+ Firebird wins Ware Gambit +

Discussion of computer chess matches and engine tournaments.

Moderators: hgm, Rebel, chrisw

rainhaus
Posts: 185
Joined: Sun Feb 01, 2009 7:26 pm
Location: Germany
Full name: Rainer Neuhäusler

+ Firebird wins Ware Gambit +

Post by rainhaus »

AOO Ware Gambit

1.a4 e5 2. a5 d5 3. e3 f5 4.a6

The old Ware Gambit is not played in tournament chess and you will search it in databases in vain (perhaps you can find one!?) It's no wonder.., with these moves!
Preston Ware Jr. (1821 – 1890, Massachusetts) was an enthusiastic U.S chess player who was specialized in unusual systems. Probably the greatest success in his life was his participation in the Second International Chess Tournament in Vienna 1882, the finest chess tournament of its time. There he won his game against non other than the first world champion, Wilhelm Steinitz.
The Ware Gambit is even more crazy than the Amar Gambit. To me it is the opening of a Pawn's Dream. Only powered by the desire fulfilling the wonderful transformation on the highest rank, the white pawn flies unstoppable up to a6.., but there he dies too early in the arms of a defending knight.. Yeah, if you has held a little bit of your childish fantasy, then the chess board becomes alive :)
In the following citation it is written in the language of chess:
The Ware opening attacks the b5 square and threatens to bring the a1 rook to the game. The b5 square is non-essential and if Black plays 1...e5, the f8 bishop prevents developing the white rook for the moment. The reply 1...e5 also gains space for Black in the center... The Ware Opening is something one could only expect to see from players completely new to chess.

Win%
It's unbelievable what is playable and winnable in chess, and the algorithms have to work right with these exotics, too. Ok, a white win was only done 2x by Firebird and 1x by Shredder, but a Ware gambit draw is good enough for White.

Code: Select all

White Wins   :      3 ( 7.1 %)
Black Wins   :     22 (52.4 %)
Draws        :     17 (40.5 %)
Games        :     42 (finished)
White Perf. : 27.4 %
Black Perf. : 72.6 %

Starting evaluation

is again surprisingly low, Firebird, the coolest of all.

Code: Select all

After 4.a6: 
                 cp      
Firebird        -0,18 
Stockfish       -0,44 
Deep Rybka      -0,46    
Zappa           -0,62   
Deep Shredder   -0,70  
Naum            -0,75
Deep Fritz      -0,80 
Opening Table (SCID)

Code: Select all

-------------------------------------------------------------------------------
  +3 =17 -22 (11.5/42: 27%)
-------------------------------------------------------------------------------
       1        2        3        4        5        6        7        8     
-------------------------------------------------------------------------------
 1  a4       a5       e3       a6       d4       La6[2]   Se2      OO[3]     7:
    e5       d5       f5       Sa6      e4[1]    ba6      Sf6      Ld6[4]   29%

 2  ...      ...      ...      ...      Sf3      Se5      c3[6]    Sc4       6:
    ...      ...      ...      ...      e4[5]    Sb4      Ld6      dc4[7]   17%

 3  ...      ...      ...      ...      ...      ...      d4       cd3[8]    4:
    ...      ...      ...      ...      ...      ...      ed3      Sf6[9]   13%

 4  ...      ...      ...      ...      ...      ...      ...      ...       6:
    ...      ...      ...      ...      ...      ...      ...      Ld6[10]  17%

 5  ...      ...      ...      ...      La6      d4       de5      Sf3[11]   5:
    ...      ...      ...      ...      ba6      Ld6      Le5      Lf6[12]  20%

 6  ...      ...      ...      ...      ...      ...      Se2      OO[14]    6:
    ...      ...      ...      ...      ...      e4       Sf6[13]  Ld6[15]  42%

 7  ...      ...      ...      ...      ...      ...      ...      ...       3:
    ...      ...      ...      ...      ...      ...      ...      Lb4[16]  33%

 8  ...      ...      ...      ...      ...      ...      ...      b3        5:
    ...      ...      ...      ...      ...      ...      ...      Lb4[17]  50%
Scores
Ranking Crosstable, Amar
(SCID)
GGTJan10_4221_10+12_A00Ware

Code: Select all

                                    Score     Fi Ry Na St DF DS Za
-------------------------------------------------------------------
 1: FireBird 1.0 betax64           8.5 / 12   XX == == == =1 11 11
 2: Deep Rybka 3 x 64              6.5 / 12   == XX 10 10 01 =1 01
 3: Naum 4 MPx64                   6.5 / 12   == 01 XX 01 == 10 1=
 4: Stockfish 1.6JAx64             5.5 / 12   == 01 10 XX 0= 0= =1
 5: Deep Fritz 11                  5.5 / 12   =0 10 == 1= XX == 0=
 6: Deep Shredder 12 x64           5.0 / 12   00 =0 01 1= == XX 10
 7: Zappa Mexico II x64            4.5 / 12   00 10 0= =0 1= 01 XX
42 games
--------------------------------------------------------

--------------------------------------------------------
Ranking Crosstable: 2 gambits, Amar+Ware
(FRITZ11-GUI)
GGTJan10_4221_10+12_A00Amar 2010

Code: Select all

                               1    2    3    4    5    6    7    
1   FireBird 1.0 betax64       **** ½1½½ 1111 10½½ 1½11 11½½ 11½1  18.5 /24
2   Deep Rybka 3x64            ½0½½ **** 01½1 1½10 ½½01 1010 1101  13.5 /24
3   Deep Shredder 12 x64       0000 10½0 **** 1101 1½10 111½ 10½½  12.5 /24
4   Naum 4 MP x64              01½½ 0½01 0010 **** 0½1½ 0101 1½½½  11.0 /24
5   Zappa Mexico II x64        0½00 ½½10 0½01 1½0½ **** 1½½0 101½  10.5 /24
6   Stockfish 1.6 JAx64        00½½ 0101 000½ 1010 0½½1 **** 1½0½   9.5 /24
7   Deep Fritz 11              00½0 0010 01½½ 0½½½ 010½ 0½1½ ****   8.5 /24
84 games
----------------------------------------------------------
Here is the pgn-link for the interested folk:
http://www.file-upload.net/download-222 ... e.pgn.html

Next: Sosolsky Gambit

Test conditions
Starting positions: pool of 100 gambits sorted by Elo number.
Time Control: short tournament level with 40/20', 20/10', 10'+12''.
System: Intel Core i7 920, oc 3700 MHz, 6 GB RAM. Hyperthreading off, Turbo Mode off. Vista 64Bit
Parameter: 3 Threads, ponder off, 1.2 MB Hash, EGTB 3-4-5 (not used by Firebird and Stockfish, but sometimes used by the GUI, even for both engines)
Fritz11-GUI: remis never, resign late.

Rainer Neuhäusler
beram
Posts: 1187
Joined: Wed Jan 06, 2010 3:11 pm

Re: + Firebird wins Ware Gambit +

Post by beram »

Again this good results confirms FB is stronger than Rybka also at longer timecontrols !
(Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz 3677 MHz Rybka W=20.6 plies; 318kN/ s; 56 TBAs B Firebird =25.7 plies; 6.899kN/s)

Total now after 2 of the 100 (!) tournaments:
GGTJan10_4221_10+12_A00 Amar+Ware 2010 84 games

1 2 3 4 5 6 7
1 FireBird 1.0 beta x64 mp **** ½1½½ 1111 10½½ 1½11 11½½ 11½1 18.5/24
2 Rybka 3 ½0½½ **** 01½1 1½10 ½½01 1010 1101 13.5/24
3 Deep Shredder 12 x64 FrGui 0000 10½0 **** 1101 1½10 111½ 10½½ 12.5/24
4 Naum 4 MPx64 01½½ 0½01 0010 **** 0½1½ 0101 1½½½ 11.0/24
5 Zappa Mexico II 0½00 ½½10 0½01 1½0½ **** 1½½0 101½ 10.5/24
6 Stockfish 1.6 JA 64bit_dec09 00½½ 0101 000½ 1010 0½½1 **** 1½0½ 9.5/24
7 Deep Fritz 11 00½0 0010 01½½ 0½½½ 010½ 0½1½ **** 8.5/24

Nice work Rainer

grts Bram
Dann Corbit
Posts: 12541
Joined: Wed Mar 08, 2006 8:57 pm
Location: Redmond, WA USA

Re: + Firebird wins Ware Gambit +

Post by Dann Corbit »

beram wrote:Again this good results confirms FB is stronger than Rybka also at longer timecontrols !
(Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz 3677 MHz Rybka W=20.6 plies; 318kN/ s; 56 TBAs B Firebird =25.7 plies; 6.899kN/s)

Total now after 2 of the 100 (!) tournaments:
GGTJan10_4221_10+12_A00 Amar+Ware 2010 84 games

1 2 3 4 5 6 7
1 FireBird 1.0 beta x64 mp **** ½1½½ 1111 10½½ 1½11 11½½ 11½1 18.5/24
2 Rybka 3 ½0½½ **** 01½1 1½10 ½½01 1010 1101 13.5/24
3 Deep Shredder 12 x64 FrGui 0000 10½0 **** 1101 1½10 111½ 10½½ 12.5/24
4 Naum 4 MPx64 01½½ 0½01 0010 **** 0½1½ 0101 1½½½ 11.0/24
5 Zappa Mexico II 0½00 ½½10 0½01 1½0½ **** 1½½0 101½ 10.5/24
6 Stockfish 1.6 JA 64bit_dec09 00½½ 0101 000½ 1010 0½½1 **** 1½0½ 9.5/24
7 Deep Fritz 11 00½0 0010 01½½ 0½½½ 010½ 0½1½ **** 8.5/24

Nice work Rainer

grts Bram
It takes about 400 games between approximate peers to have any idea if one engine is really stronger than the other.

Firebird might be stronger. But 24 games is only an indication.
beram
Posts: 1187
Joined: Wed Jan 06, 2010 3:11 pm

Re: + Firebird wins Ware Gambit +

Post by beram »

Dann Corbit wrote:
beram wrote:Again this good results confirms FB is stronger than Rybka also at longer timecontrols !
(Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz 3677 MHz Rybka W=20.6 plies; 318kN/ s; 56 TBAs B Firebird =25.7 plies; 6.899kN/s)

Total now after 2 of the 100 (!) tournaments:
GGTJan10_4221_10+12_A00 Amar+Ware 2010 84 games

1 2 3 4 5 6 7
1 FireBird 1.0 beta x64 mp **** ½1½½ 1111 10½½ 1½11 11½½ 11½1 18.5/24
2 Rybka 3 ½0½½ **** 01½1 1½10 ½½01 1010 1101 13.5/24
3 Deep Shredder 12 x64 FrGui 0000 10½0 **** 1101 1½10 111½ 10½½ 12.5/24
4 Naum 4 MPx64 01½½ 0½01 0010 **** 0½1½ 0101 1½½½ 11.0/24
5 Zappa Mexico II 0½00 ½½10 0½01 1½0½ **** 1½½0 101½ 10.5/24
6 Stockfish 1.6 JA 64bit_dec09 00½½ 0101 000½ 1010 0½½1 **** 1½0½ 9.5/24
7 Deep Fritz 11 00½0 0010 01½½ 0½½½ 010½ 0½1½ **** 8.5/24

Nice work Rainer

grts Bram
It takes about 400 games between approximate peers to have any idea if one engine is really stronger than the other.

Firebird might be stronger. But 24 games is only an indication.
For your pleisure -
It indicates once more...
rainhaus
Posts: 185
Joined: Sun Feb 01, 2009 7:26 pm
Location: Germany
Full name: Rainer Neuhäusler

Re: + Firebird wins Ware Gambit +

Post by rainhaus »

beram wrote:
Dann Corbit wrote:
beram wrote:Again this good results confirms FB is stronger than Rybka also at longer timecontrols !
(Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz 3677 MHz Rybka W=20.6 plies; 318kN/ s; 56 TBAs B Firebird =25.7 plies; 6.899kN/s)

Total now after 2 of the 100 (!) tournaments:
GGTJan10_4221_10+12_A00 Amar+Ware 2010 84 games

1 2 3 4 5 6 7
1 FireBird 1.0 beta x64 mp **** ½1½½ 1111 10½½ 1½11 11½½ 11½1 18.5/24
2 Rybka 3 ½0½½ **** 01½1 1½10 ½½01 1010 1101 13.5/24
3 Deep Shredder 12 x64 FrGui 0000 10½0 **** 1101 1½10 111½ 10½½ 12.5/24
4 Naum 4 MPx64 01½½ 0½01 0010 **** 0½1½ 0101 1½½½ 11.0/24
5 Zappa Mexico II 0½00 ½½10 0½01 1½0½ **** 1½½0 101½ 10.5/24
6 Stockfish 1.6 JA 64bit_dec09 00½½ 0101 000½ 1010 0½½1 **** 1½0½ 9.5/24
7 Deep Fritz 11 00½0 0010 01½½ 0½½½ 010½ 0½1½ **** 8.5/24

Nice work Rainer

grts Bram
It takes about 400 games between approximate peers to have any idea if one engine is really stronger than the other.

Firebird might be stronger. But 24 games is only an indication.
For your pleisure -
It indicates once more...
and what you'll say if Rybka should win the next gambit?
I guess you didn't have much clue about statistics, but a firm believe isn't to be sneezed and probably in the end you will say: I knew it right from the start :)
Carpe diem
Rainer
govert
Posts: 270
Joined: Thu Jan 15, 2009 12:52 pm

Re: + Firebird wins Ware Gambit +

Post by govert »

Rainer Marian wrote:
beram wrote:
Dann Corbit wrote:
beram wrote:Again this good results confirms FB is stronger than Rybka also at longer timecontrols !
(Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz 3677 MHz Rybka W=20.6 plies; 318kN/ s; 56 TBAs B Firebird =25.7 plies; 6.899kN/s)

Total now after 2 of the 100 (!) tournaments:
GGTJan10_4221_10+12_A00 Amar+Ware 2010 84 games

1 2 3 4 5 6 7
1 FireBird 1.0 beta x64 mp **** ½1½½ 1111 10½½ 1½11 11½½ 11½1 18.5/24
2 Rybka 3 ½0½½ **** 01½1 1½10 ½½01 1010 1101 13.5/24
3 Deep Shredder 12 x64 FrGui 0000 10½0 **** 1101 1½10 111½ 10½½ 12.5/24
4 Naum 4 MPx64 01½½ 0½01 0010 **** 0½1½ 0101 1½½½ 11.0/24
5 Zappa Mexico II 0½00 ½½10 0½01 1½0½ **** 1½½0 101½ 10.5/24
6 Stockfish 1.6 JA 64bit_dec09 00½½ 0101 000½ 1010 0½½1 **** 1½0½ 9.5/24
7 Deep Fritz 11 00½0 0010 01½½ 0½½½ 010½ 0½1½ **** 8.5/24

Nice work Rainer

grts Bram
It takes about 400 games between approximate peers to have any idea if one engine is really stronger than the other.

Firebird might be stronger. But 24 games is only an indication.
For your pleisure -
It indicates once more...
and what you'll say if Rybka should win the next gambit?
I guess you didn't have much clue about statistics, but a firm believe isn't to be sneezed and probably in the end you will say: I knew it right from the start :)
Carpe diem
Rainer
How about posting some data instead of putting each other down.
Bayeselo:

Code: Select all

ResultSet-EloRating>los
                              Fi Ry Na De St De Za
FireBird 1.0 beta x64 mp         73 75 86 86 89 93
Rybka 3                       26    52 67 67 73 79
Naum 4 MPx64                  24 47    66 65 72 79
Deep Fritz 11                 13 32 33    50 58 67
Stockfish 1.6 JA 64bit_dec09  13 32 34 49    57 66
Deep Shredder 12 x64 FrGui    10 26 27 41 42    59
Zappa Mexico II                6 20 20 32 33 40
ResultSet-EloRating>ratings
Rank Name                           Elo    +    - games score oppo. draws
   1 FireBird 1.0 beta x64 mp        87  140  140    12   71%   -15   58%
   2 Rybka 3                         27  154  154    12   54%    -4   25%
   3 Naum 4 MPx64                    21  146  146    12   54%    -4   42%
   4 Deep Fritz 11                  -18  139  139    12   46%     3   58%
   5 Stockfish 1.6 JA 64bit_dec09   -19  146  146    12   46%     3   42%
   6 Deep Shredder 12 x64 FrGui     -38  148  148    12   42%     6   33%
   7 Zappa Mexico II                -61  154  154    12   38%    10   25%
ResultSet-EloRating>



So, ELO diff is 60 Elo, and the error bars are at least +-140, and the likelihood that Firefox is better than Rybka is 73%
govert
Posts: 270
Joined: Thu Jan 15, 2009 12:52 pm

Re: + Firebird wins Ware Gambit +

Post by govert »

instead of putting each other down
I take that back. The tone is pretty civil compared to other places around here...
rainhaus
Posts: 185
Joined: Sun Feb 01, 2009 7:26 pm
Location: Germany
Full name: Rainer Neuhäusler

Re: + Firebird wins Ware Gambit +

Post by rainhaus »

govert wrote:
instead of putting each other down
I take that back. The tone is pretty civil compared to other places around here...
In view of your reflection, no problem. I think Bram shouldn't interpret my post like you, isn't it Bram?
You see yourself that it does not make much sense presenting Elo tables with barn door wide error margins. Therefore i decided to start it at a later time.
nice day
rn
rainhaus
Posts: 185
Joined: Sun Feb 01, 2009 7:26 pm
Location: Germany
Full name: Rainer Neuhäusler

Re: + Firebird wins Ware Gambit +

Post by rainhaus »

It takes about 400 games between approximate peers to have any idea if one engine is really stronger than the other.

Firebird might be stronger. But 24 games is only an indication.
Yes, in this long tournament you can check statistics practically and not abstract theoretically. Each additional gambit round will stabilize the ranking. But several engines will reach the final ranking much faster than the others. W'll see what's about your 400 games :) On the basis of what data this across-the- board estimation comes from?
rn
beram
Posts: 1187
Joined: Wed Jan 06, 2010 3:11 pm

Re: + Firebird wins Ware Gambit +

Post by beram »

Rainer Marian wrote:
govert wrote:
instead of putting each other down
I take that back. The tone is pretty civil compared to other places around here...
In view of your reflection, no problem. I think Bram shouldn't interpret my post like you, isn't it Bram?
You see yourself that it does not make much sense presenting Elo tables with barn door wide error margins. Therefore i decided to start it at a later time.
nice day
rn
No Problem for me.
Btw Rainer, what contempt setting for Rybka you use ?
According to mr Ansari this setting makes a significant difference, but in my testing I get no better results for Rybka, in Nunn2 matchgames, against Firebird, when I lower it to from 15 (standard) to 0