Comments/Ratings for a Single Item

I tried your PawnX.zrf (computer vs. computer). Following happened:

1.Nf3 Nf6 2.e4 N:e4 3.h3 N:d2 4.K:d2 g5 5.N:g5 Bh6 "Stalemate!" Black
wins.

Is this what you expected?

Hi Alfred, thanks for trying PawnX.

Yes this is what I expected, consistent with the rules. After Bh6, the
white knight should capture Pf7 or Ph7 but he can't, otherwise the white
king would be in check. So white is in stalemate, losing the game.

In all my tests (AI vs. AI) the games never ended in so few moves.
Moreover, games usually end with a checkmate. What thinking time did you
set for the AI? In PawnX, AI has to look ahead more so it needs more
thinking time.

Did you do other tests? How did they go?

P.S. For now, I have noticed is a losing strategy to develop non-pawn too
early in the game. You should move more pawns in the opening, setting traps
for the enemy. Yes, I see pawns as traps.

I'm implementing a variant with bishops, rooks and queen move up to 4
squares. I think this will be better because decrease the "pawn appeal"
factor.
I will let you know how it goes.

Anyway, I need a little help with zillion programming. How can I implement
the rule "if the king is attacked ..." before a move of a non-king piece?

Hi Winther,

>>>If you want to know whether your king is threatened, you'll have to
search out where it is, and then check this square with the "attacked?"
command.

Yes, I need to know where the king is, but I don't know how to do that.
Could you please explain it to me more in detail?

I need this piece of code because I want to change my first rule in this
one:

1. if possible and if the king is not in check, it is mandatory to capture
an enemy pawn


Thanks.

In that case, provided that you have win-condition checkmate set, you
don't need to check whether the king is checked, because Zillions does
that automatically. Zillions forces you to make a move that protects the
king, and you cannot make another move.

So it is easy to enforce the rule that an enemy pawn must be captured.
Simply add another "move-type" in every piece definition. This move type
is exactly the same as the normal move-type except that (1) it is set to
higher priority (2) you must verify before the move is executed that an
enemy pawn is captured: "(verify enemy?)(verify (piece? Pawn))". 

This will enforce the capture of an enemy pawn provided that the king is
not in check. It will still enforce the capture of a pawn if the king is
protected thereby. However, this technique will increase the value of the
pieces relative to the pawns. So you need perhaps to tweak the value of the
pawns (and other pieces, too) by increasing the number of "adds". You
could use my technique of "tweak-shift", which is simple. Zillions tries
to evaluate the pieces, but this is highly complicated, and it is necessary
to improve this evaluation by tweaking the value of pieces. This improves
the playing strength very much. My version of Chinese Chess beats the
Zillions standard version every time. It is much better only because I have
given the pieces a more appropriate value.
M. Winther

**M. Winther wrote:** Zillions forces you to make a move that protects the
king, and you cannot make another move.

So it is easy to enforce the rule that an enemy pawn must be captured.
Simply add another "move-type" in every piece definition. This move type
is exactly the same as the normal move-type except that (1) it is set to
higher priority (2) you must verify before the move is executed that an
enemy pawn is captured: "(verify enemy?)(verify (piece? Pawn))". 

This will enforce the capture of an enemy pawn provided that the king is
not in check. It will still enforce the capture of a pawn if the king is
protected thereby.
***

I have done this procedure, as you can see in the zillions file I have
posted in the first post of this discussion. I have a higher priority
move-type, called “tasty-moves”, for each piece that check if it is
possible to capture a pawn. If this is not possible , then you can make
“normal” moves.

But there is a problem, a big problem. If the king is in check and you can
capture a pawn, for Zillions’ AI the only possible moves are those in
which you capture a pawn, but with those moves you cannot get the king out
of check, so it is checkmate. Not what I want.

There is another problem. The situation is that a knight is protecting a
king from an enemy bishop: white has Ke1, Nd2; black has Bb4. But there is
an enemy pawn in e4, so the knight can capture it. Now, I want that the
knight doesn’t move because if it do so, the king would be in check. But
for Zillions’ AI there is no possible move, so it is stalemate. Not what
I want.

Any ideas?

p.s. I sent you an e-mail.

Sorry, I didn't know that it worked that way. It is generally recommended
for newcomers to Zillions programming to begin with a simpler game. This is
a very, very, complex game. I suppose, the simplest way to solve this is to
introduce yet another move-type which has the highest priority of all. This
move-type is exactly the same as the normal moves (where pawn capture is
not enforced). The only difference is that you must verify that the king is
threatened before execution of this move-type is allowed. So you must loop
through the board squares until you find the friendly king, and verify that
it is threatened. Thus, only if the king is attacked, this priority move
will be executed, and the player can thus make a move that is the same as a
normal move. Should there be no move that can save the king, then Zillions
will signal mate, which is correct. However, for the king itself, it's not
necessary to make this loop, because you can check (verify attacked?) where
it is placed.
/M. Winther

Matteo, it is the search tree which is the heavy operation. The problem is
here that the king search is involved in the search tree, which has
thousands of branches. You should consider changing the rules a bit to make
it easier to program, until you know how to do it. Anyway, these 'create'
and 'capture' commands do nothing if not followed by an 'add'. After
the last 'create' there is no 'add', so it won't happen. You probably
only need an 'add' in the last position, and that cascade seems
superfluous(?). You don't need to 'capture' the king-check piece. It is
superfluous. You could also use 'change-type' instead of 'create', to
avoid a capture sound. 
/Mats

Mats, with "add" the program still doesn't work. Anyway, I'm following
your advice and simplify the rules. ;-)

Matteo, 

there are some curios effects with your PawnTrap.zrf.

1) King may move into check if there are possibilities to capture a pawn.
   Example:  1.d4 Nf6 2.e4 Nxe4 3.Kd2?! Nxf2

2) No defense against checks if there are possibilities to capture a pawn.
   Example:  1.d4 e6 2.b3 Be7 3.f3 g5 4.Nh3 Bb4# 
             "Checkmate: Black wins!"

I assume this is not intended by you but a weakness of your zrf.

Hi Alfred,

thanks for taking the time to test my game.

1) In that situation you can move the King in d2 because d2 is not under
attack of the Knight. For each piece, if a piece can capture a Pawn, that
piece have to capture that Pawn. So, in your situation the King is not
moving into check.

2) This is a problem. If you read the previous messages in this topic you
see I have tried to solve this problem, without success. I'm not able to
implement the rule "if the King is in check, capturing a Pawn is no more
mandatory". If you can solve this problem I would be grateful to you. :-)
Anyway, I have found without this rule the game is playable as well.