< Earlier Kibitzing · PAGE 38 OF 79 ·
|Nov-08-06|| ||notyetagm: Rybka eval, please, for the following position from the game Leko vs Grischuk, 2006:|
Black to play: 55 ... ?
click for larger view
Here Black (Grischuk) actually played 55 ... e5 and the game ended in a draw. But Black could win a pawn with 55 ... xb3 or create four monstrous connected passed pawns after 55 ... xh2+ 56 xh2 xg3.
Which of these continuations does Rybka like best, 55 ... e5, 55 ... xb3, or 55 ... xh2+ ? Thanks.
|Nov-08-06|| ||notyetagm: Or, even simpler, does Rybka think that Black is winning in the variation 55 ... xh2+ <remove the guard> 56 xh2 xg3, shown below:|
click for larger view
|Nov-09-06|| ||you vs yourself: From Anand vs Topalov, 2005 |
Had Anand played 50.Rh3+, this would've been the position:
click for larger view
Chessmaster 10 gives it a +0.79. What's rybka's eval for the above position?
|Nov-09-06|| ||jhoro: free Rybka and Toga, both think it is a dead draw:
Rybka 1.0 Beta
0.00 50... Kg8 51.Rxh2 e4 52.Rch1 Qb6+ 53.Kc1 Qe3+ 54.Kb2 Qb6+ 55.Ka1
Toga II 1.2 Beta2a
0.00 50. ... Kg8 51.Rxh2 e4 52.Rch1 Qb6+ 53.Kc1 Qe3+ 54.Kd1 Qf3+ 55.Ke1 Qe3+ 56.Kf1 Qc1+ 57.Kg2 Qg5+ 58.Kf1 Qc1+
|Nov-09-06|| ||you vs yourself: <jhoro> Thanks!|
|Nov-10-06|| ||notyetagm: From www.rybkachess.com
Rybka in Dutch Open Computer Chess Championship 2006
By Jeroen Noomen
Round 1: Rybka - Loop 1-0
A very interesting game, from a theoretical point of view. The line with 13... Ne8 is hot in computerchess and it leads to very complicated positions. Some lines are a real minefield, so I had to come up with an idea to play it safe as white. Black constantly has mating power available, with queen and rook on the a-file, the knight that can hop to b5 via c7 and not to forget the thrust d6-d5! I think with 22.Qg2! and 24.Rd5! white takes the sting out of the attack and emerges with a slight edge. With the black rook gone from the a-file and d6-d5 stopped forever, white need not worry about an attack anymore. Loop felt unhappy and rushed to exchange queens, but that meant white had a beautiful passed pawn on a3, supported by a strong bishop. Rybka played this ending strongly and gave Loop no chance. A clean win against one of our main rivals, a good way to start the tournament.
Round 2: Shredder - Rybka 0-1
Needless to say that after Turin I was very eager to win this game. Rybka played a solid Sicilian and had no problems after book was over. Shredder sacrificed a pawn to keep the black king in the centre, showing a +0,50 score. But this was really a misevaluation. Rybka already felt better as black and when Shredder started to sideline pieces (Qe3-h6-h7 for instance), it was all over quickly. Rybka gave a mate in 20 announcement at move 46. Sweet revenge (sorry, Ernst!).
Round 3: Rybka - Hiarcs 1-0
Another Najdorf with Hiarcs deviating from the Loop game. But IMO 10... Qc7 and 11... Rc8 cannot trouble white. With black's white squared bishop gone, the black minors passive at e7 and h5 and a firm white grip on the square d5, black can only wait and see what white will do. Rybka took some time to decide on action, but the pretty 30.Na1! - heading for d5 - increased the white advantage. Still, Hiarcs put up a tough fight and the ending needed precise play. We really wondered why Hiarcs gave away its g-pawn so easily at move 62, after that the game was over.
Round 4: Deep Gandalf - Rybka 0-1
The fourth Sicilian in a row, and no problems for black after leaving book. 10... Qa7! prevents a bishop trade by 11.Be3 and equalises. Suddenly Gandalf put oil on the fire with the questionable thrust 11.g4?! Black lost her castling rights, but white was in trouble right away. At move 17 white surely had to exchange twice at b7, hoping to escape in a worse endgame, but 17.Kd1? made matters worse. The rest of the game was a real Rybka demonstration, leading to the shortest win in a mere 28 moves. Anticipating 29.Bc3 Rybka announced mate in 15, so Gandalf's operator resigned.
Round 5: Rybka - Fruit 1-0
Another topical line, the Anti-Meran with 7.g4. This leads to lively play. IMO 10... cxd5 is questionable, while 15... exd4? is probably already the losing mistake. 15... Kh7 was necessary, but white is better anyway. Rybka played the resulting ending flawlessly and cashed in another win after performing some nice rook manoeuvres.
|Nov-10-06|| ||notyetagm: Round 6: The Baron - Rybka 0-1
A quiet opening line suddenly led to a closed position, when both programs decided to swap a lot of pieces but no pawns. White's 24.g4?! meant that only black had chances to go for a win, connected with the plan f7-f5. Rybka decided to wait and see and I was already making myself up for a very long game. Then The Baron helped us out by playing 55.Bxg6? that immediately led to a lost position. The Baron was already in mild time trouble and couldn't evaluate properly what was wrong with its decision. A tough game.
Round 7: Rybka - Isichess 1-0
In the opening - a Ruy Lopez - Rybka decided to sacrifice a pawn for positional compensation. Isichess had to play very carefully to avoid being entangled, but failed to do so. 24... Nh5? was a clear mistake and 28... f5? was something the black position couldn't take anymore. Still, Rybka finished the game in nice style, with an excellent exchange sac 33.Rc6!
Round 8: Rybka - Deep Sjeng 1-0
A Petroff was played 15 moves out of book and after 15.Bg5 black already has slight problems. Sjeng decided to get rid of the pressure by giving away a pawn, but then the game was essentially already over. With 57.g5! and 58.f6! - boxing in the black bishop - the finish was cute. Funilly enough Rybka operator Hans van der Zijden saw it already coming and when we decided that it was winning, Rybka agreed and jumped to +9.
Round 9: The King - Rybka 0-1
The King played the English as white. Who could have thought that this quiet opening would end in such a tactical fight! 12... Bh3 was played from the book (through a transposition of moves) and in practice white never took the bishop, as the black queen is very strong on h3. The King was happy to swap anyway and proceeded with 14.Nd5, asking Rybka 'prove your point'. This duly came: 14... Ng4! leads to a strong attack and when The King took the bait on c7, the game was already over. 15... f5! and the manoeuvre Nd5-c7-e6-g5 were pretty moves, underlining why white shouldn't capture on h3. I haven't seen The King losing so quickly very often, in the earlier days it was The King that was crushing the opponents like this! At move 24 we had yet another stunning mate announcement, this time in 18 moves.
So Rybka scored an amazing tournament victory, winning al her games. In a sense it was the perfect tournament: a great engine, a superb machine (a big thank you to Joe Soney is in place here!) and the openingbook worked fine. Normally you always have at least one bad game in a tournament, but here all went smoothly. I want to thank:
1. The organisers for yet another well organised and super relaxed tournament in Leiden
2. Joe Soney for making it possible to play on his machine (yes, it is a monster!)
3. Theo van der Storm for helping me out with the internet connection
4. All the participants, it was nice to meet you and the atmosphere was absolutely relaxed and fine
Second place went to Loop, that played an excellent tournament. Third came Hiarcs, a deserved and fine performance. Fruit was fourth and - like Loop and Hiarcs - only lost to Rybka. It is good to see Fabien Letouzey returning to the scene!
See you next time!
brief game comments by Vasik Rajlich
all games from tournament
Rybka games (with Rybka search info)
|Nov-10-06|| ||aw1988: Yes, impressive, although Fritz and Junior were not present...|
|Nov-10-06|| ||chessmoron: Because they are scared. They know that Rybka will clobber them up unless you say Hydra wasn't present, then I'll shut up for a moment.|
|Nov-11-06|| ||notyetagm: <aw1988: Yes, impressive, although Fritz and Junior were not present...>|
Did you see Rybka declare mate in 20(!) in one game and mate in 18(!) in another? And you wonder why Fritz and Junior were no shows.
|Nov-11-06|| ||THE pawn: It's not that hard for a computer to declare mates in 20+ moves...|
|Nov-11-06|| ||notyetagm: <THE pawn: It's not that hard for a computer to declare mates in 20+ moves...>|
Yes, I understand that, but the point was that Rybka got into -positions- against these other top computer programs where the mate in xx was possible.
Go play over the crushing Rybka win over The King again to see what I mean:
[Event "26th DOCC"]
[Site "Leiden, NED"]
[White "The King"]
1. c4 Nf6 2. Nc3 e5 3. Nf3 Nc6 4. g3 Bc5 5. Bg2 d6 6. O-O O-O 7. d3 h6 8. a3 a6 9. b4 Ba7 10. e3 Be6 11. Bb2 Qd7 12. Qc2 Bh3 13. Bxh3 Qxh3 14. Nd5 Ng4 15. Nxc7 f5 16. Nd5 Rf7 17. b5 Ne7 18. b6 Nxd5 19. bxa7 Nc7 20. Rfd1 Ne6 21. Rd2 f4 22. exf4 Ng5 23. Qd1 exf4 24. Bd4 fxg3 25. fxg3 Nxf3+ 26. Qxf3 Rxf3 27. Rg2 Raf8 28. Kh1 Rxg3 29. Rag1 Qh4 30. a8=Q Rxa8 31. Re2 Rxd3 0-1
Position after 24 Bd4 (Black mates in 18!):
click for larger view
|Nov-11-06|| ||aw1988: That's due to Rybka's superior play in general... I doubt Rybka played 2...e5 and thought okay, gotta get into a position with a mate in 18.|
Full congratulations to the developers and such and the usual praise that comes out of this situation but this mate argument is flawed.
|Nov-11-06|| ||whatthefat: Amazing performance by Rybka! By the way, does anyone know how Rybka fares compared to other engines in Fischerandom? I ask because it might give some indication of how big a role opening books play in computer chess.|
|Nov-12-06|| ||Mameluk: Which strong chess software is free available? Fruit/Toga, Spike, first Rybka version... any else?|
|Nov-12-06|| ||WillC21: <notyetagm><Did you see Rybka declare mate in 20(!) in one game and mate in 18(!) in another? And you wonder why Fritz and Junior were no shows.>|
Yes, it has becoming extremely clear that Rybka has one of the most robust(and efficient) pruning routines in the business. All this talk about Rybka's positional strength is being used by Rajlich as smokescreen, IMO; I bet, positionally, Rybka is only marginally different from Shredder 10 and the newest version of Fruit.
Just observe Rybka's low node count and very deep search. Go deep enough, and bam! you know what happens? Those extra tactics(pruning extensions) make the engine "appear" better positionally in most instances. Although, granted, to know for sure I would need Rybka's source code and about 5 hours of free time, neither of which is likely to happen :)
|Nov-12-06|| ||WillC21: And, yes, I am aware Rajlich is an International Master, but I still think the positional strength is not much different from Shredder/Fruit.|
|Nov-13-06|| ||notyetagm: <whatthefat: Amazing performance by Rybka! By the way, does anyone know how Rybka fares compared to other engines in Fischerandom? I ask because it might give some indication of how big a role opening books play in computer chess.>|
Try posting your question over at http://www.rybkaforum.com.
|Nov-13-06|| ||whatthefat: <notyetagm>
Will do, cheers.
|Nov-18-06|| ||whatthefat: Has anyone tried setting Rybka on this?
|Nov-19-06|| ||classF: <whatthefat> Rybka 1.0 beta 32-bit finds the correct first several moves within one second, but does not understand that black can underpromote to knight on the seventh move.|
|Nov-19-06|| ||whatthefat: Thanks, would you be able to post Rybka's analysis with increasing ply? At what point does it see that white's winning?|
|Nov-19-06|| ||RandomVisitor: <whatthefat>I do not believe that Rybka can play Fischer random. It does not understand the unique castling rules.|
|Nov-19-06|| ||classF: <whatthefat> as chessbase explains, there is no win, it's a draw. The puzzle is busted by 4...Kg4|
click for larger view
Rybka v1.0 Beta.w32 :
3 00:00 128 7.710 0.00 Ng4f6+
4 00:00 228 13.733 0.00 Ng4f6+ Kh7g7
5 00:00 477 14.801 0.00 Ng4f6+ Kh7g7 Nf6h5+ Kg7g6
6 00:00 5.725 74.207 -3.71 Ng4f6+ Kh7g7 Nf6h5+ Kg7g6 Nh5f4+ Kg6f5
6 00:00 11.038 101.828 -3.07 d7d8Q Ng5f7+ Kd6e7 Nf7xd8
7 00:00 23.026 124.754 -3.59 d7d8Q Ng5f7+ Kd6e7 Nf7xd8 Ke7xd8 Bb4a5+ Kd8d7
8 00:00 39.638 123.371 -3.88 d7d8Q Ng5f7+ Kd6e7 Nf7xd8 Ke7xd8 Bb4a5+ Kd8d7 Ba5c7
8 00:00 52.224 126.424 -3.77 Ng4f6+ Kh7g7 Nf6h5+ Kg7g6 Nh5f4+ Kg6f5 Bd1c2+ Kf5g4 Bc2d1+ Kg4xg3
9 00:00 111.262 107.079 +0.43 Ng4f6+ Kh7g7 Nf6h5+ Kg7g6 Bd1c2+ Kg6xh5 d7d8Q Kh5g4 Qd8e8
10 00:01 125.803 102.975 +0.44 Ng4f6+ Kh7g7 Nf6h5+ Kg7g6 Bd1c2+ Kg6xh5 d7d8Q Kh5g4 Qd8e8 c5c4+
11 00:01 187.308 88.880 +1.91 Ng4f6+ Kh7g7 Nf6h5+ Kg7g6 Bd1c2+ Kg6xh5 d7d8Q Kh5g4 Kd6c6 Kg4xg3
12 00:03 278.344 88.079 +2.43 Ng4f6+ Kh7g7 Nf6h5+ Kg7g6 Bd1c2+ Kg6xh5 d7d8Q Kh5g4 Kd6c6 Kg4xg3
13 00:05 456.741 87.241 +1.62 Ng4f6+ Kh7g7 Nf6h5+ Kg7g6 Bd1c2+ Kg6xh5 d7d8Q Kh5g4 Kd6c6 Ng5f3
14 00:09 824.811 89.037 +2.43 Ng4f6+ Kh7g7 Nf6h5+ Kg7g6 Bd1c2+ Kg6xh5 d7d8Q Kh5g4 Qd8e8 c5c4+
15 00:34 3.033.374 91.920 +4.16 Ng4f6+ Kh7g7 Nf6h5+ Kg7g6 Bd1c2+ Kg6xh5 d7d8Q Kh5g4 Kd6c6 Ng5f3
16 01:03 5.523.293 90.545 +4.16 Ng4f6+ Kh7g7 Nf6h5+ Kg7g6 Bd1c2+ Kg6xh5 d7d8Q Kh5g4 Kd6c6 Kg4xg3
17 02:04 10.510.080 86.860 +4.16 Ng4f6+ Kh7g7 Nf6h5+ Kg7g6 Bd1c2+ Kg6xh5 d7d8Q Kh5g4 Kd6c6 Kg4xg3
|Nov-19-06|| ||whatthefat: Cheers. So Rybka found the critical line at 9 ply (including 4...Kg4), but increasingly assesses the position as winning for white from there?|
< Earlier Kibitzing · PAGE 38 OF 79 ·