Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Wrong Rybka 3 evaluation - what about cluster?
- - By Xaggard (**) Date 2009-10-08 14:14
r2q1rk1/ppp2ppp/1bnp1nb1/8/2PP1PP1/P1N4P/1P2N1B1/R1BQK2R b KQ -


In this position Rybka shows even adavantage for Black, but this is totally wrong!
Shredder 12 does well here, his eval is about +1 for white. I made a match between Rybka and Shredder here. Shred was right he won the game :)
I wonder what's Rybka Cluster evaluation here?
Parent - - By Lukas Cimiotti (Bronze) [de] Date 2009-10-08 15:08
r2q1rk1/ppp2ppp/1bnp1nb1/8/2PP1PP1/P1N4P/1P2N1B1/R1BQK2R b KQ - 0 1


Analysis by Rybka 3:

1...h6 2.f5 Bh7
  =  (-0.01)   Depth: 6   00:00:00  7kN
1...h6 2.f5 Bh7 3.0-0 Re8 4.b4
  =  (-0.09)   Depth: 7   00:00:00  21kN
1...h6 2.f5 Bh7 3.0-0 Re8 4.b4
  =  (-0.09)   Depth: 8   00:00:00  32kN
1...h6 2.f5 Bh7 3.0-0 Re8 4.b4 a6
  =  (-0.02)   Depth: 9   00:00:00  49kN
1...h6 2.f5 Bh7 3.0-0 Re8 4.Bf4 Qd7 5.b3
  =  (-0.13)   Depth: 10   00:00:00  106kN
1...h6 2.f5 Bh7 3.0-0 Re8 4.Bf4 Qd7 5.Rf2 Na5 6.Qd3
  =  (-0.10)   Depth: 11   00:00:00  162kN
1...h6 2.f5 Bh7 3.0-0 Re8 4.b4 a6 5.Bf4 Qd7 6.Qd2
  =  (-0.11)   Depth: 12   00:00:00  250kN
1...h6 2.f5 Bh7 3.Qd3 Re8 4.Kd1 Qd7 5.Kc2 d5 6.c5 Ba5
  =  (0.08)   Depth: 13   00:00:04  998kN
1...h6 2.f5 Bh7[] 3.Qd3 Re8 4.Kd1 Qd7 5.Bd2 h5 6.Bf3 hxg4 7.hxg4 Bxd4 8.Nxd4 Ne5 9.Qf1 c5
  =  (0.05)   Depth: 14   00:00:07  1767kN
1...h6 2.f5 Bh7[] 3.Qd3 Re8 4.Kd1 Qd7 5.Bd2 h5 6.Bf3 hxg4 7.hxg4 Bxd4 8.Nxd4 Ne5 9.Qf1 c5
  =  (0.05)   Depth: 15   00:00:09  2313kN
1...h6 2.f5 Bh7[] 3.Qd3 Re8 4.Kd1 Qd7 5.Bd2 h5 6.Bf3 hxg4 7.hxg4 Bxd4 8.Nxd4 Ne5 9.Qf1 c5
  =  (0.05)   Depth: 16   00:00:18  4540kN
1...h6 2.f5 Bh7[] 3.b4 a6 4.0-0 Qd7 5.Be3 Rae8 6.Qd2 Qe7 7.Bf4 Rd8 8.Rf3
  +/=  (0.30)   Depth: 17   00:01:55  28595kN

I don't know what you did, but my Rybka 3 shows an advantage for white.
Parent - By chessprofessional (**) [rs] Date 2009-10-08 15:24
ha ha ha 1:55 rybka 3 only
Parent - By Highendman (****) Date 2009-10-08 16:31
:)
Parent - - By Kappatoo (*****) [de] Date 2009-10-09 08:42
Still - the score isn't that great. I think the position illustrates that Rybka tends to underestimate space.
Parent - - By Felix Kling (Gold) [de] Date 2009-10-10 11:10
yes, indeed, space and mobility of pieces...
Parent - By Banned for Life (Gold) Date 2009-10-10 17:25
Most engines evaluate space too highly. In many openings, e.g. B38, white gets more space and an easier game, but it is not clear that this would translate into a good winning percentage in freestyle or cc.

I also don't usually have a problem with Rybka's mobility scoring, except when it locks pieces up forever as it did with a bishop in one of the games against Zappa in Mexico.
Parent - - By Xaggard (**) Date 2009-10-09 11:34
Let' look at Shredders info:

r2q1rk1/ppp2ppp/1bnp1nb1/8/2PP1PP1/P1N4P/1P2N1B1/R1BQK2R b KQ - 0 1


Analysis by Deep Shredder 12 x64:

1...Qe7 2.Bxc6 bxc6 3.f5 Bxf5 4.gxf5 d5 5.cxd5 cxd5
  (1.18)   6/16   00:00:00  34тП
1...Qe8 2.0-0 h6 3.f5 Bh7 4.Kh2 Qe7
  (1.00)   6/20   00:00:00  86тП
1...h6 2.f5 Bh7 3.0-0 Re8 4.Kh2 Na5
  (0.80)   6/20   00:00:00  100тП
1...h6 2.f5 Bh7 3.0-0 Re8 4.Bf4 a6 5.Kh2
  (0.82)   7/20   00:00:00  112тП
1...Qe8 2.f5 Bxd4 3.fxg6 fxg6 4.Nb5 Bb6 5.Rf1
  (0.57)   7/20   00:00:00  127тП
1...Qe8 2.f5 Nxd4 3.fxg6 fxg6 4.Rf1 c6 5.Ne4 Nf3+ 6.Rxf3 Qxe4 7.Qxd6 Qxc4
  (0.70)   8/21   00:00:00  187тП
1...Qe8 2.f5 Nxd4 3.fxg6 fxg6 4.Rf1 c6 5.Na4 Bc5 6.Rf4 Nxe2 7.Qxe2
  (0.71)   9/22   00:00:00  303тП
1...Qe8 2.f5 Nxd4 3.fxg6 fxg6 4.Rf1 c6 5.Na4 Bc5 6.Bf4 Qe6 7.Bxd6 Nf3+ 8.Rxf3 Bxd6
  (0.94)   10/24   00:00:00  515тП
1...h6 2.f5 Bh7 3.0-0 Re8 4.Bf4 a6 5.Kh2 Qe7 6.b4 a5 7.b5
  (0.89)   10/26   00:00:00  686тП
1...h6 2.f5 Bh7 3.b4 a6 4.c5 Ba7 5.0-0 Re8 6.Qa4 Qd7 7.d5 Ne5 8.Qxd7 Nfxd7
  (1.15)   11/34   00:00:00  1701тП
1...Qe8 2.f5 Nxd4 3.fxg6 fxg6 4.Rf1 c6 5.Na4 Bc5 6.Bf4 Qe6 7.Bxd6 Nf3+ 8.Rxf3 Bxd6
  (0.94)   11/34   00:00:01  1939тП
1...Qe8 2.f5 Nxd4 3.fxg6 fxg6 4.Rf1 Qe7 5.Bxb7 Rab8 6.Bg2 Nb3 7.Nd5 Ba5+ 8.Bd2 Bxd2+ 9.Qxd2 Nxd5 10.Qxd5+ Kh8 11.Rxf8+ Rxf8
  (1.12)   12/32   00:00:01  3152тП
1...Qe8 2.f5 Nxd4 3.fxg6 fxg6 4.Kd2 c6 5.Nxd4 Bxd4 6.Qe2 Bxc3+ 7.bxc3 Qf7 8.Kc2 Rae8 9.Qd3 d5 10.cxd5 Nxd5
  (1.17)   13/32   00:00:02  5290тП
1...Qe8 2.f5 Nxd4 3.fxg6 fxg6 4.Kd2 c6 5.Nxd4 Bxd4 6.Qe2 Bxc3+ 7.bxc3 Qd7 8.Kc2 Rae8 9.Be3 b6 10.Rhf1 d5 11.cxd5 cxd5
  (1.16)   14/30   00:00:05  10567тП
1...Qe8 2.0-0 Be4 3.Nxe4 Nxe4 4.Qd3 f5 5.gxf5 Nf6 6.b4 a6 7.c5 Ba7 8.Kh2 Qd7 9.Rd1 d5 10.Ng3 Rae8 11.Rg1 Kh8
  (1.24)   15/37   00:00:10  19054тП
1...Qe8 2.0-0 Be4 3.Nxe4 Nxe4 4.Qd3 f5 5.gxf5 Nf6 6.b4 a6 7.Bb2 Qd7 8.c5 Ba7 9.Ng3 Kh8 10.Rad1 Rad8 11.Qc2 Qf7 12.Rde1 dxc5 13.dxc5 Bb8 14.Re6
  (1.36)   16/44   00:00:21  40270тП

Rybka s*cks here :)
Parent - - By Gaмßito (****) [cr] Date 2009-10-09 22:33 Edited 2009-10-09 22:39
I think white has a comfortable position. Apparently there is something a bit strange with some engines here.

This is what the new Fritz 12 engine thinks; but evaluation doesn't looks very accurately. Look:

r2q1rk1/ppp2ppp/1bnp1nb1/8/2PP1PP1/P1N4P/1P2N1B1/R1BQK2R b KQ - 0 1


Analysis by Fritz 12:

1...h6 2.f5 Bh7 3.0-0 Re8 4.Bf4 Qd7 5.Nd5 Ne4
  =/+  (-0.63)   Depth: 9/25   00:00:00  142kN
1...h6 2.f5 Bh7 3.0-0 Re8 4.Bf4 Qd7 5.Nd5 Nxd5 6.Bxd5 a6
  =/+  (-0.38)   Depth: 10/27   00:00:00  310kN
1...h6 2.f5 Bh7 3.0-0 Re8 4.Bf4 Rb8 5.Rf2 Qd7 6.Qd3 a6
  =/+  (-0.40)   Depth: 11/27   00:00:00  653kN
1...h6 2.f5 Bh7 3.0-0 Re8 4.Bf4 h5 5.g5 Nd7 6.Qd2 Bxf5 7.Bxd6
  =/+  (-0.39)   Depth: 12/31   00:00:01  1313kN
1...h6 2.f5 Bh7 3.0-0 Re8 4.Bf4 Qd7 5.Nd5 Nxd5 6.Bxd5 Qe7 7.Re1 Qf6 8.Kg2
  =/+  (-0.63)   Depth: 13/35   00:00:03  3153kN
1...h6 2.f5 Bh7 3.b4 a6 4.c5 Ba7 5.0-0 Re8 6.Bf4 Qd7 7.Re1 Rad8 8.cxd6 cxd6
  =  (-0.18)   Depth: 14/36   00:00:09  7502kN
1...h6 2.f5 Bh7 3.b4 a6 4.c5 Ba7 5.0-0 Re8 6.Bf4 Qd7 7.Re1 Rad8 8.Rb1 d5
  =  (-0.15)   Depth: 15/37   00:00:20  17146kN
1...h6 2.f5 Bh7 3.b4 a6 4.c5 Ba7 5.0-0 Re8 6.Bf4 Qd7 7.Re1 Rad8 8.Rb1 dxc5
  =  (-0.13)   Depth: 16/41   00:00:45  39336kN
1...h6 2.f5 Bh7 3.b4 a6 4.c5 Ba7 5.0-0 Re8 6.Bf4 Qd7 7.Re1 Rad8 8.Bf3 dxc5
  =  (-0.16)   Depth: 17/41   00:01:32  82480kN
1...h6 2.f5 Bh7 3.b4 a6 4.c5 Ba7 5.0-0 Re8 6.Bf4 Qd7 7.Re1
  =  (-0.03)   Depth: 18/42   00:03:08  172mN
1...h6 2.f5 Bh7 3.b4 a6 4.c5 Ba7 5.0-0 Re8 6.Bf4 Qd7 7.Re1
  =  (-0.10)   Depth: 19/45   00:07:03  393mN
1...h6 2.f5 Bh7 3.b4 a6 4.c5 Ba7 5.0-0 Re8 6.Bf4 Qd7 7.Re1
  =  (0.08)   Depth: 20/45   00:14:25  815mN
1...h6 2.f5 Bh7 3.b4 a6 4.c5 Ba7 5.0-0 Re8 6.Bf4 Qd7 7.Re1
  =  (0.08)   Depth: 21/49   00:34:21  1929mN

(AMD, 09.10.2009)

Regards,
Gaмßito.
Parent - By Xaggard (**) Date 2009-10-10 06:25
[quote]I think white has a comfortable position.[/quote]

Yes and even more! This is a totally lost position for Black.
Try for white Deep Fritz 10 or Deep Shredder 12 and Rybka 3 for Black. Results will be always the same - Rybka will lose!
Parent - By Eelco de Groot (***) Date 2009-10-10 11:01
I got a very up- and down evaluation here, I do not trust the evaluation of Ancalagon very much. I would not know if Black is really lost, he just seems to have less space to maneuver, development is okay though. Can Black break open the position then White's King is still standing a little exposed there, not having castled. What do the better players say? I have not much idea how to evaluate this or why Ancalagon does not play h6, apparently Black invests some material in breaking open the position :)!?

r2q1rk1/ppp2ppp/1bnp1nb1/8/2PP1PP1/P1N4P/1P2N1B1/R1BQK2R b KQ -


Engine: Ancalagon 1.3 Weak Squares 180 Board Control middlegame 50 endgame 50
Build 339 (Athlon 2009 MHz, 256 MB) by Romstad, Costalba, Kiiski, de Groot

2.00   0:09   -0.61    1...Qe7 2.Bxc6 bxc6 3.f5 c5 4.fxg6 cxd4
                       5.gxh7+ Nxh7 6.Nd5 Qe4 7.Kf2 f5
                       8.Nxb6 fxg4+ 9.Kg1 axb6 10.Nxd4 (4.087.478) 428

2.02   0:09   -0.61    1...Qe8 (4.096.162) 428

2.03   0:09   -0.35    1...Re8 2.f5 Bxd4 3.Bxc6 Bxc3+
                       4.bxc3 bxc6 5.fxg6 hxg6 6.O-O Ne4 (4.185.227) 428

3.01   0:13   -1.84    1...Re8 2.Bxc6 bxc6 3.f5 Bxf5 4.gxf5 c5
                       5.d5 Qd7 6.O-O c6 (5.691.760) 433

3.02   0:21   -1.25    1...Qe8 2.O-O Be4 3.Nxe4 Nxe4 4.Qd3 Nf6
                       5.Bd2 Qe7 (9.171.381) 434

3.05   0:21   -0.78    1...d5 2.f5 Bxf5 3.gxf5 dxc4 (9.441.012) 435

4.01   0:27   -1.41    1...d5 2.f5 Bxf5 3.gxf5 dxc4 4.Bxc6 bxc6
                       5.Bg5 Re8 6.O-O c5 7.d5 Rb8 (11.815.652) 435

4.02   0:27   -1.25    1...Qe8 2.O-O Be4 3.Nxe4 Nxe4 4.Qd3 Nf6
                       5.Bd2 Qe7 (11.844.896) 435

4.04   0:27   -1.14    1...h5 2.f5 Bh7 3.g5 Nd7 4.O-O Re8
                       5.b4 Nde5 (12.005.443) 436

4.05   0:30   -0.53    1...Nxd4 2.Nxd4 Re8+ 3.Nce2 Bxd4
                       4.Qxd4 Be4 5.Bxe4 Nxe4 6.O-O Qh4
                       7.Qd3 h5 8.g5 (13.295.784) 439

5.01   3:59   -0.75    1...Nxd4 2.Nxd4 Re8+ 3.Nce2 Be4
                       4.O-O d5 5.Bxe4 Rxe4 6.c5 Bxc5 7.b4 Bb6
                       8.Bb2 c6 9.Ng3 Re8 (106.147.576) 443

5.20   4:10   -0.42    1...Nd7 2.O-O f5 3.b4 Qh4 4.c5 dxc5
                       5.bxc5 Nxc5 6.Bxc6 bxc6 7.dxc5 Bxc5+
                       8.Nd4 Qxh3 (111.026.346) 442

6.01   4:36   -0.14    1...Nd7 2.O-O f5 3.Be3 Nf6 4.gxf5 Bxf5
                       5.Ng3 Qd7 6.c5 Bxh3 7.cxb6 Bxg2
                       8.Kxg2 Ng4 9.Qb3+ Rf7 (121.586.496) 439

7.01   12:38  -0.87    1...Nd7 2.f5 Qh4+ 3.Kf1 Rae8 4.Ne4 (332.533.369) 438

7.14   13:15  -0.71    1...Nxd4 2.Nxd4 Re8+ 3.Nce2 Be4
                       4.O-O d5 5.Bxe4 Rxe4 6.c5 Bxc5 7.b4 Bb6
                       8.Bb2 c6 (348.447.851) 437

8.01   40:07  -0.37    1...Nxd4 2.Nxd4 Re8+ 3.Nce2 Be4
                       4.O-O d5 5.c5 Bxc5 6.b4 Bf8 (1.040.868.779) 432

9.01   89:16  -0.84    1...Nxd4 2.Nxd4 Re8+ 3.Kf1 Ne4
                       4.Bxe4 Bxe4 5.Rh2 Qf6 6.Be3 Rad8
                       7.Rc1 c6 8.Qd2 h6 9.b4 Bxd4 10.Qxd4 Qxd4
                       11.Bxd4 Bd3+ 12.Kg2 Bxc4 13.Nd1 Bd5+
                       14.Kg3 Re1 (2.365.809.920) 441

9.02   94:55  -0.64    1...Nd7 2.Be3 f5 3.Bf2 Na5 4.b4 Nxc4
                       5.Bd5+ Bf7 6.Qb3 Bxd5 7.Nxd5 c6
                       8.Qxc4 cxd5 9.Qxd5+ Rf7 10.Qxd6 Bc7 (2.502.108.835) 439

10.01  188:25 -1.05    1...Nd7 2.f5 Qh4+ 3.Kf1 Nxd4 4.Nxd4 Bxf5
                       5.gxf5 Bxd4 6.Ne4 Bf6 (4.821.542.442) 426

best move: Nf6-d7 time: 198:41.657 min  n/s: 426.489  nodes: 5.084.430.243
Parent - - By George Tsavdaris (****) Date 2009-10-11 10:20

>Let' look at Shredders info:
>Analysis by Deep Shredder 12 x64:
>1...Qe8 2.0-0 Be4 3.Nxe4 Nxe4 4.Qd3 f5 5.gxf5 Nf6 6.b4 a6 7.Bb2 Qd7 8.c5 Ba7 9.Ng3 Kh8 10.Rad1 Rad8 11.Qc2 Qf7 12.Rde1 dxc5 >13.dxc5 Bb8 14.Re6
(1.36)   16/44   00:00:21  40270тП


>Rybka s*cks here :-)


•Rybka 3 having a 0.30 evaluation most probably does not mean the same of what a 0.30 evaluation means for Shredder 12.
•Rybka's 3 0.30 evaluation means it believes it will score from this position as white, about 62.7%.
•Shredder's 12 1.36 evaluation for example may mean it believes it will score from this position as white, about 65%. So that would mean R3 and S12 believe about the same things for this position (the whole point is that 1.36 of Shredder may mean 0.35 e.g for Rybka and that you should not compare evaluations of different engines).
•Also Shredder's 12 1.36 may reduce in the next ply so things may be even more equal for Rybka's and Shredder's beliefs in this position.
Parent - - By Xaggard (**) Date 2009-10-11 10:37
Rather than philosophize just run engine match Shredder 12 - Rybka 3 from this position. ;)
Parent - - By George Tsavdaris (****) Date 2009-10-11 10:46
I can't since i don't have Shredder 12, but since you have it then you can run 2-3 matches with (2-3) different time controls and give as the results as also the games.

The best would be 6 matches for example with each engine play once with white and once with black and at time controls like: 4+2" , 40/15, 40/40.
Parent - - By Xaggard (**) Date 2009-10-11 11:15 Edited 2009-10-11 11:22
Ok! I'll show you the game (Time Control 15'+10")

[Event "Game 15м+10с"]
[Site "PC"]
[Date "2009.10.11"]
[Round "1"]
[White "Deep Shredder 12 x64"]
[Black "Rybka 3"]
[Result "1-0"]
[Annotator "Someone"]
[SetUp "1"]
[FEN "r2q1rk1/ppp2ppp/1bnp1nb1/8/2PP1PP1/P1N4P/1P2N1B1/R1BQK2R b KQ - 0 1"]
[PlyCount "28"]
[TimeControl "900+10"]

1... h6 {0.03/15 40} 2. f5 {1.41/17 40} Bh7
{0.03/13 0} 3. b4 {1.41/17 55 (Qd3)} a6 {0.34/15 157 (Qe7)} 4. O-O {1.33/18 52}
Qd7 {0.34/15 41 (Ba7)} 5. c5 {1.41/18 48 (Be3)} Ba7 {0.23/15 27} 6. Qa4 {1.36/
16 41 (Nf4)} Rad8 {0.07/14 41} 7. Be3 {1.48/16 49 (Rd1)} h5 {0.00/14 39 (Bb8)}
8. Bxc6 {1.87/16 39 (g5)} Qxc6 {0.77/15 93} 9. Qxc6 {2.21/18 18} bxc6 {0.77/14
0} 10. Bg5 {1.96/19 7} Rfe8 {0.80/15 37 (hxg4)} 11. Bxf6 {2.29/20 59 (gxh5)}
gxf6 {1.15/15 31} 12. Ng3 {2.09/19 64} Kh8 {1.47/15 62} 13. Nxh5 {2.47/18 9}
Re3 {1.47/16 12} 14. Rac1 {2.47/18 38} dxc5 {1.47/16 138 (a5)} 15. dxc5 {2.55/
17
7 adjud.} 1-0

I aborted the game here since there was no sence to play more! Have a look at these funny Bishops :)



Shredder understands this type of position much better than Rybka.
I also once played corr. game, where I believed Rybka and locked my bishop right as on the picture. As s result I must resign :( after about 20 moves.
Parent - By exigentsky (***) [us] Date 2009-10-11 12:16
Yes, I've also noticed this as a weakness for Rybka when analyzing the Maroczy bind. It seemed like Rybka didn't care about space even when it mattered and Black was a bit worse.
Parent - By Vinvin (***) [be] Date 2009-10-11 13:28
Great post !
Thanks
Parent - - By George Tsavdaris (****) Date 2009-10-11 14:18

>Shredder understands this type of position much better than Rybka.


Perhaps, but perhaps not.
The fact that Deep Shredder 12 beat Rybka 3 mt with white in this position does not mean in this position white is better.
So perhaps it's a draw or black is better but just Rybka 3 mt didn't play correctly, so Shredder's evaluation of this position is wrong, so Shredder does not understand this position better than Rybka.
Parent - - By Xaggard (**) Date 2009-10-11 14:29

> The fact that Deep Shredder 12 beat Rybka 3 mt with white in this position does not mean in this position white is better.
> So perhaps it's a draw or black is better but just Rybka 3 mt didn't play correctly, so Shredder's evaluation of this position is wrong, so Shredder does not understand this position better than Rybka.


entia non sunt multiplicanda praeter necessitatem :)
Parent - - By George Tsavdaris (****) Date 2009-10-11 14:34

>> The fact that Deep Shredder 12 beat Rybka 3 mt with white in this position does not mean in this position white is better.
>> So perhaps it's a draw or black is better but just Rybka 3 mt didn't play correctly, so Shredder's evaluation of this position is wrong, so Shredder does not understand this position better than Rybka.


>entia non sunt multiplicanda praeter necessitatem :-)


Occam's razor does not apply here since what i say is another possibility of what actually happens and not a more complex way of describing your theory of "Shredder understands this position better than Rybka".
Parent - - By Xaggard (**) Date 2009-10-11 15:23

> Occam's razor does not apply here since what i say is another possibility of what actually happens and not a more complex way of describing your theory of "Shredder understands this position better than Rybka".


OK! Another example. SmarThink 1.10 (1 CPU, ELO 2818) vs Rybka 3 (4 CPU, ELO 3235). ELO difference is 417!!! And result is only a draw for Rybka!

[Event "15м+10с"]
[Site "PC"]
[Date "2009.10.11"]
[Round "1"]
[White "SmarThink 1.10"]
[Black "Rybka 3"]
[Result "1/2-1/2"]
[Annotator "1.52;-0.06"]
[SetUp "1"]
[FEN "r2q1rk1/ppp2ppp/1bnp1nb1/8/2PP1PP1/P1N4P/1P2N1B1/R1BQK2R b KQ - 0 1"]
[PlyCount "75"]
[TimeControl "900+10"]

1... h6 {-0.06/15 34} 2. f5 {1.52/15 69} Bh7
{-0.06/13 0} 3. b4 {1.56/15 66 (0-0)} a6 {0.25/15 81} 4. O-O {1.69/15 52} Rb8 {
0.27/15 35 (Re8)} 5. c5 {2.27/16 50 (g5)} Ba7 {0.37/15 21} 6. Bxc6 {2.25/15 48
(g5)} bxc6 {0.00/17 20} 7. Qa4 {2.10/15 47} dxc5 {0.00/17 24} 8. bxc5 {2.01/16
45} Nd5 {0.03/17 126} 9. Nxd5 {2.02/15 43 (Rf3)} Qxd5 {0.00/15 17} 10. Bf4 {2.
03/14 42} Rb3 {0.07/15 50 (Qb3)} 11. Kh2 {2.02/13 40} Bb8 {0.10/15 33 (g5)} 12.
Qxa6 {2.28/14 39 (Rae1)} g5 {0.28/16 91 (Kh8)} 13. Bd2 {2.42/15 37} f6 {0.49/
15 98 (Kh8)} 14. Rab1 {2.43/13 36} Rxb1 {0.65/16 16} 15. Rxb1 {2.41/14 12} Qf3
{0.95/17 56 (Kg7)} 16. Rg1 {2.41/15 35} h5 {1.02/17 36} 17. Rg3 {2.23/15 34}
Qf1 {1.01/17 16} 18. Qd3 {2.23/15 32} hxg4 {1.01/18 13} 19. hxg4 {2.24/16 31}
Kg7 {1.01/18 13} 20. Bxg5 {2.67/15 30 (Nc1)} fxg5 {0.74/14 15} 21. Rf3 {2.74/
15 29 (Qe3)} Bxf5 {0.46/16 38 (Qe1)} 22. gxf5 {3.00/14 28} Rh8+ {0.46/15 0} 23.
Kg3 {3.20/13 28} Qh1 {0.46/16 88 (Qe1+)} 24. Kf2 {2.78/13 27} Rh2+ {0.46/15 65}
25. Ke3 {2.72/16 26} Qe1 {0.46/15 0} 26. f6+ {2.25/14 25 (Qd2)} Kf8 {0.31/17 18
} 27. Qd2 {2.04/14 25} Qa1 {0.31/17 49} 28. Qc3 {1.55/13 24 (a4)} Qa2 {0.00/17
8} 29. Qe1 {2.05/13 23 (Qd3)} Qc4 {0.00/16 22 (g4)} 30. Qd1 {1.57/12 23 (a4)}
Ba7 {0.00/15 13} 31. Rf5 {1.05/13 22} Qe6+ {0.00/15 18} 32. Re5 {0.12/15 11}
Qh3+ {0.00/16 23 (Rh3+)} 33. Kd2 {1.14/14 21} Qxa3 {0.00/15 0} 34. Rxg5 {0.88/
13 21} Rh3 {0.00/16 6} 35. Ke1 {0.88/13 20} Qe3 {0.00/16 16 (Bxc5)} 36. Rf5 {
0.00/15 15} Qe4 {0.00/16 11} 37. Rf4 {0.00/15 10} Rh1+ {0.00/16 5} 38. Rf1 {0.
00/15 10} Rh2 {0.00/17 31 adjud. Draw accepted} 1/2-1/2

Rybka played the hole game without Black Bishop! here is the final standing:

5k2/b1p5/2p2P2/2P5/3Pq3/8/4N2r/3QKR2 w - - 0 39


Actually she played also without white bishop, but she can finally unlock it due to ST weak play level.
Parent - - By Uri Blass (*****) [il] Date 2009-10-11 15:53
Rybka was clearly handicapped in this game
Maybe rybka could win this game without adjudication and rybka did not play without a bishop but only played with a bad bishop.

remove the bishop from the board and I believe that rybka is going to score worse.
Parent - By Uri Blass (*****) [il] Date 2009-10-11 17:41
I can add that rybka probably started from a lost position

You should play the same game with opposite colors if you are interested in being fair
Here is the opinion of rybka3 960 and there is no big difference between 960 and not 960:

It shows 0.46 pawns for white at depth 17 and cannot find a better move than h6

SmarThink 1.10 - Rybka 3, 15?+10? PC 2009
r2q1rk1/ppp2ppp/1bnp1nb1/8/2PP1PP1/P1N4P/1P2N1B1/R1BQK2R b KQ - 0 1


Analysis by Rybka 3 960 1-cpu 32-bit :

1...Nf6-d7
  =  (-0.09)   Depth: 2   00:00:00
1...h7-h5
  =  (-0.10)   Depth: 2   00:00:00
1...Nf6-d7
  =  (0.01)   Depth: 3   00:00:00
1...h7-h6
  =  (-0.03)   Depth: 3   00:00:00
1...h7-h6
  =  (-0.15)   Depth: 4   00:00:00
1...h7-h6 2.f4-f5
  =  (-0.22)   Depth: 5   00:00:00
1...h7-h6 2.f4-f5 Bg6-h7 3.0-0 Qd8-e7
  =  (-0.03)   Depth: 6   00:00:00  3kN
1...h7-h6 2.f4-f5 Bg6-h7 3.0-0 Nc6-a5 4.b2-b3 Qd8-d7 5.Bc1-f4
  =  (-0.03)   Depth: 7   00:00:00  16kN
1...h7-h6 2.f4-f5 Bg6-h7 3.0-0 Nc6-a5 4.b2-b3 Qd8-d7 5.Bc1-f4
  =  (-0.03)   Depth: 8   00:00:01  31kN
1...h7-h6 2.f4-f5 Bg6-h7 3.0-0 Rf8-e8 4.Bc1-f4 Qd8-d7 5.b2-b4 a7-a6 6.c4-c5
  =  (-0.02)   Depth: 9   00:00:02  59kN
1...h7-h6 2.f4-f5 Bg6-h7 3.0-0 Rf8-e8 4.Bc1-f4 Qd8-d7 5.b2-b4 a7-a6 6.c4-c5
  =  (-0.02)   Depth: 10   00:00:05  110kN
1...h7-h6 2.f4-f5 Bg6-h7 3.0-0 Nc6-a5 4.b2-b3 Qd8-d7 5.Bc1-f4 Ra8-e8 6.Rf1-e1 a7-a6 7.Bf4-g3
  =  (0.02)   Depth: 11   00:00:15  342kN
1...h7-h6 2.f4-f5 Bg6-h7 3.0-0 Nc6-a5 4.b2-b3 Qd8-d7 5.Bc1-f4 Ra8-e8 6.Rf1-e1 a7-a6 7.Bf4-g3
  =  (0.02)   Depth: 12   00:00:25  560kN
1...h7-h6 2.f4-f5 Bg6-h7 3.0-0 Nc6-a5 4.b2-b3 Qd8-d7 5.Bc1-f4 Ra8-e8 6.Rf1-e1 a7-a6 7.Bf4-g3
  =  (0.02)   Depth: 13   00:00:55  1256kN
1...h7-h6 2.f4-f5 Bg6-h7 3.0-0 Nc6-a5 4.b2-b3 Qd8-d7 5.Bc1-f4 Ra8-e8 6.Rf1-e1 a7-a6 7.Bf4-g3
  =  (0.02)   Depth: 14   00:01:23  1862kN
1...h7-h6 2.f4-f5 Bg6-h7 3.0-0 Nc6-a5 4.b2-b3 Qd8-e7 5.Bc1-f4 a7-a6 6.Qd1-c2 Na5-c6 7.Qc2-d2 Nc6-a5
  =  (0.06)   Depth: 15   00:02:39  3608kN
1...h7-h6 2.f4-f5 Bg6-h7 3.0-0 Nc6-a5 4.b2-b3 Qd8-e7 5.Bc1-f4 Rf8-e8 6.Qd1-c2 Na5-c6 7.Qc2-d2 a7-a6 8.Ra1-e1
  =  (0.07)   Depth: 16   00:07:50  10869kN
1...h7-h6 2.f4-f5 Bg6-h7 3.b2-b4 a7-a6 4.0-0 Qd8-d7 5.Bc1-e3 Rf8-e8 6.Qd1-d2 Bb6-a7 7.Ra1-e1 Ra8-d8 8.Be3-f2 Rd8-a8 9.Ne2-f4
  ²  (0.46)   Depth: 17   00:47:54  68699kN

(so k,  11.10.2009)
Parent - - By InspectorGadget (*****) [za] Date 2009-10-12 20:16
Great posts Xaggard. The point you are making is very simple and it is self explanatory. Rybka does not understand the position, plain as that. It is not a matter of she loses when she is black. It is a matter of understanding the position that it is lost for black and show the correct evaluation, simple as that :)
Parent - By George Tsavdaris (****) Date 2009-10-12 21:27
Rybka 3:

1...h6 2.f5 Bh7[] 3.b4 a6 4.0-0 Qd7 5.Be3 Rae8 6.Qd2 Qe7 7.Bf4 Rd8 8.Rf3
  +/=  (0.30)   Depth: 17   00:01:55  28595kN

That means Rybka's 3 believes it will score from this position as white, about 62.7%.
That means Rybka believes white is a lot better here.

Having that in mind why do you say the:

>Rybka does not understand the position, plain as that.
>It is a matter of understanding the position that it is lost for black and show the correct evaluation, simple as that


And what would be the correct evaluation for you?
A +1.20 for example? And why not a +1.96? A +2.20 maybe? Will you be satisfied with that?
With what criteria do you prefer the evaluation to be a certain value in order to be correct?
Parent - - By Xaggard (**) Date 2009-10-11 15:50
example number 2

Deep Fritz 10 (not 11 or 12 - they are garbage!) running on Single CPU vs Rybka 3 (4 CPU)

[Event "15м+10с"]
[Site "PC"]
[Date "2009.10.11"]
[Round "1"]
[White "Deep Fritz 10"]
[Black "Rybka 3"]
[Result "1-0"]
[Annotator "1.22;-0.06"]
[SetUp "1"]
[FEN "r2q1rk1/ppp2ppp/1bnp1nb1/8/2PP1PP1/P1N4P/1P2N1B1/R1BQK2R b KQ - 0 1"]
[PlyCount "31"]
[TimeControl "900+10"]

1... h6 {-0.06/15 29} 2. f5 {1.22/15 42} Bh7
{-0.06/13 0} 3. b4 {1.18/16 86 (0-0)} a6 {0.01/14 17} 4. Be3 {1.18/15 75 (0-0)}
Re8 {0.14/14 28} 5. Bf2 {1.48/15 52 (Qd3)} h5 {0.16/14 30} 6. c5 {1.78/15 47
(Bf3)} Ba7 {0.13/13 48 (hxg4)} 7. Bh4 {1.59/14 28 (Bf3)} dxc5 {0.71/15 244
(Qe7)} 8. Bxc6 {1.76/14 25 (dxc5)} bxc6 {-0.09/15 21} 9. bxc5 {1.51/16 49
(dxc5)} Qe7 {0.00/14 37} 10. Qc1 {1.46/15 27 (Bg5)} a5 {0.12/15 59 (Rab8)} 11.
Rb1 {2.18/15 23 (gxh5)} Qd8 {0.36/14 22 (Kh8)} 12. O-O {2.35/15 30 (Qf4)} hxg4
{0.61/15 12 (Qe7)} 13. hxg4 {2.72/15 15} Kh8 {0.75/16 52 (Qe7)} 14. Rf3 {3.33/
16 45 (Re1)} Rc8 {1.04/13 19 (Qe7)} 15. Rh3 {4.18/14 9 (Qc2)} Kg8 {2.94/13 18}
16. Qf1 {4.48/15 12 (Rb7)} Qe7 {2.71/14 17 adjud.} 1-0

2r1r1k1/b1p1qppb/2p2n2/p1P2P2/3P2PB/P1N4R/4N3/1R3QK1 w - - 0 17


It takes only 15 moves for Fritz 10 (1 CPU) to beat Rybka 3 on 4 threads.
Parent - - By George Tsavdaris (****) Date 2009-10-11 16:07
I don't see a win of DF10 here.
Adjudicating the game as 1-0 here is a clear mistake. Rybka may(or may not) win the game.
For example look at the following position from the following game:

Fritz(the black player) played Qd5 here with -4.24 evaluation.
Rybka(the white player) thought she is losing with a -1.31 evaluation.
Yet Rybka won the game.
4rr1k/P1n4p/1n6/1p1q1bpN/2ppp3/6P1/3N1PBP/R1Q1R1K1 w - - 0 39


[Event "15'/60"]
[Site "??"]
[Date "??"]
[Round "1"]
[White "Rybka 1.01 Beta"]
[Black "Fritz 9"]
[Result "*"]
[ECO "A82"]
[Annotator "??"]
[PlyCount "147"]
[TimeControl "60/900:0"]

{Intel(R) Pentium(R) 4 CPU 2.40GHz 2394 MHz W=11.2 ply; 46kN/s B=11.8 ply;
874kN/s} 1. d4 f5 2. e4 {Both last book move} fxe4 {0.22/12 11} 3. Nc3 {-0.12/
12 21} Nf6 {0.12/13 35} 4. Nh3 {(Bg5) -0.15/12 30} d5 {-0.48/12 18} 5. Nf4 {
(f3) -0.16/12 18} e6 {(Nc6) -0.60/12 15} 6. Be3 {(g4) -0.19/12 30} Bd6 {-1.08/
12 18} 7. Be2 {-0.22/11 6} O-O {-1.25/12 14} 8. Qd2 {-0.23/12 25} c6 {(Nc6) -1.
32/12 20} 9. O-O {(h4) -0.21/11 27} e5 {(Na6) -1.82/12 18} 10. dxe5 {-0.45/10
10} Bxe5 {-1.91/11 11} 11. Bc5 {(Rad1) -0.61/10 19} Qc7 {(Rf7) -1.71/11 14} 12.
g3 {-0.69/12 29} Rd8 {(Bxf4) -1.74/11 29} 13. Bd4 {(f3) -0.64/11 26} Bxd4 {-1.
83/12 13} 14. Qxd4 {-0.79/12 10} Qb6 {-1.76/12 11} 15. Qd2 {(Qxb6) -0.81/12 25}
Qxb2 {(Na6) -2.39/12 13} 16. a3 {-0.78/11 43} Qb6 {-2.59/13 28} 17. Rab1 {(Na4)
-0.90/10 17} Qc5 {-2.71/12 16} 18. Rb4 {(Na4) -1.08/10 22} Re8 {(Na6) -3.14/12
31} 19. Na4 {(Qe3) -1.22/10 18} Qd6 {(e3) -3.36/11 15} 20. c4 {(Nb2) -1.35/11
26} Na6 {(c5) -3.32/11 15} 21. Rbb1 {-1.74/11 21} d4 {-3.68/12 19} 22. Rfd1 {
(f3) -1.91/11 32} c5 {(Rd8) -4.11/12 20} 23. Re1 {(f3) -1.68/10 10} Qc6 {(Nc7)
-4.44/12 13} 24. Nb2 {-2.05/11 23} Bf5 {(Bg4) -4.41/11 10} 25. Ra1 {(Bf1) -2.
10/12 17} Nc7 {(Ng4) -4.78/12 13} 26. Bf1 {-2.22/10 10} Rf8 {(Ne6) -4.78/12 15}
27. Bg2 {-2.12/11 11} Rae8 {(Ne6) -4.69/12 23} 28. Nbd3 {(h3) -2.22/10 11} Qd6
{(Ne6) -5.15/12 17} 29. Nc1 {(Nb2) -2.53/11 8} a6 {(Ne6) -5.19/11 18} 30. Nb3 {
-2.39/10 14} b6 {(Ne6) -4.84/11 21} 31. a4 {-2.33/10 14} Nd7 {(Ne6) -4.81/11 17
} 32. a5 {-2.02/10 11} b5 {(Ne5) -4.44/11 15} 33. cxb5 {-2.26/10 23} axb5 {
(Nxb5) -4.38/11 12} 34. a6 {(Ne2) -2.01/9 10} Nb6 {(Nxa6) -4.86/10 9} 35. a7 {
-1.49/9 6} Kh8 {(Nc4) -4.49/10 14} 36. Qc1 {-1.24/10 17} c4 {(Nc4) -4.66/11 9}
37. Nd2 {-1.50/11 8} g5 {-3.75/11 22} 38. Nh5 {-1.31/11 17} Qd5 {(e3) -4.24/10
10} 39. Qa3 {(Bxe4) -0.90/11 21} Rf7 {(Kg8) -4.02/10 11} 40. Qa5 {-1.06/10 7}
Nba8 {-3.71/11 11} 41. g4 {-0.88/11 13} Bg6 {(Bxg4) -3.36/11 16} 42. Ng3 {-0.
88/11 12} Rf4 {-3.07/11 14} 43. Qb4 {-1.02/10 13} Kg8 {-3.40/10 15} 44. h3 {
(Ra5) -1.14/9 8} Qc6 {(d3) -2.75/10 13} 45. Ne2 {-0.87/10 19} Ne6 {-2.65/11 20}
46. Nxf4 {(Qb1) -0.81/10 10} Nxf4 {-2.79/12 14} 47. Bf1 {-1.02/10 9} d3 {-2.96/
10 10} 48. Nb1 {-1.28/10 18} Nd5 {-3.66/12 19} 49. Qd2 {-0.95/10 5} Qf6 {(e3)
-3.66/12 10} 50. Ra5 {-1.44/11 9} c3 {-3.67/12 14} 51. Qa2 {-1.50/11 6} Bf7 {
-2.04/10 10} 52. Rxb5 {-1.43/11 7} Ndc7 {-3.36/12 22} 53. Rb3 {-1.76/12 8} c2 {
(Qe6) -2.46/12 18} 54. Qb2 {-0.72/9 2} Qf4 {(Kg7) -1.76/10 8} 55. Rb8 {0.00/10
9} Na6 {(Ne6) -1.29/8 7} 56. Rxe8+ {1.14/8 4} Bxe8 {0.87/10 5} 57. Bxd3 {(Qa2+)
1.32/10 5} exd3 {0.39/12 9} 58. Rxe8+ {1.49/10 8} Kf7 {0.91/12 9} 59. Re3 {2.
14/11 21} Qf6 {1.34/12 17} 60. Qb3+ {2.69/12 16} Kg7 {1.66/11 2} 61. Re1 {3.11/
13 34} cxb1=B {(Nc5) 2.10/11 15} 62. Qxb1 {3.60/12 36} Nc5 {2.30/13 32} 63. Qb8
{4.02/13 43} d2 {3.46/14 30} 64. Rd1 {4.80/14 150} Nd3 {3.94/13 25} 65. Rxd2 {
4.85/13 42} Qa1+ {4.27/14 31} 66. Kh2 {5.09/14 45} Qe5+ {4.28/14 23} 67. Qxe5+
{5.09/15 51} Nxe5 {4.40/16 34} 68. Rd5 {5.15/17 46} Nc6 {(Nf7) 4.55/16 30} 69.
Rd7+ {(Rxg5+) 5.29/17 29} Kg8 {(Kf8) 3.65/17 24} 70. Kg3 {5.33/19 81} Nxa7 {
(Kf8) 3.82/16 22} 71. Rxa7 {5.29/18 12} Nb6 {4.28/16 31} 72. f4 {(Kf3) 5.42/18
47} gxf4+ {(Nd5) 4.46/15 35} 73. Kxf4 {5.85/16 9} Nc4 {(Kf8) 4.95/14 25} 74. h4
{adjud. 6.51/19 28} *
Parent - - By Xaggard (**) Date 2009-10-11 16:28

> I don't see a win of DF10 here.
> Adjudicating the game as 1-0 here is a clear mistake. Rybka may(or may not) win the game.


Very funny.
Do you have DF10? Run the game and ensure rybka has no chances to win at all!
What I want to explain: Rybka - running on 4 threads - can nothing to do against engines (even on single thread) that understand this position better.
Parent - - By George Tsavdaris (****) Date 2009-10-11 18:01 Edited 2009-10-11 18:08

>Do you have DF10? Run the game and ensure rybka has no chances to win at all!


Do other engines like DS12 for example have any chances against Rybka 3 when Rybka is white?
Can you run the same match but with Rybka 3 to be white? I guess the score will be 1-0 for Rybka 3.

Are you going then to say that Deep Shredder 12 does not understand this position?
Parent - - By Xaggard (**) Date 2009-10-11 18:21
You do not want to understand me, or simply can't do it. Sorry for my bad English. :)

r2q1rk1/ppp2ppp/1bnp1nb1/8/2PP1PP1/P1N4P/1P2N1B1/R1BQK2R b KQ - 0 1


Analysis by Deep Shredder 12 x64:

1...Qe8 2.0-0 Be4 3.Nxe4 Nxe4 4.Qd3 f5 5.gxf5 Nf6 6.b4 a6 7.Bb2 Qd7 8.c5 Ba7 9.Ng3 d5 10.Kh2 Ne7 11.Rfe1 Bb8 12.Re6 c6 13.Rae1
  (1.35)   13/39   00:00:02  3933тП
----------------------

r2q1rk1/ppp2ppp/1bnp1nb1/8/2PP1PP1/P1N4P/1P2N1B1/R1BQK2R b KQ - 0 1


Analysis by Rybka 3:

1...h6 2.f5 Bh7[] 3.Qd3 Re8 4.Kd2 d5 5.c5 Bxc5 6.dxc5 Ne5 7.Qd4 Nc6 8.Qd3 Ne5 9.Qd4 Nc6 10.Qd3 Ne5 11.Qd4 Nc6 12.Qd3 Ne5 13.Qd4 Nc6 14.Qd3 Ne5 15.Qd4 Nc6 16.Qd3 Ne5
  (0.00)   15   00:00:16  2458тП

I wanted to show you only this difference. Rybka and other «fruit-based» engines (like all Togas, Fritz 11 & 12 - they are Rybka 2 «clones»  underestimate white's play here, from the early beginning. Of course later Rybka will see the things right, but unfortunatly, too late for her.
Parent - - By George Tsavdaris (****) Date 2009-10-11 18:30 Edited 2009-10-11 18:35

>Analysis by Rybka 3:


>1...h6 2.f5 Bh7[] 3.Qd3 Re8 4.Kd2 d5 5.c5 Bxc5 6.dxc5 Ne5 7.Qd4 Nc6 8.Qd3 Ne5 9.Qd4 Nc6 10.Qd3 Ne5 11.Qd4 Nc6 12.Qd3 Ne5 13.Qd4 >Nc6 14.Qd3 Ne5 15.Qd4 Nc6 16.Qd3 Ne5
>  (0.00)   15   00:00:16  2458тП


Oh 16 seconds and 2 seconds for Shredder. Nice analysis method.:-P And nice method of getting your point to be seem valid. :-P
I would only be satisfied if i saw exact analysis times for both engines as also a considerable amount of time let's say 3 minutes on a normal quad.

Also:

>I wanted to show you only this difference. Rybka and other «fruit-based» engines (like all Togas, Fritz 11 & 12 - they are Rybka 2 >«clones»   )   underestimate white's play here, from the early beginning.


But you were using games where Rybka 3 lost as black from this position to make your point and you even half based your point on these games.
And i say with the same logic: Show us the games from this position of R3-DS12 and R3-DF10 and show us the result.
And if the result is 1-0 (as i suspect it will be) then are you going to allow me to say that DS12 and DF10 do not understand the position? :-)
Parent - By Xaggard (**) Date 2009-10-11 18:50
I examine this position from the advanced chess player point of view.
Let's consider you play a corrgame using your Rybka. You trust her evaluation and, following her hints, you came to this position. As we both may see this position is totally lost for Black. And even Rybka can «understand» it but too late to save your game.
When using Shredder, instead of Rybka or additionaly to Rybka, :) you can initially avoid this trap.
Shredder can spare a lot of your time in such game.
What is this if not a defect of rybka evaluation?
Parent - By InspectorGadget (*****) [za] Date 2009-10-13 11:05

> Can you run the same match but with Rybka 3 to be white? I guess the score will be 1-0 for Rybka 3.


Of course Rybka should win it with white because Xaggard has been saying that the position favours white :)

> Are you going then to say that Deep Shredder 12 does not understand this position?


No, Shredder already understands that the position is lost for black :)
Parent - By Xaggard (**) Date 2009-10-11 16:44
example #3

[Event "15м+10с"]
[Site "PC"]
[Date "2009.10.11"]
[Round "1"]
[White "Stockfish 1.5.1 64bit (1 CPU)"]
[Black "Rybka 3 (4 CPU)"]
[Result "1-0"]
[Annotator "0.84;-0.09"]
[SetUp "1"]
[FEN "r2q1rk1/ppp2ppp/1bnp1nb1/8/2PP1PP1/P1N4P/1P2N1B1/R1BQK2R b KQ - 0 1"]
[PlyCount "110"]
[TimeControl "900+10"]

1... h6 {-0.09/15 27} 2. b4 {0.84/16 28} a6
{0.07/15 49} 3. O-O {0.76/17 46 (f5)} Bh7 {0.22/16 39} 4. f5 {0.80/17 31} Qd7 {
0.35/16 116 (Ba7)} 5. Be3 {1.13/16 48} Rfe8 {0.42/15 24 (Rae8)} 6. Bf2 {1.17/
16 24 (Qd2)} Kh8 {0.53/16 43 (Qe7)} 7. Bh4 {1.73/16 25 (Nf4)} Ng8 {0.70/14 60
(Ba7)} 8. c5 {1.77/16 30} Ba7 {0.76/15 43} 9. Rc1 {1.61/15 43 (Nf4)} h5 {0.65/
14 23 (g5)} 10. Nd5 {1.81/16 68 (Nf4)} hxg4 {0.63/14 35} 11. hxg4 {2.02/17 46}
dxc5 {0.63/15 19} 12. bxc5 {1.77/17 43} Nce7 {0.68/15 36} 13. Nxe7 {1.61/17 82}
Nxe7 {0.90/16 38} 14. Bxb7 {1.73/17 28} Rab8 {0.95/17 71} 15. Bf3 {1.73/16 27}
Nd5 {0.95/17 43} 16. Qd2 {1.89/16 25 (Qd3)} Rb3 {0.98/15 64 (f6)} 17. Rb1 {1.
57/16 48 (Nc3)} Rxa3 {0.79/14 11} 18. Ra1 {1.57/16 21} Rxa1 {0.90/15 19} 19.
Rxa1 {1.57/15 4} Bb8 {1.05/15 47 (c6)} 20. Nc3 {2.62/18 35 (Rxa6)} Nb4 {0.95/
16 87 (Nxc3)} 21. Ra4 {2.50/17 31 (Ne4)} Nc6 {1.13/15 17} 22. Be4 {2.50/17 92
(Qd1)} a5 {0.98/15 19} 23. d5 {2.54/16 17 (Qd1)} Ne5 {1.06/15 29} 24. Qd1 {2.
46/16 20 (Qg5)} f6 {1.21/13 12 (Kg8)} 25. Rxa5 {2.50/17 30} Rf8 {1.26/14 27
(Qe7)} 26. Ra8 {2.66/15 13 (Bg3)} Qe7 {1.49/14 15 (Rd8)} 27. Bf2 {2.78/15 15}
Rd8 {1.59/15 20} 28. Bd4 {3.23/17 31} Nf7 {1.69/16 17 (Qe8)} 29. Kg2 {3.23/16
14 (Qb1)} Ng5 {1.75/15 14 (Qf8)} 30. Bd3 {3.47/18 38} Qf8 {1.75/16 40 (Qe8)}
31. Qb1 {3.51/17 14 (Bb5)} Bg8 {2.28/15 22 (Qe8)} 32. Qh1+ {3.87/17 16} Nh7 {
2.57/16 24} 33. Be4 {4.28/17 16} c6 {2.63/16 23 (Re8)} 34. d6 {4.56/17 13} Bxd6
{2.63/17 10} 35. Rxd8 {4.64/18 15} Qxd8 {2.63/16 0} 36. cxd6 {4.60/18 16} Qxd6
{2.63/16 0} 37. Be3 {4.56/19 19} Qb4 {2.71/17 13} 38. Qc1 {4.56/17 15} c5 {2.
84/17 24 (Nf8)} 39. Kf2 {4.72/17 21 (Qd2)} Bf7 {2.56/15 16} 40. Bf3 {5.01/17 11
} c4 {2.94/15 14} 41. Ne4 {5.05/17 11 (Qd2)} Bg8 {2.59/16 17} 42. Be2 {5.09/17
15 (Qd2)} Qb7 {2.76/16 12 (Qb3)} 43. Nd6 {5.81/19 29} Qc7 {2.76/15 14} 44. Bf4
{5.81/18 14} Qa7+ {3.30/15 13 (Qb6+)} 45. Qe3 {5.81/18 14} Qa4 {3.73/16 10} 46.
Qd4 {6.02/18 14 (Qe8)} Qa3 {3.55/15 10 (Qd7)} 47. Nxc4 {6.10/17 10} Qe7 {3.76/
16 9 (Qa4)} 48. Nd2 {6.22/17 9 (Nd6)} Nf8 {3.87/16 12 (Bf7)} 49. Qd6 {6.38/18
10} Qe8 {4.00/17 8} 50. Nf3 {6.46/17 13 (Qb8)} Qc8 {3.98/17 11 (Ba2)} 51. Nd4 {
6.58/16 9 (Qb8)} Ba2 {4.04/15 12 (Bf7)} 52. Bd2 {6.78/17 12 (Bf3)} Bg8 {4.62/
14 12} 53. Bb5 {6.86/17 11 (Bb4)} Bf7 {4.62/13 11 (Qa8)} 54. Nc6 {8.40/18 21}
Qa8 {5.11/13 11} 55. Ne7 {8.92/18 22 (Be2)} Qh1 {5.12/12 10} 56. Be2 {9.97/18
10} 1-0


what else?
Parent - - By rivaldo (***) [de] Date 2009-10-11 17:40
there's a simple way of getting a more reliable eval on this position than rybkas, fritz or whatelses.

ask 10 GM whether they think this position objectively is a win for white or a draw.

I can tell you, that the human eval will be 'won for white'.
Parent - - By George Tsavdaris (****) Date 2009-10-11 17:56 Edited 2009-10-11 18:36

>there's a simple way of getting a more reliable eval on this position than rybkas, fritz or whatelses.


>ask 10 GM whether they think this position objectively is a win for white or a draw.


I don't believe that the opinions of 10 GMs or even 100 GMs, even if all of them have the same opinion, can always show the absolute truth for a position(if it's won, lost for white or draw).

I also don't believe that the opinions of 10 GMs of a position, even if it's the same, is more reliable than the opinion of Rybka or Fritz or Shredder, etc....

>I can tell you, that the human eval will be 'won for white'.


You can't tell me that, because you are not 10 GMs(you do not have the wisdom of 10 GMs).
Parent - By rivaldo (***) [de] Date 2009-10-12 04:09
of course the GMs don't know the absolute truth, but their abilities to evaluate such a position are much more subtle than a simple algorithm.
here's really not much to calculate, or something, that could be overlooked by the GMs.
my own assessment of the position is good enough to know what the GMs would say.
Parent - By Geomusic (*****) Date 2009-10-13 08:42
lol at this convo
Parent - - By Jim Walker (***) [us] Date 2009-10-11 13:32
The main difference between Rybka and Rybka cluster is that Rybka cluster comes to the wrong evaluation much faster ! :)
Parent - By InspectorGadget (*****) [za] Date 2009-10-13 10:36

> The main difference between Rybka and Rybka cluster is that Rybka cluster comes to the wrong evaluation much faster !  :-)


Very creative, LOL
Parent - By JohnL (***) Date 2009-10-11 17:23
Nice example of what seems to be a serious mis-evaluation!

I just ran a randomizer match R3 default-R3 default, 10cp window, 6 ply.

The score was a pretty convincing +60=34-6 <=>77% for white.
Parent - - By jpe (*) Date 2009-10-11 21:08
Well, as others have already said, any deeper analysis of the position by R3 would show a definitive advantage for White. There are many possibilities for Black at moves 4 and 5. Here I just picked up one variation that looked quite promising...

[Event "?"]
[Site "?"]
[Date "1917.01.01"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]
[FEN "r2q1rk1/ppp2ppp/1bnp1nb1/8/2PP1PP1/P1N4P/1P2N1B1/R1BQK2R b KQ - 0 1"]

1... h6 2. f5 Bh7 3. b4 a6 4. O-O Qd7 5. Be3 Kh8 6. Qd2 Rae8 7. Rae1 Rg8 8. Bf2
Ba7 9. Bh4 Qd8 *

After 5. Be3:

[+0.64]  d=18  5...Kh8 6.Nf4 (0:13:17) 157265kN
[+0.64]  d=17  5...Kh8 6.Nf4 (0:13:17) 157265kN
[+0.65]  d=17  5...Rae8 6.Bf2 Qc8 (0:06:00) 70920kN
[+0.61]  d=16  5...Rad8 6.Qd2 (0:02:50) 33087kN
[+0.64]  d=16  5...Rae8 6.Qd2 Re7 7.Rae1 Rfe8 (0:01:56) 22967kN
[+0.65]  d=16  5...Ba7 6.Qd2 Kh8 7.Rae1 Rad8 8.Nf4 Rfe8 9.Bf2 Bg8 10.Ncd5 Nxd5 11.Nxd5 f6 (0:01:22) 16051kN
[+0.60]  d=15  5...Ba7 6.Qd2 Rfe8 7.Rae1 Qd8 8.Nd5 Rb8 9.Nec3 Rxe3 10.Rxe3 Nxd5 11.Nxd5 (0:00:44) 8441kN
[+0.61]  d=15  5...Rae8 6.Bf2 Qd8 (0:00:41) 7770kN
[+0.62]  d=15  5...Rfe8 6.Bf2 Qe7 7.Qd2 (0:00:30) 5661kN

After 5...Kh8 6. Qd2:

[+0.70]  d=18  6...Rae8 7.Bf2 h5 (0:04:18) 56442kN
[+0.64]  d=17  6...Rae8 7.Bf2 h5 (0:00:08) 1773kN
[+0.64]  d=16  6...Rae8 7.Bf2 h5 (0:00:04) 1126kN
[+0.64]  d=15  6...Rae8 7.Bf2 h5 (0:00:04) 971kN
[+0.75]  d=15  6...Rad8 7.Nf4 (0:00:03) 846kN

After 6...Rae8:

[+0.83]  d=18  7.Rae1 Qc8 8.c5 Ba7 (0:08:13) 107457kN
[+0.84]  d=17  7.Rae1 Ba7 (0:03:15) 42039kN
[+0.74]  d=16  7.Rae1 Ba7 (0:00:37) 8367kN
[+0.74]  d=15  7.Rae1 Nb8 8.Nd5 Nxd5 (0:00:14) 3009kN

After 7. Rae1:

[+0.78]  d=19  7...Rg8 8.c5 dxc5 9.dxc5 (0:10:10) 139576kN
[+0.67]  d=18  7...Rg8 8.c5 Ba7 9.Nf4 (0:03:24) 43950kN
[+0.69]  d=17  7...Rg8 8.c5 Ba7 9.Nf4 Rd8 10.Nfd5 Nxd5 (0:03:01) 39001kN
[+0.70]  d=17  7...Ba7 8.Nd5 Nxd5 (0:02:14) 28979kN
[+0.79]  d=17  7...Qd8 8.Nd5 Ba7 9.Nec3 (0:00:41) 9198kN

After 7...Rg8:

[+0.87]  d=19  8.Bf2 Ba7 9.Nf4 (0:10:03) 133190kN
[+0.89]  d=18  8.Bf2 Rgf8 9.Nf4 Bg8 (0:04:01) 52999kN
[+0.94]  d=17  8.Bf2 Ba7 9.Nf4 (0:01:10) 15034kN
[+0.91]  d=16  8.Bf2 Ba7 9.Nf4 (0:00:26) 5464kN
[+0.91]  d=15  8.Bf2 Ba7 9.Nf4 (0:00:12) 2803kN

After 8. Bf2:

[+1.00]  d=19  8...Ba7 9.Nf4 (0:24:02) 322794kN
[+0.87]  d=18  8...Ba7 9.Nf4 (0:00:19) 3390kN

After 8...Ba7:

[+1.27]  d=19  9.Bh4 Qd8 10.Kh1 (0:16:01) 210350kN
[+1.23]  d=18  9.Bh4 Qd8 10.Kh1 Ne4 (0:07:25) 96831kN
[+1.20]  d=18  9.Bh4 (0:02:50) 32909kN
[+1.00]  d=18  9.Nf4 Rxe1 10.Rxe1 Re8 (0:00:18) 3609kN

After 9. Bh4:

[+1.34]  d=19  9...Qd8 10.Kh1 Bb6 11.Qd3 (0:12:17) 156351kN
[+1.23]  d=18  9...Qd8 10.Kh1 Ne4 11.Qxh6 (0:02:33) 32998kN
[+1.16]  d=17  9...Qd8 10.Kh1 Ne4 (0:01:07) 13899kN
[+1.15]  d=16  9...Qd8 10.Kh1 Ne4 11.Qxh6 gxh6 (0:00:28) 5736kN
Parent - By Mark (****) [us] Date 2009-10-12 13:42
Here's a 12-hour IDeA analysis with d=15 showing a white advantage.
- - By rocket (***) [se] Date 2015-05-26 23:20 Edited 2015-05-26 23:27
People in here don't understand computer chess. The reason Rybka 3 initially evalutes this as microscopically better for black (instead of white!) at lower depths, is because of the pawn advancements around whites own king, which R3 evalutes as compromising (weakening) its own king safety. Thats also why its still not higher than around +0.68 at greater depth. It is strange though that it doesn't view the bishop inactivity for black after f5 (locked in) as immediately worse than pawn pushes around whites own king (which is not dogmatically bad).
Parent - By Labyrinth (*****) [us] Date 2015-05-27 01:20
Epic necro
Up Topic Rybka Support & Discussion / Rybka Discussion / Wrong Rybka 3 evaluation - what about cluster?

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill