Winboard Forum

Chess Engine: Dabbaba 0.98wb JA (winboard version)
Author : Jens B‘k Nielsen
Website : http://www.jens-musik.dk/dabbaba.htm


Dabbaba 0.98wb JA
-----------------
1. Winboard protocol 2 support.
2. Added 'resign' function.
3. New logo.


Instructions:
-------------
Just install as a standard winboard chess engine.
No extra command line switches needed.

Info:
----
Here is Dabbaba 0.98wb by Jens B'K Nielsen.
Dabbaba (Arabic meaning 'war machine') is now a winboard chess engine.
Originally written in 1993, the last update/release of original (dos) Dabbaba was Nov 06 1998.

Dabbaba original also supports variants, but they are not
to my knowledge variants supported by winboard, so my winboard
version is 'standard chess' only. (also I'm not that interested
in variants, and it would be a lot more work to implement!)
I estimate it's strength around Tscp level.

Things still to do:
------------------
1. Enable ponder. Original Dabbaba supports ponder, but it is disabled in
this version until I can make work with winboard.
2. Fix the 'force command'. I've found it hard to implement the winboard
'force' command so for now it is disabled. That means using chess gui
opening books won't work at the moment.
3. Clean the pv output up. There's a lot of unneeded stuff being spat out.
4. Make time allocation more accurate.

Available from my homepage.

Have fun,
Jim Ablett.

Jim Ablett wrote:Dabbaba original also supports variants, but they are not to my knowledge variants supported by winboard, so my winboard version is 'standard chess' only. (also I'm not that interested in variants, and it would be a lot more work to implement!) I estimate it's strength around Tscp level.

Would it really be that much more work to support it? It seems from the description that most variants are little more than normal Chess with some strange pieces in it. And in WB protocol the piece indicators are hardly ever used anyway (the moves sent forth and back only contain the squares, if they are not promotions).

So it seems hardly any support is needed at all, except for a method for somehow telling the engine that it is now playing the particular variant.

WinBoard actually does support "variant knightmate", and it supports variants with the mentioned pieces in it under the collective name "variant fairy". Would it be very hard for you to make sure Dabbaba lists those two variants (knightmate and fairy) in the features command (if you rigged it as a protocol 2 engine)? Or at least retain some way to tell the engine it should play any of the variants it knows, even if only through a command-line argument to the engine?

From the description of the variants that seems pretty much all the support it needs: if it says for 'Chancellor Chess' that Queen moves as R+N, there is really nothing that WB has to know about, except that you would have to play with /testLegality=false. It could still use 'Q' for indicating the Chancellor, when setting up a position through FEN or 'edit'. (Although it would be nicer if you could make it such that it would consider 'C' or 'A' synonyms for 'Q', and 'H' (nightrider) a synonym for 'N', in 'edit' or FEN. Assuming it always uses the same representation internally for the exo-pieces as for the FIDE counterparts they replace, and just makes them move differently in the various variants.)

It seems to me that the major attraction of this engine is the fact that it does play these variants, and it would be a pity if this possibility was lost. TSCP-class normal engines already come in several hundreds... I would be interested to have it play Knightmate against Fairy-Max. I have no WinBoard-compatible opponents for that now.

H.G.Muller wrote:

Jim Ablett wrote:Dabbaba original also supports variants, but they are not to my knowledge variants supported by winboard, so my winboard version is 'standard chess' only. (also I'm not that interested in variants, and it would be a lot more work to implement!) I estimate it's strength around Tscp level.

Would it really be that much more work to support it? It seems from the description that most variants are little more than normal Chess with some strange pieces in it. And in WB protocol the piece indicators are hardly ever used anyway (the moves sent forth and back only contain the squares, if they are not promotions).

So it seems hardly any support is needed at all, except for a method for somehow telling the engine that it is now playing the particular variant.

WinBoard actually does support "variant knightmate", and it supports variants with the mentioned pieces in it under the collective name "variant fairy". Would it be very hard for you to make sure Dabbaba lists those two variants (knightmate and fairy) in the features command (if you rigged it as a protocol 2 engine)? Or at least retain some way to tell the engine it should play any of the variants it knows, even if only through a command-line argument to the engine?

From the description of the variants that seems pretty much all the support it needs: if it says for 'Chancellor Chess' that Queen moves as R+N, there is really nothing that WB has to know about, except that you would have to play with /testLegality=false. It could still use 'Q' for indicating the Chancellor, when setting up a position through FEN or 'edit'. (Although it would be nicer if you could make it such that it would consider 'C' or 'A' synonyms for 'Q', and 'H' (nightrider) a synonym for 'N', in 'edit' or FEN. Assuming it always uses the same representation internally for the exo-pieces as for the FIDE counterparts they replace, and just makes them move differently in the various variants.)

It seems to me that the major attraction of this engine is the fact that it does play these variants, and it would be a pity if this possibility was lost. TSCP-class normal engines already come in several hundreds... I would be interested to have it play Knightmate against Fairy-Max. I have no WinBoard-compatible opponents for that now.

Hi Harm,

You make a strong case for variants
The src is there (both modifed & original) if you wish to put variant capability back in yourself
and I'm sure you could do it a lot better than me as you're a much more skillful coder. One problem though,
I stripped out most of the variant code when I was porting it, so you'd have to compare both sources and replace
what I threw out or it maybe easier to do the reverse of that and just add my changes to the original src.

Jim.

Well, if you still need to do work on it, it seems pointless to try repair what you deleted, as it will be gone again when you produce the next version...

H.G.Muller wrote:Well, if you still need to do work on it, it seems pointless to try repair what you deleted, as it will be gone again when you produce the next version...

OK. I'll put variant support back in at next update.

Jim.

Hi Harm,

I've added the variant support back into Dabbaba.
Download a test build here > http://www.zshare.net/download/118332456ba00455/

It wasn't as straightforward as I thought it would be as Dabbaba normally requires two separate builds - one for normal chess and one for variants, so I had to change quite a lot for them to co-exist in the same exe.

Following your suggestions I created an ini file to select the variants that can be used with 'fairy' and also internal variant support for all the others including 'knightmate'. I added currently unsupported variants also. You never know, they might be supported in the future.
I'm a convert! I've just been playing 'knightmate' against Dabbaba and it's great fun - seems to be working ok as far as I can tell, but It's early days and I'm sure bugs will surface.

best,
Jim.

OK, thanks. I will have a look at it. Perhaps I can setup a public Knightmate match between Dabbaba and Fairy-Max, for live viewing on my Battle-of-theGoths website. But this weekend I will be out of town.

And yes, Knightmate is an interesting variant. A Knight is still weak enough to be checkmated with reasonable ease. There are some differences with normal Chess though. My EGTB generator is not very picky as to which piece is Royal, so I could also generate tablebases for Knightmate. This shows:

NQN is won, but NRN is draw. So a Rook no longer has mating potential.
NKN is draw (like its normal 'equivalent' KNK)
NBN is draw (as usual).

All 3+1 4-men end-games are won (NBBN, NBKN, NKKN, NRKN, NRBN, NRRN). In particular, NKKN, the 'equivalent' of KNNK, which is draw in FIDE Chess, is now won.

So Rooks get weaker, but the equivalent of the Knights are stronger in the end-game. Queen is still overwhelmingly strong. NQNR, NQNB, NQNK are all won.

Hello HG,

Sorry for hijacking this thread, but did you ever see my bug report
6 weeks ago in the Winboard_F thread or did you get my PM?
(I don't have any email address of yours...)

Regards,
Guenther

New version of Dabbaba available on my homepage.

what's new:

1. Experimental variant support (currently 'Knightmate+Fairy')
You will need to use HG Muller's 'Winboard-F' gui for this.
2. Improved time-allocation.
3. Adjustments to evaluation.

Jim.

Guenther:

Yes, I got it. Sorry I didn't reply, but I had no time to work on WinBoard_F at all, the past month. As soon as I fix it, you'll be the first to know!

New Update:

Dabbaba 1.0wb JA
----------------
what's new:

1. Found and fixed a few bugs relating to winboard variant support.
2. More evaluation/search tuning.

Jim.