is not for making a ZRF. I already made an ugly ZRF in which I used
25 as maximum, becuase when higher stacks are brought back to 25, the same
moves are possible and the outcome (win/loss/draw) will always be the same.
Only the point difference can be different.

A very important point in Pawns. In FIDE chess the Pawn skeleton is a key
strategic element. Pawns on adjacent files protect each other. The idea
'Pawns are the soul of chess' certainly applies more for FIDE chess than
for Xiang Qi. Piece strategy in the middle game and in the endgame are much
related to Pawn structure. The Pawn structure defines your playing space in
the middle game. If you want to penetrate through the opponent's Pawn
fortification (with brute force), you have to sacrifice at least a piece
with thrice the value of the Pawn (Knight or Bishop). In Xiang Qi, a Pawn
isn't worth much less than an Elephant (at least when the Pawn moved
twice, getting it across the river). In the endgame, Pawn promotion is a
much bigger issue in FIDE chess. With little material the mobile FIDE King
isn't easily checkmated. The idea of the endgame is to use the King as an
attacker and the goal is to get a Pawn across the board. In Xiang Qi, the
goal of the endgame is still to attack the King, not to eliminate Pawns
with the King.

For two Camels, two Knights or a Wildebeest we can prove that the longest
forced mate is one move, because the geometry allows only a few mating
patterns. This two-move example with two cannons is not trivial. Is it
possible to prove that two moves is the maximum. Known is that the mating
patterns is always with the Black King on the side (X1), White's King on
X3 or on b3 against when X = a. And White's cannons are at Y1 and Z1 with
Y between X and Z and Y not adjacent to Z. The last move is a vertical move
by a cannon (C YA-Y1 or C ZA-Z1). Blacks last move is a horizontal King
move, which can only be forced if the end file is involved (Second rank
squares can only be covered by the White King), so this must be Ka1-b1,
which means that X = b. One retromove by a cannon later, c1 must be
covered. This is impossible. With two (Cannon + passive Bishop)-pieces
(passive Bishop is a Bishop that does not capture), it should work (from:
White CmB on c4 and c5, White King King b3, Black King b1, Black to move)
1. ...,Kb1-a1 2. CmB c4-f1, Ka1-b1 3. CmB c5-g1#. Probably (not certainly,
it should be possible to force this with two CmB's and a King against a
lone King.

It would be interesting to prove this (and of course the
King+Cannon+Knight against King)

By the Way: The possibility with both the Wildebeest and the King on b3
would be stalemate anyway.

For the Camels: the Checkmate positions are:

White King (a3 or b3), White Camel (b4 or d2) White Camel (a4, c4 or e2)
Black King a1
White's (b4 or d2)-Camel moved last, so before that move, the Camel was
on (c1, e3, e5, c7, a7, a3, c5, g3 or g1) (No a1 because of the Black
King).
Black moved Kb1-a1:
Position before that:
White King (a3 or b3), White Camel (c1, e3, e5, c7, a7, a3, c5, g3 or g1)
White Camel (a4, c4 or e2) Black King b1.
Blacks Kb1-a1 must be forced, so c1 must be covered by a Camel (The King
doesn't cover it). So there must be a Camel on b4, d4 or f2, but there
isn't. So no mate with King + two Camels against lone King.

I knew that the move-priorities-trick does not work. Zillions resolves
move-priorities before the checkmate condition. So it first concludes
there is a normal move and then thus renders the special move illegal
versus the move-priorities. And then it renders all normal moves illegal
because of check. A possible solution is to take into account that
defenders can be pinned by Bishops, Rooks, Dragon Horses and Dragon Kings.
It is easy to implement (A defending capture move is either like a Bishop,
like a Rook or like a forwardmost Knight, the relative position of the
mating Pawn with respect to the mated King is fixed. This leaves six
pinnable positions for defenders (four diagonal and two orthogonal).

However, this does not solve the problem. It just reformulates the Pawn
drop rule to 'A Pawn may not be dropped to check the enemy King when ...
(long formulation involving the geometric explanation of some specific
pins) ...'. It should be 'A Pawn may not be dropped to give checkmate'.
In Shogi these rules may have the same effect, but it doesn't give a
checkmate-detection that always works. If someone wants to use Shogi.zrf
to make a Shogi variant with some different pieces, he or she can never
know that the Pawn Drop Mate rule is well implemented. The same problem is
there for Tamerlane 2000, where Princes can become Kings when the original
King is mated (It is not implemented because detecting checkmate in
Tamerlane 2000 is a nightmare). Another example is 'Thirty-Nine squares
Chess' where you may leave your King in check, but you lose if you are
mated (Kings return when captured). I have an ugly solution for the last
example, but the ZRF is still too ugly and buggy to publish.

A (dirty) solution would be that the Pawn Drop Mated player can declare
checkmate after a pawn drop. On such a declaration, the whole position if
flipped (A Black Gold on 3f becomes a White Gold on 7d, etc) and the
player that dropped the Pawn is automatically checkmated if it were
checkmate, but that player should win the game if he or she can continue
with a legal move (Penaly for a false declaration). It takes a while to
implement. You have to know whether the opponent just did a Pawn drop, the
flip mechanism must be implemented. The flip must be registered (for
instance by dropping a Sign piece on a dummy position). These Sign pieces
should also enable a 'death penalty'-move if the dropping player manages
to prove that it isn't checkmate. Anyway, it really fucks up the ZRF just
to use the (checkmated ...)-command in a different context then ending the
game.

Like Christine, I think that Miserable and the stuff below is useless. What
is the difference between a 'Hideous' and 'Loathsome' item? Is the
'Hideous' one better? I think both 'Hideous' and 'Loathsome' (and
all those low rating) means that the item is has no value. If you think a
submission is 'Loathsome', you should say what the problem is and so
increase the chance that the next submission of the same inventor is not
'Loathsome'. Not add a negative atmoshpere by crying 'Loathsome!!'.

About specifications like Playability: Neutral, Graphics: Good, etc. I
think it is good enough if those specifications are said in the comment
text. At least if I can say something between 'Good' and 'Poor', it
should be fine. I think that ratings are less important than the comment
text.

A good thing about a new system is that there is something rated between
Good and Poor. If I rate something 'Good', I think it is Good and not
the second best out of three possible ratings.

A problem with popularity lists is that it could become a competition.
Comments should be used to give feedback, to ask and answer question or
something like that. Feedback and discussion can be used as inspiration
for new projects. I think it shouldn't be about who scores the most
comment-points.

This game is very good. I first thought the Transmitter would dominate the
game, but that is certainly not the case. Especially the endgame can be
very interesting with Kings interacting with Courtesans and Transmitters.
I like the idea a vulnerable piece (the King) is required to fulfill
certain tasks (Guide Courtesans, protect and attack the Transmitter)

On a moment I saw that my opponent had three Chakra's and I had only one.
The problem is that a Chakra becomes an enemy Chakra if you use your King
to move off yor Courtesan from your Chakra. I found the bug in the ZRF:

On line 347 (in courtslide-copy) '(slide-ecopysub' should be
'(slide-copysub'

If you go to Game Courier (Via Chess Variants main page -> Play -> Game
Courier). Click on the Editor. You can also reach this page by going to any
Game Courier preset and choose 'Edit this preset'. Then, you see a table on
which the game is defined. If you set the 'Set group' to 'Chess', then you
can set the 'Set' to 'Galactic Graphics'. Then, click on update and the
game pieces will change into the Galactic graphics. If you then scroll
down, you see a whole alphabet of pieces for both White and Black. You can
see other piece sets as well.

Update: I see that my link didn't work and that Christine had the link to an
even more complete set of Galactic graphics. I used only the 26 from Game
Courier. Nice :)

I think such ideas fits well in the idea of a contest. A contest should
make people creative by giving a thing to start. It is hard to just create
a game from scrap, but if there is one idea given, it is much easier to be
creative.

Previous contests were almost all about a number of squares on the board.
It would be interesting to begin a contest with another given idea than
number of squares. An inventive goal would be a nice example.

...
(capture ThisAndThat)
add add add add
...

ThisAndThat is only captured in the first add. This action is canceled after that add. The three other add's do a move without performing the capture-action. Maybe it also works if you even further fool Zillions by adding 'completely useless add's'

(piece
  (name Coordinator)
  ...
  (moves
    (move-type Regular)
    (CoordinatorMove n)
    ...
    (move-type Monkey)
    (add add add add add add)
  )
)

And say in the turn-order that White and Black do only Regular moves. I don't know much about Zillions AI. But maybe this works.