DanaSah vs GM Mario Gomez (Chess960) 2 - 0 (Blitz 20')

[pedrox]

Sunday I had the opportunity to confront DanaSah against GM Mario Gómez (2496) at Chess960 for the opening of a chess tournament.

Mario Gomez was 2 times champion of Spain and 8 times champion of the Basque Country.

The version that played against Mario is almost identical to the version 3.13 but with support for chess960. If you want to download it, this is the link, the version is not official, do not use for tournaments.

http://danasah.googlepages.com/DanaSahv.3.zip

DanaSah is winboard, it play chess960 in Arena or in ChessGUI of Mattias Gemuh

The evaluation on the normal variants is the same that he used to variant fischer and so the program made bad opening moves, for example the first game (position 840) a4 then make rook a3 to (this does not happen in the normal variant because the program plays openings book and castling. This is preventable by changing the table of values that I have where the rook is located on the board. Also I want to put more code for bishops trapped in a1, h1, a8 and h8 at the beginning of the game. Curiously this seems to juice worse against other programs, I have to do more tests.

First game: http://www.eiretaberna.com/danamario/danamario.htm
Second game: http://www.eiretaberna.com/mariodana/mariodana.htm

In the second game the viewer think that is not working well after castling for black. You can download the file pgn of the page and if you want to display in the GUI Fritz 10 or Shredder.

Pedro

[swami]

Congrats Pedro!!
btw the pgn for 2nd game is wrong, check the 24th move..

[pedrox]

Thanks Swami,

If you download the file pg from the page and you open it with Shredder (fritz10) then the game is OK.

It seems that it is a problem with Fritz 10 generating htm page and then open the file pgn, move 15 ... O-O the viewer does not want to do that.

This is the pgn:

Code: Select all

[Event "Blitz:20' (FRC 960)"]
[Site "?"]
[Date "2008.01.13"]
[Round "2"]
[White "Gomez Esteban, Juan Mario"]
[Black "DanaSah v.3.56 FRC"]
[Result "0-1"]
[Annotator "Fritz 960 (10s)"]
[Variant "chess 960"]
[SetUp "1"]
[FEN "rkqbbnrn/pppppppp/8/8/8/8/PPPPPPPP/RKQBBNRN w GAga - 0 1"]
[PlyCount "96"]
[EventDate "2008.??.??"]

{Chess960-Position809} 1. d4 Nhg6 2. e4 d6 3. Nhg3 e5 4. d5 {
White gets more space} Nf4 5. Ne3 a5 6. Qd2 Ra6 7. Bg4 Bd7 8. Bxd7 Nxd7 9. Ne2
Bg5 10. Nxf4 Bxf4 11. g3 Bg5 12. Qe2 Nf6 13. Nc4 a4 14. f3 {Covers g4} Qh3 15.
a3 O-O 16. Bb4 b5 17. Na5 Raa8 18. Nc6 Nd7 19. Ka2 Nb6 $2 (19... Nb8 $142 $5 {
is an interesting idea} 20. Nxb8 Raxb8 $14) 20. Qxb5 (20. f4 $5 Bh6 21. fxe5
Rfe8 22. Qxb5 dxe5 $18) 20... Qxh2 21. Qd3 Rfe8 22. Raf1 Qh3 23. Qe2 Nd7 $2 (
23... Bh6 24. Ba5 $18) 24. Rd1 (24. f4 $142 {secures victory} exf4 25. gxf4 $18
) 24... Nb8 $16 25. Na5 Na6 26. Be1 f5 27. Nc6 Nb8 28. Nb4 fxe4 29. Qxe4 Nd7
30. Bc3 Nc5 31. Qc4 (31. Qe2 $5 $16) 31... Be3 $14 32. Rge1 Bf2 33. Rh1 Qxg3
34. Nc6 {Here comes the goal-getter} (34. Be1 $142 $5 $11 {
and White can hope to survive}) 34... Qxf3 $17 35. Rdf1 $2 (35. Qf1 $142 h6 36.
Rh3 $17) 35... Qe4 $19 36. Qxe4 Nxe4 37. Ba5 Rxa5 $1 {Decoy: a5} 38. Nxa5 Ng3
39. Rxf2 Nxh1 40. Rg2 Rf8 41. b3 axb3+ 42. Kxb3 Nf2 43. a4 e4 (43... Rf3+ $142
$5 {makes it even easier for Black} 44. Kb4 h5 45. Rh2 $19) 44. Nc4 Nd1 45. a5
Rb8+ 46. Ka3 Nc3 47. Rg3 (47. Ne3 {is not the saving move} g6 $19) 47... Nxd5
48. Ka4 $4 (48. Rb3 {otherwise it's curtains at once} Ra8 49. Ka4 $19) 48...
Rb4+ 0-1

[hgm]

Your engine seems to have some protocl problems.

Under WinBoard it almost always loses because of forfeit by false illegal-move claim. This happens each time BigLion castles:

Code: Select all

[Event "Computer Chess Game"]
[Site "FOM-RHKA8J2A5WY"]
[Date "2008.01.18"]
[Round "-"]
[White "BigLion 2.23w"]
[Black "DanaSah v.3.56 FRC"]
[Result "1-0"]
[TimeControl "40/60"]
[Variant "fischerandom"]
[FEN "rkrnbnqb/pppppppp/8/8/8/8/PPPPPPPP/RKRNBNQB w CAca - 0 1"]
[SetUp "1"]

{--------------
r k r n b n q b
p p p p p p p p
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
P P P P P P P P
R K R N B N Q B
white to play
--------------}
1. e4 {+0.12/7 1} a5 {-0.24/8 0} 2. Nde3 {+0.20/5 1} Ra6 {+0.09/7 0} 3.
d3 {+0.27/5 2} Nc6 {+0.40/6 0} 4. Bc3 {-0.16/5 1} g5 {+0.61/6 0} 5.
Bxh8 {-0.44/4 1} Qxh8 {+0.51/5 0} 6. Ng3 {-0.51/4 1} e6 {+0.55/6 0} 7.
Ne2 {-0.49/5 1} Ng6 {+0.77/6 0} 8. g3 {-0.54/4 2} d6 {+0.88/6 0} 9.
Qg2 {-0.66/4 1} Bd7 {+1.18/5 0} 10. Nc4 {+0.14/4 1} Nce5 {+1.21/6 0} 11.
Nxe5 {+0.18/5 2} Nxe5 {+1.17/6 0}
{False illegal-move claim} 1-0

The move was O-O, which I think is legal in this position:
[d]1kr4q/1ppb1p1p/r2pp3/p3n1p1/4P3/3P2P1/PPP1NPQP/RKR4B w CAc - 0 12

In WinBoard protocol FRC castlings must be written as O-O and O-O-O (oh, not 0).

Another problem is the variants feature: You have no double quotes around the variants. WinBoard requires this, so it does not recognize the feature, and sends "rejected variants" (in stead of "accepted variants". This leads to an infinite loop, where Danasah sends "feature rejected" and WinBoard (correctly) responds with "rejected rejected". This then goes on the entire game, and fills Gigabytes of the hard disk!

So:
* accept "move O-O" and "move O-O-O"
* do not ever send "feature rejected". If you want debug output, always put something in front of it that WinBoard ignores, or is harmless (like "telluser feature rejected")
* put quotes around the variant names: feature variants="normal,fischerandom"

Code: Select all

5017 <second: DanaSah v.3.56 FRC, web: http://danasah.tk 
5017 <second: 13/01/2008, Pedro Castro, Spain
5017 <second: 
5017 <second: DanaSah only supports the winboard interface.
5017 <second: Type command quit to exit
5017 <second: 
5017 <second: Ini file danasah.ini no found.
5017 <second: Ownbook: off
5017 <second: Table hash size: 64 Mb.
5017 <second: Eval cache size: 8 Mb.
5017 <second: Bitbases path: ./egbb/
5017 <second: Bitbases men: 4
5017 <second: Bitbases cache size: 4
5017 <second: Kibitz: off.
5017 <second: danasah> feature done=0
5027 <second: feature setboard=1 draw=0 sigint=0 sigterm=0 myname="DanaSah v.3.56 FRC" variants=normal,fischerandom colors=0 ics=1
5027 >second: accepted setboard
5027 >second: accepted draw
5027 >second: accepted sigint
5027 >second: accepted sigterm
5027 >second: accepted myname
5027 >second: rejected variants
5027 >second: accepted colors
5027 >second: accepted ics
5027 <second: EgbbProbe not Loaded!
5027 <second: feature done=1
5027 >second: accepted done
5027 <second: feature rejected
5027 >second: rejected rejected
5027 <second: feature rejected
5027 >second: rejected rejected
5027 <second: feature rejected
5027 >second: rejected rejected
5027 <second: feature rejected
5027 >second: rejected rejected
5027 <second: feature rejected
5027 >second: rejected rejected
5027 <second: feature rejected
5027 >second: rejected rejected
5027 <second: feature rejected
5027 >second: rejected rejected
5027 <second: feature rejected
5027 >second: rejected rejected
5027 <second: feature rejected
...

[Graham Banks]

Sunday I had the opportunity to confront DanaSah against GM Mario Gómez (2496) at Chess960 for the opening of a chess tournament.

Mario Gomez was 2 times champion of Spain and 8 times champion of the Basque Country.

The version that played against Mario is almost identical to the version 3.13 but with support for chess960. If you want to download it, this is the link, the version is not official, do not use for tournaments.

http://danasah.googlepages.com/DanaSahv.3.zip

DanaSah is winboard, it play chess960 in Arena or in ChessGUI of Mattias Gemuh

The evaluation on the normal variants is the same that he used to variant fischer and so the program made bad opening moves, for example the first game (position 840) a4 then make rook a3 to (this does not happen in the normal variant because the program plays openings book and castling. This is preventable by changing the table of values that I have where the rook is located on the board. Also I want to put more code for bishops trapped in a1, h1, a8 and h8 at the beginning of the game. Curiously this seems to juice worse against other programs, I have to do more tests.

First game: http://www.eiretaberna.com/danamario/danamario.htm
Second game: http://www.eiretaberna.com/mariodana/mariodana.htm

In the second game the viewer think that is not working well after castling for black. You can download the file pgn of the page and if you want to display in the GUI Fritz 10 or Shredder.

Pedro

Congratulations Pedro.

[pedrox]

Hi H.G,

This version of DanaSah is compatible with Arena and ChessGUI (Matthias Gemut), Winboard so far not played FRC

For these two programs I receive the command setboard, but castling rights with the letters KQkq, Is it possible to receive it that way? All programs winboard playing in Arena Chess960 receive this.

DanaSah if accepts O-O and O-O-O. In your game rejected the castling probably because at the beginning of the game when receiving the command setboard had no proper rights for the castling.

In forum Winboard you will see the debug file for a previous version, in this case I do not have the problem of command rejected.

I also put quotes around the variant names: feature variants = "normal fischerandom" for Winboard.

Best,

Pedro

[hgm]

If you want 'setboard' to operate correctly not only for the opening position, but also for positions later in the game, you will have to make your FEN reader understand A-H castling rights. In FRC K or Q is not sufficient do describe all positions: later in the game you might have two Rooks on the same side of the King (say K-side), and you can castle with at most one of these Rooks, and there has to be a way to indicate the difference.

Therefore notation of castling in FRC FEN is done by giving the file of the piece that you may castle with.

Now I know there is kind of a war going on between different FEN standards (x-FEN and shredder-FEN, which have different preference for A-H or KQ in positions where the latter would be unambiguous). But as you have to understand both in in some positions anyway, there really is no need for taking sides in this battle and explicitly adding code that figures out if the message you already understand could also have been encoded in another way, and then act like it did not understand it.

I would recommend to always understand both notations. Mentioning the file (A-H notation) is always unambiguous (except on boards with 13 files or more ).

[pedrox]

All GUI when edit a position FRC have a problem, the problem is that they forget to ask whether there are 2 rooks on the same side with which one can castle, this problem is for Arena, Winboard, I believe Fritz and only one not forget to ask is Shredder GUI (You can download a trial version).

Shredder GUI is only compatible UCI (winboard with Wb2uci but only for the normal variant), so for the moment I think the engines xboard not have a proper GUI for edit positions at FRC therefore is not as important if we use for the moment x-FEN system or shredder-FEN.

There are already a few engines playing FRC in Arena, I think it would be interesting to try them if they play well on Winboard unchanged.

In DanaSah I had to make a small change. AICE winboard plays correctly.

I have a version that plays well in Arena, I now have another version that plays well in Winboard, I try to combine the 2 versions, but I have a problem with the start of the engine, Arena and Winboard seem to work differently, I consult this in the forum programming.

Best,

Pedro