FRC standards (again)

[Richard Allbert]

Hi,

I've been hunting through the winboard and talkchess forums this morning, as I'm trying to add FRC to Jabba.

It would be nice to know all of the FEN castling permission inputs Jabba should understand... so far, I understand this to be

1. any of KQkq, where the outmost rook is taken as the castling rook.
2. any of AaFf where the letter describes the file of the castling rook
3. A mix of the two above.

If 3. is true, is that any mix? e.g kAhq or mixed by side e.g HAkq ?

Is there anything else?

Thanks!

Richard

[Sven]

Hi,

I've been hunting through the winboard and talkchess forums this morning, as I'm trying to add FRC to Jabba.

It would be nice to know all of the FEN castling permission inputs Jabba should understand... so far, I understand this to be

1. any of KQkq, where the outmost rook is taken as the castling rook.
2. any of AaFf where the letter describes the file of the castling rook
3. A mix of the two above.

If 3. is true, is that any mix? e.g kAhq or mixed by side e.g HAkq ?

I have never heard about 3., could you quote where you found this? IMO either 1. or 2. is possible.

Sven

[Richard Allbert]

Here

http://en.wikipedia.org/wiki/X-FEN

giving this

Code: Select all

rn2k1r1/ppp1pp1p/3p2p1/5bn1/P7/2N2B2/1PPPPP2/2BNK1RR w Gkq - 4 11

near the bottom of the page

So I just need 1. and 2. ?

[hgm]

Well, if you would make the engine understand 1 and 2, it would automatically understand 3. What would you gain by adding code to prevent it from understanding 3?

Note that WinBoard always sends A-Ha-h in FRC. But if your engine also plays variant normal, you would have to understand the KQkq anyway. You gcould get away with the fixed association K=H, Q=A. This is not anchored in the protocol, though; other interfaces could use XFEN, where K can indicate an outermost Rook that is not in the h-file. So to be fully WB compliant, the engine would have to understand KQkq as the outermost Rooks, next to A-Ha-h.

[Richard Allbert]

I've read this complaint in your other posts.

When reading the FEN, I have

Code: Select all

        if(EPDline has  'q' or 'k' or 'Q' or 'K')  { setARENAcastlebits(epdsections); break; }
        else if&#40;EPDline has >= 'a' && <='h' or >='A' && <= 'H') &#123; setSHREDDERcastlebits&#40;epdsections&#41;; break; &#125;

Where "epdsections" is is a vector of the fen string split into parts.

So I have two distinct functions to set the castle bits. Thus a mix of file description and KQkq would lead to an error.

Of course, I should just process each letter at a time, and then decide.

Richard

[Richard Allbert]

In addtion, the AHah solution seems much more elegant.

[hgm]

Indeed, and fortunately it still works for variant caparandom, where the file can run upto J, without collision with KQ. In Courier Chess the board is 12 wide and files runs up to L. But there is no castling there.

[Richard Allbert]

Well, I've played a few games vs Jabba in the shredder interface and Winboard, and all seems ok.

Fairymax doesn't support FRC, winboard tells me. That was shock - I thought it played everything!

Is there anything special needed to play variant chess using Zippy?

I'll put Jabba on the ICC playing FRC for a while, and see what happens.

Richard

[hgm]

You will have to start WinBoard with the option -zippyVariants "normal,fischerandom". I am not sure fischerandom is in the default setting for this oprion.

You might also have to do something on the ICS to tell it that you are open to challeges in variants.

Indeed, none of my engines supprt FRC so far. The castling is a problem, I should equip Fairy-Max with more general castling code. Currently it is possible only to program castlings of 2 or 3 King steps with corner Rooks. I don't like FRC much, though. It s not really a variant. And it retained all the bad properties of normal Chess (high raw rate, boring), while throwing away the elegance and beauty of the design.

[Richard Allbert]

Well, now FRC and Knightmate are done, Chinese is next on the list!

Richard