Not logged inRybka Chess Community Forum
Forum RybkaChess Help Search Register Login
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka - Ehlvest (game 3)
- - By Vasik Rajlich (Silver) [hu] Date 2007-03-06 23:02
Rybka plays without f-pawn here.

Vas
Parent - - By Henrik Dinesen (***) [dk] Date 2007-03-06 23:31
What a suprise :-)
Parent - By Eduard (***) [de] Date 2007-03-07 00:58
I do not understand Ehlvest.
It plays very aggressively. Thus no humans play against a computer!
Parent - By Alkelele (***) Date 2007-03-06 23:32
[Event "?"]
[Site "?"]
[Date "2007.03.07"]
[Round "?"]
[White "New game"]
[Black "Rybka 2.2n2 mp Local"]
[Result "*"]
[Annotator "Dagh,Nielsen"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPP1PP/RNBQKBNR w KQkq - 0 1"]
[PlyCount "14"]

{1024MB, PlayChess2006.ctg, REBEKKA} 1. Nf3 d5 2. d4 Nf6 3. c4 e6 4. Nc3 Be7 5.
e3 O-O 6. Bd3 c5 7. O-O Nc6 *

r1bq1rk1/pp2bppp/2n1pn2/2pp4/2PP4/2NBPN2/PP4PP/R1BQ1RK1 w - - 0 8
Parent - - By Alkelele (***) Date 2007-03-07 00:26
[Event "?"]
[Site "?"]
[Date "2007.03.07"]
[Round "?"]
[White "New game"]
[Black "Rybka 2.2n2 mp Local"]
[Result "*"]
[Annotator "Dagh,Nielsen"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPP1PP/RNBQKBNR w KQkq - 0 1"]
[PlyCount "41"]

{1024MB, PlayChess2006.ctg, REBEKKA} 1. Nf3 d5 2. d4 Nf6 3. c4 e6 4. Nc3 Be7 5.
e3 O-O 6. Bd3 c5 7. O-O Nc6 8. dxc5 Bxc5 9. Bd2 a6 10. Qe2 Ba7 11. Rad1 Qe7 12.
Kh1 dxc4 13. Bxc4 Ng4 14. h3 Nge5 15. Bb3 b5 16. Ne4 Bb7 17. Bc3 Bb8 18. Nfg5
h6 19. Qh5 b4 20. Be1 Bc8 21. Bh4 *

rbb2rk1/4qpp1/p1n1p2p/4n1NQ/1p2N2B/1B2P2P/PP4P1/3R1R1K b - - 0 21


It looks critical for black again...
Parent - - By Uri Blass (*****) [il] Date 2007-03-07 00:45 Edited 2007-03-07 00:54
It seems to be 3-0 based on this position

Analysis by rybka2.3.1

New game - Rybka 2.2n2 mp Local,  2007
rbb2rk1/4qpp1/p1n1p2p/4n1NQ/1p2N2B/1B2P2P/PP4P1/3R1R1K b - - 0 1


Analysis by Rybka 2.3.1 32-bit :

21...Qc7
  ²  (0.35)   Depth: 2   00:00:00
21...Qa7
  =  (0.11)   Depth: 2   00:00:00
21...f6
  =  (-0.23)   Depth: 2   00:00:00
21...f6
  =  (-0.19)   Depth: 3   00:00:00
21...f6
  ³  (-0.33)   Depth: 4   00:00:00
21...f6 22.Nf3
  ³  (-0.35)   Depth: 5   00:00:00
21...f6 22.Nf3 Bd7
  ³  (-0.33)   Depth: 6   00:00:00  10kN
21...f6 22.Nf3 Bd7 23.Rd2
  ³  (-0.35)   Depth: 7   00:00:00  16kN
21...f6 22.Nf3 Bd7 23.Rd2 Rd8
  ³  (-0.33)   Depth: 8   00:00:00  34kN
21...f6 22.Nf3 Bd7 23.Rd2 Rd8 24.Rc1
  ³  (-0.29)   Depth: 9   00:00:01  79kN
21...f6 22.Nf3 Bd7 23.Nxe5 Bxe5 24.Qg4 Kh8 25.Nc5
  =  (-0.17)   Depth: 10   00:00:03  187kN
21...f6 22.Nf3 Qf7 23.Qxf7+ Kxf7 24.Ba4 Bb7 25.Nc5 Ra7 26.Bb3
  =  (-0.11)   Depth: 11   00:00:06  399kN
21...f6 22.Nxf6+ gxf6 23.Ne4 Qf7 24.Qxh6 Qg6 25.Nxf6+ Rxf6 26.Rxf6 Qxh6 27.Rxh6 Kg7 28.Rh5
  ²  (0.57)   Depth: 12   00:00:19  1250kN
21...Qc7 22.Bg3 Qe7 23.Bh4
  =  (0.00)   Depth: 12   00:01:20  4925kN
21...Qc7 22.Bg3 Qe7 23.Bh4
  =  (0.00)   Depth: 13   00:02:15  8422kN
21...Qc7 22.Rd6 Ra7 23.Rf6 Ng6 24.Nxf7 Rxf7 25.Qxg6 Ne5 26.Qg3 Kh8 27.Rxf7 Nxf7 28.Rd4
  ²  (0.42)   Depth: 14   00:03:29  12795kN
21...Qc7 22.Nf6+ gxf6 23.Ne4 Ne7 24.Rf4 Nd7 25.Bxf6 Qxf4 26.exf4 Nxf6 27.Nxf6+ Kg7 28.Qf3
  +-  (2.20)   Depth: 15   00:05:37  20820kN
21...f6 22.Nxf6+ gxf6 23.Ne4 Qf7 24.Qxh6 Qg6 25.Nxf6+ Rxf6 26.Rxf6 Qxh6
  ²  (0.52)   Depth: 15   00:05:56  22078kN
21...f6 22.Nxf6+ Rxf6 23.Ne4 Qf7 24.Nxf6+ gxf6 25.Qxh6 Qg6 26.Qxg6+ Nxg6 27.Bxf6 a5 28.Rc1
  ±  (0.85)   Depth: 16   00:06:58  26347kN
21...f6 22.Nxf6+ Rxf6 23.Ne4 Qf7 24.Nxf6+ gxf6 25.Qxh6 Qg6 26.Qxg6+ Nxg6 27.Bxf6 a5 28.Rc1
  ±  (0.84)   Depth: 17   00:08:20  31565kN

(,  07.03.2007)

If you do not believe buggy rybka that can see wrong scores then here is glaurung's analysis:

New game - Rybka 2.2n2 mp Local,  2007
rbb2rk1/4qpp1/p1n1p2p/4n1NQ/1p2N2B/1B2P2P/PP4P1/3R1R1K b - - 0 1


Analysis by Glaurung 1.2.1:

21...hxg5 22.Nxg5
  µ  (-0.78)   Depth: 2   00:00:00
21...hxg5 22.Nxg5 Qxg5 23.Bxg5
  +-  (5.25)   Depth: 3   00:00:00
21...Ng6 22.Ba4 Nxh4 23.Bxc6
  ²  (0.35)   Depth: 3   00:00:00
21...f6 22.Nf3 Nxf3 23.Rxf3
  =  (0.18)   Depth: 3   00:00:00
21...Qa7 22.Rd2 hxg5 23.Bxg5
  ³  (-0.67)   Depth: 3   00:00:00
21...Qa7 22.Rde1 hxg5 23.Nxg5
  ³  (-0.70)   Depth: 4   00:00:00
21...Qa7 22.Bf2 Bc7 23.Nf3 Nxf3 24.Qxf3
  ³  (-0.43)   Depth: 5   00:00:00
21...Qa7 22.Bf2 Bc7 23.Nf3 Nxf3 24.Qxf3
  ³  (-0.43)   Depth: 6   00:00:00  24kN
21...Qa7 22.Bg3 a5 23.Nf3 Qxe3 24.Nxe5 Qxe4
  ³  (-0.57)   Depth: 7   00:00:01  113kN
21...Qa7 22.Bg3 Rd8 23.Bxe5 Rxd1 24.Qxd1 Bxe5 25.Rxf7 Qxe3
  ³  (-0.31)   Depth: 8   00:00:01  281kN
21...Qa7 22.Rf6 Qb7 23.Rxh6 gxh6 24.Qxh6 f5 25.Nxe6 Bxe6 26.Bxe6+ Nf7
  ²  (0.67)   Depth: 9   00:00:03  961kN
21...Qc7 22.Bg3 Qb6 23.Nf3 Ng6 24.Bxb8 Qxb8 25.Nd6 Bd7
  =  (0.04)   Depth: 9   00:00:04  1457kN
21...Qc7 22.Bg3 Qb6 23.Nf3 Ng6 24.Bxb8 Qxb8 25.Rd6 Qc7 26.Bc4
  =  (0.04)   Depth: 10   00:00:08  2548kN
21...Qc7 22.Bg3 Qe7 23.Nf3 Nxf3 24.Rxf3 Bxg3 25.Rxg3 Kh8 26.Bc4 e5
  =  (0.03)   Depth: 11   00:00:20  6768kN
21...Qc7 22.Nf6+ gxf6 23.Ne4 Ne7 24.Rc1 Qb7 25.Nxf6+ Kg7 26.Bc2 N5g6 27.Be4 Nd5 28.Nxd5 Nxh4
  ²  (0.39)   Depth: 12   00:00:49  16718kN
21...Qc7 22.Nf6+ Kh8 23.Bg3 Qe7 24.Nfh7 Kg8 25.Bxe5 Nxe5 26.Nxf8 Qxg5 27.Qxg5 hxg5 28.Rd8
  ±  (1.28)   Depth: 13   00:01:32  31333kN
21...Qc7 22.Nf6+ gxf6 23.Ne4 Ne7 24.Rc1 Qb7 25.Qxh6 Qxe4 26.Bc2 Qxc2 27.Rxc2 Nf5 28.Rxf5 exf5 29.Bxf6
  ±  (1.15)   Depth: 14   00:03:18  69973kN
21...Qc7 22.Nf6+ Kh8 23.Bg3 Qe7 24.Nfh7 Kg8 25.Bc2 Bb7 26.Bh4 f5 27.Nxf8 hxg5 28.Nxe6 Qxe6 29.Bxf5 Qxa2
  +-  (1.45)   Depth: 15   00:04:47  102688kN

(,  07.03.2007)
Parent - - By Alkelele (***) Date 2007-03-07 00:52 Edited 2007-03-07 00:55
Yes, it turned out to be 3-0.

Full game below:

[Event "?"]
[Site "?"]
[Date "2007.03.07"]
[Round "?"]
[White "New game"]
[Black "Rybka 2.2n2 mp Local"]
[Result "*"]
[Annotator "Dagh,Nielsen"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPP1PP/RNBQKBNR w KQkq - 0 1"]
[PlyCount "55"]

{1024MB, PlayChess2006.ctg, REBEKKA} 1. Nf3 d5 2. d4 Nf6 3. c4 e6 4. Nc3 Be7 5.
e3 O-O 6. Bd3 c5 7. O-O Nc6 8. dxc5 Bxc5 9. Bd2 a6 10. Qe2 Ba7 11. Rad1 Qe7 12.
Kh1 dxc4 13. Bxc4 Ng4 14. h3 Nge5 15. Bb3 b5 16. Ne4 Bb7 17. Bc3 Bb8 18. Nfg5
h6 19. Qh5 b4 20. Be1 Bc8 21. Bh4 f6 22. Nxf6+ gxf6 23. Ne4 Qf7 24. Nxf6+ Kg7
25. Qe2 Qc7 26. Bc2 Kh8 27. Qh5 Nf7 28. Ng4 *

And final position:

rbb2r1k/2q2n2/p1n1p2p/7Q/1p4NB/4P2P/PPB3P1/3R1R1K b - - 0 28


With a mate in 5 or so.

A truly incredible game. White created an attack out of nothing, and from move 17 or so until the end, everything was reversed. Black did quite well in the opening, but it was not enough.
Parent - - By Uri Blass (*****) [il] Date 2007-03-07 01:19 Edited 2007-03-07 01:22
As far as I see moves 17-20 of black could be played by rybka herself and after them it seems that black is losing the game.
It seems that rybka does not understand the position and maybe black is already worse at move 17.

some analysis
Notice that moves 17-19 suggested by rybka and move 20 is in the pv but I admit that rybka does not suggest move 20 in analysis mode with clear hash.

New game - Rybka 2.2n2 mp Local,  2007
r4rk1/bb2qppp/p1n1p3/1p2n3/4N3/1BB1PN1P/PP2Q1P1/3R1R1K b - - 0 1


Analysis by Rybka 2.3.1 32-bit :

17...Rad8
  ³  (-0.34)   Depth: 2   00:00:00
17...Rad8
  ³  (-0.31)   Depth: 3   00:00:00
17...Rad8
  =  (-0.19)   Depth: 4   00:00:00
17...Rad8 18.Rxd8
  ³  (-0.29)   Depth: 5   00:00:00
17...Rad8 18.Rxd8 Nxd8 19.Bxe5 Bxe4
  =  (-0.14)   Depth: 6   00:00:00  22kN
17...Bb8 18.Nxe5 Bxe5 19.Qh5
  ³  (-0.39)   Depth: 6   00:00:00  38kN
17...Bb8 18.Nxe5 Bxe5 19.Qh5 Bxc3 20.bxc3 Rad8
  ³  (-0.41)   Depth: 7   00:00:00  51kN
17...Bb8 18.Nxe5 Bxe5 19.Qh5 Bxc3 20.bxc3 Rad8 21.Rxd8
  ³  (-0.40)   Depth: 8   00:00:01  76kN
17...Bb8 18.Nxe5 Bxe5 19.Qh5 Bxc3 20.bxc3 Rad8 21.Rxd8 Nxd8 22.Qc5
  ³  (-0.48)   Depth: 9   00:00:01  131kN
17...Bb8 18.Nxe5 Bxe5 19.Qh5 Bxc3 20.bxc3 Rad8 21.Rxd8 Nxd8 22.Qc5 Qh4
  ³  (-0.48)   Depth: 10   00:00:03  265kN
17...Bb8 18.Nxe5 Bxe5 19.Qh5 Bxc3 20.bxc3 Rad8 21.Rxd8 Nxd8 22.Qc5 Qxc5 23.Nxc5 a5 24.Kh2
  ³  (-0.50)   Depth: 11   00:00:07  528kN
17...Bb8 18.Nd4 b4 19.Be1 Rc8 20.Qh5 Nxd4 21.Bh4 f6
  ³  (-0.59)   Depth: 12   00:00:29  1917kN
17...Bb8 18.Nd4 b4 19.Be1 Rc8 20.Qh5 Nxd4 21.Bh4 f6 22.Rxd4 Bxe4
  ³  (-0.61)   Depth: 13   00:00:34  2298kN
17...Bb8 18.Nd4 b4 19.Be1 Rc8 20.Qh5 Nxd4 21.Rxd4 Bxe4 22.Rxe4 Nd3 23.Bd2 g6 24.Qf3
  ³  (-0.67)   Depth: 14   00:00:54  3590kN
17...Bb8 18.Nd4 b4 19.Be1 Rc8 20.Qh5 Nxd4 21.Rxd4 Bxe4 22.Rxe4 Nd3 23.Bd2 g6 24.Qf3
  ³  (-0.57)   Depth: 15   00:01:26  5640kN

(,  07.03.2007)

New game - Rybka 2.2n2 mp Local,  2007
rb3rk1/1b2qppp/p1n1p3/1p2n1N1/4N3/1BB1P2P/PP2Q1P1/3R1R1K b - - 0 1


Analysis by Rybka 2.3.1 32-bit :

18...h6
  =  (-0.03)   Depth: 2   00:00:00
18...h6
  µ  (-0.81)   Depth: 3   00:00:00
18...h6
  ³  (-0.59)   Depth: 4   00:00:00
18...h6 19.Nf3
  ³  (-0.67)   Depth: 5   00:00:00
18...h6 19.Qh5 b4
  ³  (-0.55)   Depth: 6   00:00:00  16kN
18...h6 19.Qh5 b4 20.Be1
  µ  (-0.83)   Depth: 7   00:00:00  27kN
18...h6 19.Qh5 b4 20.Be1 Rd8
  µ  (-0.91)   Depth: 8   00:00:01  59kN
18...h6 19.Nf3 b4 20.Be1 Nxf3 21.Rxf3 Be5
  µ  (-0.86)   Depth: 9   00:00:01  113kN
18...h6 19.Nf3 b4 20.Be1 Nxf3 21.Rxf3 Be5 22.Bg3
  µ  (-0.86)   Depth: 10   00:00:03  146kN
18...h6 19.Nf3 b4 20.Be1 Nxf3 21.Rxf3 Be5 22.Bg3 Rac8
  µ  (-0.86)   Depth: 11   00:00:04  216kN
18...h6 19.Nf3 b4 20.Be1 Nxf3 21.Rxf3 Be5 22.Bg3 Rac8 23.Nd6
  µ  (-0.87)   Depth: 12   00:00:13  852kN
18...h6 19.Qh5 Bc7 20.Nf3 Rad8 21.Rxd8
  ³  (-0.37)   Depth: 13   00:00:49  3119kN
18...h6 19.Qh5 b4 20.Be1 Nd8 21.Bg3 Ng6 22.Rd6 Bxe4 23.Nxe4 Qa7 24.Qf3
  =  (-0.23)   Depth: 14   00:02:30  9703kN

(,  07.03.2007)

New game - Rybka 2.2n2 mp Local,  2007
rb3rk1/1b2qpp1/p1n1p2p/1p2n1NQ/4N3/1BB1P2P/PP4P1/3R1R1K b - - 0 1


Analysis by Rybka 2.3.1 32-bit :

19...b4
  ³  (-0.65)   Depth: 2   00:00:00
19...b4
  ³  (-0.65)   Depth: 3   00:00:00
19...b4
  ³  (-0.55)   Depth: 4   00:00:00
19...b4 20.Be1
  µ  (-0.83)   Depth: 5   00:00:00
19...b4 20.Be1 Rd8
  µ  (-0.83)   Depth: 6   00:00:00  43kN
19...b4 20.Be1 Rd8 21.Rxd8+ Nxd8 22.Nf3
  µ  (-0.99)   Depth: 7   00:00:01  99kN
19...b4 20.Be1 Bc7 21.Nf3 Nxf3
  µ  (-0.75)   Depth: 8   00:00:02  145kN
19...b4 20.Be1 Bc7 21.Nf3 Nxf3 22.gxf3 Ne5
  ³  (-0.60)   Depth: 9   00:00:03  203kN
19...b4 20.Be1 Bc7 21.Nf3 Nxf3 22.gxf3 Ne5 23.Rd4
  ³  (-0.56)   Depth: 10   00:00:04  289kN
19...b4 20.Be1 Nd8 21.Bg3 Ng6 22.Rd6 Bxe4 23.Nxe4 Nb7
  ³  (-0.42)   Depth: 11   00:00:20  1265kN
19...b4 20.Be1 Nd8 21.Rd4 Nec6 22.Rd1 Be5 23.Bh4 f6
  ³  (-0.39)   Depth: 12   00:00:26  1632kN
19...b4 20.Be1 Na5 21.Rd4 Nxb3 22.Bxb4 Qe8 23.axb3 f5 24.Qd1 hxg5 25.Bxf8
  ³  (-0.35)   Depth: 13   00:00:51  3259kN
19...b4 20.Be1 Bc8 21.Nf3 a5 22.Bg3 f6 23.Nd4 Nxd4 24.exd4 Qf7 25.Qxf7+ Nxf7 26.Bxb8
  =  (-0.21)   Depth: 14   00:02:09  8367kN

(,  07.03.2007)
Uri
Parent - By Alkelele (***) Date 2007-03-07 11:09 Edited 2007-03-07 11:38
Hi Uri, it's true that Rybka could maybe also have been behind the black pieces and have suffered the same fate (speaking of moves 17-20 or so).

The reason I think this game is incredible is that there was a human behind the black pieces. He had a "safe" advantage out of the opening. He had many years of experience with this kind of structure (with the small difference that the f2-pawn was missing). He seemed to be very careful about his play in this transition into the middlegame. And yet, WHAM!, suddenly his opponent finds a surprising attack and finishes him off in 10 moves. I don't know if you followed the game live -- I think that to many of the spectators, this was an extremely surprising and exciting turn of events in this game.

I have not analysed the game deeply (it will take a lot of CPU time since there are so many different lines to check and refute), but I suspect that the game became untenable for the human when he embarked on the Nf6-g4-e5 maneuvre. It may not be objectively lost yet, but it happens that white is allowed a complicated attack that it would require extreme precision to fend off.

The point is -- it is not so relevant that Rybka was behind the white pieces. Maybe Toga would have played the attack in the same manner. The interesting aspect is that this was a spectacular piece of evidence about what can happen in a computer-human game, even when the human is trying hard to play it safe and is up a pawn.
Parent - - By h1a8 (***) [us] Date 2007-03-07 01:55
Pawn odds with first to move isn't really a pure disadvantage. It is exactly like a sound gambit.
This is because the side with the missing pawn has first move (a tempi ahead) and has open lines
for better mobility and faster development (creating an even extra tempi).
Parent - - By Jim Walker (***) [us] Date 2007-03-07 03:55
I agree.  Rybka didn't even know the pawn was missing in the first 3 games.  If this is done again with a more fair setup I hope GM Ehlvest is given the right of first refusal.
Parent - - By Quapsel (****) [de] Date 2007-03-07 09:43
Fair Setup?

Maybe, that the advantage for human was not big enough.
OK, perhaps, this is to learn at this point.
Or maybe, Jaan doesn't have the right general plan, to do the matches.
Or a stronger human Player should be choosen.

Perhaps one will do it again, having thougt about this, an some things will be changed.
But I dont think, that there ist someting unfair at this point.

Quap
Parent - - By Jim Walker (***) [us] Date 2007-03-07 12:46
What I mean by a fair setup is to give the human equal chances to win.  That means different odds for different humans.  For instance I would need at least Queen odds for any chance vs Rybka.  Considering Jaan Ehlvest is only 2600 GM he needs more advantage vs Rybka on the hardware used.  I know this is an experiment and more data is needed to find out what is fair for who but it seems obvious that Rybka didn't even notice it's pawn was missing.  A human who is more "practiced" at playing computers might do a lot better even if his rating is lower.  This experiment is interesting only if terms are near equal.
Parent - - By Michael Waesch [de] Date 2007-03-07 12:52
Come on. Even I, Patzer cubed, can defeat Rybka at Knight odds now and then.

Mike
Parent - - By Jim Walker (***) [us] Date 2007-03-07 14:06
What's your point?
Parent - - By Michael Waesch [de] Date 2007-03-07 14:26
You surely don´t need queen odds to defeat Rybka and GMs don´t need Knight odds.

Mike
Parent - - By Quapsel (****) [de] Date 2007-03-15 08:20
Ist this a shure cognition?
(I don't know enough about chess to say this. In german we say: "Asche auf mein Haupt!")

What do you think:
A human player with which ELO-strength would have equal play againt Rybka with one knight less?

Benno
Parent - - By Michael Waesch [de] Date 2007-03-15 13:32
Well, I am a Patzer about 1400. I usually lose 3 games, have one draw and one win out of 5 games.

People with 1800 or more reported that they have no difficulties at all beating Rybka with Knight odds.

Mike
Parent - - By grolich (***) Date 2007-03-15 14:53
It actually depends on the computer itself, the exact Rybka version, Rybka's settings, and the time control.

On the older versions of Rybka it was easier to win with knight odds than it is today.

If the computer is relatively slow or old (or both...) Rybka's ELO without a piece drops a lot more rapidly with the computer speed than it does with equal material on the same compared machines.
Not sure why, perhaps Rybka needs a certain minimum depth in order to exploit common mistakes to the fullest, when it is playing a piece down, and in contrast to an equal material game, not exploiting them to the fullest here just leads to a more stable position (which is lost...).

The time control is also important: in 5 minute games on strong computers, I've seen candidate masters lose to it (and Larry Kauffman mentioned a master friend of his who encountered the same problem).
Parent - By Michael Waesch [de] Date 2007-03-15 17:00
Well, I was talking about 2.3.1 MP on a Athlon 4400+ x2.

Mike
Parent - By grolich (***) Date 2007-03-15 15:04
if the question is at what level he can be better than Rybka, that is, win a long MATCH with Rybka at knight odds, not just the odd game here and there, than probably a pretty strong player is needed... of about master level strength (ELO ~2300).

Just to win some games (even a lot, as long as it's not most), is possible at a lower level.
Parent - - By grolich (***) Date 2007-03-15 17:31
What's the contempt value Rybka is set for in these matches?
It should be REALLY high with a piece.

The problem is that with an extra piece, whenever Rybka sees in some variation a strong (actually, even just a not so strong) initiative (or attack, or some other strong play) for the side with the extra piece it prefers to go for the simpler route (including, but not limited to, exchanges),  but that only helps the other side to simplify to reach a more easily won position.

It also causes more draws as Rybka tends to force a draw the first opportunity it can find under these circumstances.

The contempt value must be used for handicap games. Otherwise you make the program's logic work against her (I'm worse, so a value close to 0.00 is better... even if I'm switching a -1.50 complicated position for a -1.00 ending which can be a lot simpler to win for the other player, and maybe even without any risk of losing...).

If you just play it without raising the contempt value, or with a too low (or too high...) setting on the contempt value,
Play it with a contempt value at about 100 and report the results again... They should be quite different.
Parent - - By Michael Waesch [de] Date 2007-03-15 19:43
Sounds interesting and of course I didn´t change the contempt. I just forgot it and I don´t take these Knight odd games too seriously. I just didn´t believe that Knight odds can be given to a grandmaster which it was suggested.

Mike
Parent - By grolich (***) Date 2007-03-15 19:57
Of course not to a GM.  Well, not for the next few years for sure anyway.
Don't know if ever, But there are arguments to the contrary as well.

Also, Larry Kauffman used a limited version of Rybka, which played at a level of 2400 ELO when it played against other engines.
He then found out that the Deeper Rybka can play against the 2400 Rybka at knight odds.

As always, time will tell.
Parent - - By lkaufman (*****) Date 2007-03-16 03:39
     One new bit of data on this. Yesterday I had a visitor who is an internet blitz specialist, rated about 2700 on ICC blitz (mainly 3 minute chess). Now that's not like FIDE 2700, but it is roughly the level of a fairly strong IM or a relatively weak GM on ICC. Anyway he played two three minute games with Rybka 2.3.1 (Ehlvest version, with anti-human heuristic but no other important difference, since the other differences were specific to pawn odds) on my Quad at knight odds. In both games he kept a lead for quite a while, but eventually lost material and had to resign. Of course three minute chess is faster than normal blitz, but this does suggest that on a Quad, Rybka may well be able to give knight odds successfully to weaker grandmasters in three minute chess.
Parent - By Michael Waesch [de] Date 2007-03-16 14:11
I am total crap in blitz. I like to play Rybk wit Kight odds at least with 5+5 time settings.

Mike
Parent - - By grolich (***) Date 2007-03-15 15:00
I don't know what your level is, but I believe, based on results I have seen with rook odds games, even 1200 level players can beat Rybka at queen odds (probably 1100 too, but this gets to the level where people don't have the game sorted out in their heads enough to calculate any variation properly, so I dare not make predictions).
Parent - - By Uri Blass (*****) [il] Date 2007-03-16 21:21
I think that in order to know what 1200 means it is better to give some pgn of players with that rating.

I give some normal correspondence games that I played with players with rating 1200-1300 in www.gameknot.com

I believe that rybka without a queen can in most cases beat players with rating of 1200-1300 from that site.
Players who play their first game in gameknot.com get rating of 1200 and part of them can lose games simply by 1.e4 2.Bc4 3.Qf3 4.Qxf7 checkmate

superman123 often beat beginners by this tactics(he likes to win so he always choose players with rating near 1200
and when I look at his game I often find short games like 1.e4 e5 2.Bc4 Nc6 3.Qf3 h5 4.Qxf7 mate)

When he finds strong players(for example when he played against me) he plays slowly and probably hopes that the opponent is going to lose on time(It did not help him against me and at some point he was too slow and lost on time by himself but there are cases when this strategy helps against other players).

It seems that this strategy is very efficient in achieving high rating in this site because he got rating of 1864 when I know that he probably deserves only 1400-1500 based on his playing strength.

I give also my game against superman123 from that site.

You can watch games that I play against madad(rating 1219) in the following links:

http://gameknot.com/chess.pl?bd=6913024&r=110
http://gameknot.com/chess.pl?bd=6913031&r=653

I guess that I can beat madad with queen odd.


[Event "Let's play chess"]
[Site "http://gameknot.com/chess.pl?bd=3945882"]
[Date "2005.10.20"]
[White "edcba"]
[Black "coloughl"]
[Result "1-0"]
[WhiteElo "1200"]
[BlackElo "1211"]
[TimeControl "1/259200"]
[Mode "ICS"]
[Termination "normal"]

1. d4 d5 2. c4 e6 3. Nc3 Nf6 4. Bg5 Nc6 5. cxd5 exd5
6. Bxf6 Qxf6 7. Nxd5 Qd6 8. e4 g6 9. Nf3 Bg7 10. Bc4 Be6
11. O-O O-O-O 12. Rc1 Bxd5 13. Bxd5 Nb4 14. Bxf7 Bxd4 15. Nxd4 Qxd4
16. Qg4+ Rd7 17. Be6 Nd3 18. Bxd7+ Qxd7 19. Qxd7+ Kxd7 20. Rfd1 Ke6
21. Rxd3 b5 22. Rxc7 a5 23. Rd5 Rb8 24. f4 a4 25. Rdc5 Kd6
26. Rc5c6# 1-0

[Event "Let's play chess"]
[Site "http://gameknot.com/chess.pl?bd=3945916"]
[Date "2005.10.21"]
[White "edcba"]
[Black "timbo12"]
[Result "1-0"]
[WhiteElo "1200"]
[BlackElo "1219"]
[TimeControl "1/259200"]
[Mode "ICS"]
[Termination "normal"]

1. e4 e5 2. Nf3 Nc6 3. Bb5 a6 4. Bxc6 dxc6 5. O-O Bg4
6. h3 Bh5 7. g4 Bg6 8. Nxe5 Qh4 9. Qf3 Bd6 10. d4 f6
11. Nc4 O-O-O 12. Nxd6+ cxd6 13. Nc3 h5 14. Kg2 hxg4 15. Qxg4+ Kc7
16. Qxh4 Rxh4 17. f3 Rf8 18. Be3 f5 19. Bf2 Rh5 20. h4 fxe4
21. fxe4 Rf4 22. Rae1 b5 23. Bg3 Rg4 24. Kh3 Nh6 25. Rf8 Kb6
26. b4 a5 27. a3 c5 28. dxc5+ Kc6 29. cxd6 axb4 30. axb4 Nf7
31. Kxg4 Ne5+ 32. Bxe5 Rxe5 33. Rd1 Bh5+ 34. Kf4 Re6 35. d7 Rd6
36. Rxd6+ Kxd6 37. d8=Q+ Kc6 38. Nd5 g6 39. Qc7# 1-0

[Event "Let's play chess"]
[Site "http://gameknot.com/chess.pl?bd=3960426"]
[Date "2005.10.25"]
[White "galaxy2"]
[Black "edcba"]
[Result "0-1"]
[WhiteElo "1298"]
[BlackElo "1200"]
[TimeControl "1/259200"]
[Mode "ICS"]
[Termination "normal"]

1. e4 e5 2. Bc4 Nf6 3. Qf3 Nc6 4. c3 Be7 5. d3 d6
6. Bg5 Be6 7. Bxf6 Bxf6 8. Bxe6 fxe6 9. g3 O-O 10. h4 Qd7
11. Qg4 Rf7 12. Nd2 Raf8 13. f3 Ne7 14. Ne2 Nc8 15. O-O Qb5
16. b3 Qxd3 17. Rf2 Qxd2 18. Nd4 Qxc3 19. Nc2 Re8 20. f4 exf4
21. Qxf4 Ref8 22. Qe3 Qxe3 23. Nxe3 Bxa1 24. Rf1 Rf7xf1+ 25. Nxf1 Bd4+
26. Kg2 Rf2+ 27. Kg1 Rxa2+ 28. Kh1 Rf2 29. Kg1 Rf3+ 30. Kg2 Rxb3
31. Kh3 Rb2 32. Nh2 Rxh2+ 33. Kxh2 a5 0-1

[Event "Challenge from superman123"]
[Site "http://gameknot.com/chess.pl?bd=3942987"]
[Date "2005.12.15"]
[White "superman123"]
[Black "edcba"]
[Result "0-1"]
[WhiteElo "1641"]
[BlackElo "1200"]
[TimeControl "1/259200"]
[Mode "ICS"]
[Termination "time forfeit"]

1. e4 e5 2. Bc4 Nf6 3. Qf3 Nc6 4. c3 Bc5 5. d3 O-O
6. Ne2 a6 7. O-O d5 8. exd5 Bg4 9. Qg3 Bxe2 10. Re1 Bxd3
11. Qxd3 e4 12. Qe2 Ne5 13. Rd1 Nfg4 14. Rf1 Qf6 15. h3 Nxf2
16. Be3 Bxe3 17. Qxe3 Nxh3+ 18. Qxh3 Qb6+ 19. Rf2 Nxc4 20. b3 Ne5
21. Kf1 Rad8 22. c4 Nd3 23. Rd2 Qf6+ 24. Kg1 Qxa1 25. Rd1 Qxa2
26. Nc3 Qc2 27. Qe3 Qxc3 28. Qxe4 Nc5 0-1
Parent - By grolich (***) Date 2007-03-16 22:36
Ok i'll put it another way: don't know about the ratings on that site.
Since ELO is a relative system, I'll be more specific:

I've had a 1350 (Israel ratings) player play against Rybka at queen odds... and he won without any trouble. Which made me think even 1300 and somewhat below that will succeed (with some problems, perhaps).
I'd love to repeat the experiment with weaker players... But my club currently has either stronger players, or a couple which have just started to learn the rules...

The level below which you simply don't know HOW to win (or play... for that matter) differs from place to place, country to country, and also, from site to site (I've also noticed that within the SAME system, the ratings near the absolute beginner level constitute a bit of a grey area... The standard expected performance that ELO suggests doesn't seem to apply near those ratings).
Parent - - By Uri Blass (*****) [il] Date 2007-03-07 10:03
In this case I think it is going to be fair to give the human to choose the side to play
I do not think that humans will play better with pawn disadavantage.

What may be correct is that humans are relatively weaker in this type of positions because there is no theory that they can use when in normal chess they may get into a drawn endgame thanks to their theoretical knowledge.

I expect rybka to perform better in shuffle chess relative to normal chess against humans for tnat reason.
Uri
Parent - - By DamirD81 (***) [dk] Date 2007-03-16 14:55
In my opinion the best human player against computers is Eduard Nemeth who has played plenty of games and won.
You can see some of his games on chessgames. They are simply masterpieces with a win percentage of 95%.
Even though the engines he played against at that time were much weaker than they are now, still a win is a win.
I believe he can play without a pawn odds against any engine out there and win.
Parent - - By Uri Blass (*****) [il] Date 2007-03-16 16:26
I believe that you are wrong.

Nemeth posted part of his games.
He posted the games that he won but did not post the games that he lost.

Uri
Parent - By Banned for Life (Gold) Date 2007-03-16 19:28
I've never seem him claim that he wins all of his games. He posts the newsworthy ones, i.e. the ones he wins. I suspect if anyone asks him, he would post the complete list.
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka - Ehlvest (game 3)

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill