Result Mainz Chess960 qualifier: 1) Sjeng 2) Rybka 3) Ikarus

[Teemu Pudas]

there was a version of Toga more than a year ago by Teemu P. ->
Toga CMLX (and Toga 960) ...but as far as I know, it never worked properly.
(i.e. would often crash after castling)...

IMHO, there is no known version of Toga that can play chess960...?!
(i.e. one that can accept and understand all the different castling moves....)

CMLX had a FEN bug that made it crash if the kings were on different files and one still had castling rights (it'll work from any starting position). And it doesn't work under Arena (the code is there, but I accidentally made it accept (and output) 0-0 instead of O-O).

[Alexander Schmidt]

But SMK's UCI protocol says:

"This is why castling moves are sent in the form king "takes" his own rook."

To work properly under all UCI interfaces (including Polyglot) Toga CMLX should do it like in the UCI standart.

[Teemu Pudas]

Note I said Arena. I do that, but only if I get a Shredder-FEN. Arena uses X-FEN and O-O castling.

[diep]

congratulations to Sjeng for its impressive victory.

Rybka 3.0 and Toga played at 8 core machines (real 8 cores, not 8 fake cores from i7) so similar hardware like in world champs 2009.

Sjeng clearly beated them all.
Congrats on Ikarus for its clear 2nd spot.

p.s. don't jammer about bug XYZ in the first place toga's and rybka's shouldn't have joined a final as that can seriously influence the outcome, in this game that all wasn't needed as Sjeng hammered them all.

[diep]

Code: Select all

>Automato 
> icctd version: 8.676 (Compiled Jun 27 2009 08:31:33)
> #  999   Mainz Chess960 Computer Section "help Chess960" Double Round Robin w22 5 5 rated manager 
> ******   8 players/14 round (of 14)
>          Managed by .   (Sat Jul  4 09 12:35 - Sat Jul  4 09 17:30)
>          System: main TD: Automato
>          max_bye_points:3
>          max_best:3000 last_sort:score scheduled: no
>     Name                   1  2  3  4  5  6  7  8 Score  #g
>   1 Sjeng          (2032)  xx 1= 11 10 11 11       12.5  10
>   2 Rybka          (2000)  0= xx 11 =1 =1 11 i     11.5  10
>   3 Ikarus         (1852)  00 00 xx == 11 11        9.0  10
>   4 Toga           (2000)  01 =0 == xx 00 11    i   8.5  10
>   5 Jonny          (2099)  00 =0 00 11 xx 11        8.5  10
>   6 The Baron      (2097)  00 00 00 00 00 xx        4.0  10
>   7 Glaurung       (2223)     o              xx     0.0   0
>   8 Fritz          (1400)           o           xx  0.0   0

Sjeng and Ikarus will join Shredder and Rybka in Mainz.

Congratulations! Why was Rybka playing?

It shouldn't have played in the first place. Tornado-C was not allowed to join for example but Rybka was. Very bad decision taking.

Lucky Sjeng beated it clearly.

This makes Sjeng a serious candidate for title over there, the Rybka's did not play at slow hardware here. Sjeng always was a good blitz engine and proves itself here once again.

Sjeng and Rybka do not have much of a chessknowledge. It seems that for an engine like Sjeng all it needs is equal hardware to beat Rybka. Maybe Sjeng has a far better search, especially for blitz?

Vincent

[Titu]

From the ICC notes it looks like that Rybka played on a dual core

intel e2140@1.6GHz, 1GB Ram

Obviously the notes can in incorrect, but I have seen kirkconley play on ICC vs quadcore Rybkas where it drew or lost.

Also the handle "kirkconley" does not seem to have any connection to Vas or his team, as GCP already pointed out.

[Harvey Williamson]

From the ICC notes it looks like that Rybka played on a dual core

intel e2140@1.6GHz, 1GB Ram

Obviously the notes can in incorrect, but I have seen kirkconley play on ICC vs quadcore Rybkas where it drew or lost.

Also the handle "kirkconley" does not seem to have any connection to Vas or his team, as GCP already pointed out.

Rybka and Shredder are invited to Mainz anyway - Rybka should not have played in this prelim.

[Alexander Schmidt]

Note I said Arena. I do that, but only if I get a Shredder-FEN. Arena uses X-FEN and O-O castling.

I am talking about the official UCI protocol what is used in Polyglot. Thats what Toga CMLX is supposed to use (option name UCI_Chess960 type check default false). Here Toga CMLX sends 0-0 instead of "King takes rook"

Code: Select all

POLYGLOT FEN rqkr4/1pp2ppp/p2bpnn1/1b1p4/3P4/3BPNN1/PPPB1PPP/RQ3RK1 b da - 7 8
Adapter->Engine: position fen rqkrbbnn/pppppppp/8/8/8/8/PPPPPPPP/RQKRBBNN w DAda - 0 1 moves d2d4 d7d5 g1f3 g8f6 e2e3 e7e6 f1d3 a7a6 e1a5 f8d6 h1g3 e8b5 c1d1 h8g6 a5d2
Adapter->Engine: go wtime 35180 btime 46930
.
.
.
Engine->Adapter: bestmove 0-0 ponder f1c1

Arena has an inofficial implementation of Chess960 in UCI, it was released before the official Chess960 extensions. Arena does not support the official UCI/Chess960 protocol.

[Titu]

From the ICC notes it looks like that Rybka played on a dual core

intel e2140@1.6GHz, 1GB Ram

Obviously the notes can in incorrect, but I have seen kirkconley play on ICC vs quadcore Rybkas where it drew or lost.

Also the handle "kirkconley" does not seem to have any connection to Vas or his team, as GCP already pointed out.

Rybka and Shredder are invited to Mainz anyway - Rybka should not have played in this prelim.

In the human event, Nakamura was already qualified for this years Mainz, but ICC allowed him to enter. I guess ICC have the same rules for comps.

Nakamura was knocked out in the quaters I believe. When I asked Nakamura why he is playing despite beeing qualified he said he could ge some free money in case he was in top 3.

[Teemu Pudas]

I am talking about the official UCI protocol what is used in Polyglot.

Of course I know all that. Turns out CMLX never leaves Arena mode if it ever enters it. And since the board is initialized with a KQkq fen at startup...

Fixed now. New version here.