![[cr]](/mwf/v12/flags/cr.png "Costa Rica")
Hi,
I have seen that Rybka is having some problems with these positions. It would be good, at least to take a look.
In this position, white has a winning sequence. White needs to push his pawns and break the black pawn structure. In my opinion, this secuence should be not so difficult to see for a very strong program like Rybka.
Rybka is thinking this line for a long time: 1. d5! exd5 2. exd5 c5 3. a5 bxa5 4. bxa5?? The first 3 moves are good and strong enough, but 4. bxa5 is a mistake. After this, it is simply a draw.
The engine is just missing for a good time the strongest 4. a5! and after bxa5 then b5! winning.
After b5, bxa5 cxb5 and Ke7, and now white have a passed pawn and black King needs to chase the pawn. Of course, the black king finally hunts it, but then white break the black chain pawn again in the Kingside with g5! and after hxg5 then, f5 gxf5 and h5 and again white achieves to enters with the h pawn and assure the victory easily.
Equally, if black plays fxg5 rather than hxg5, then h5 gxh5 and f5, and now the f pawn enter. White is winning the game without no problems.
Rybka cannot find the winning sequence from the initial position, (Although, I do not know if with a lot of time, it can) so we can go some moves ahead.
Here, the position change a bit, because Black does not plays c5 (Rybka move). It plays cxd5, but always white has a strong move to break the black pawn structure with their pawns. Rybka again has some troubles to find the winning sequence. The engine thinks for some minutes that cxd5 is the best white move here. But after cxd5, white cannot enter with a pawn in the Queenside, and the Black King it is not more in danger because if White tries to break now in the kingside it is now useless with the Black King so near of them.
It should not be so difficult for Rybka to see how to break the pawns and to pass with some of them, in this case first in the queenside and later in the other side of the board, the kingside. The engine is discarding a5 for some time and in my opinion it should understand it a bit more faster.
The winning line is: 3. a5! bxa5 4. b5 axb5 5. cxb5 Ke7 6. b6 Kd7 7. b7 Kc7, (The black king hunts the white pawn but then), 8. g5 and 1-0, because here black cannot stop the white pawn in the Kingside. It should not be so hard for Rybka
I think that this is the worthiest position to rescue.
In this position, black plays hxg3 but if white plays the very normal hxg3, then white cannot progress anymore. It is a draw. After white plays hxg3, black respond with g4 and white needs to take the g4 pawn with his f pawn. If not, the f white pawn falls.
After g4 and fxg4, it does not care that here White have 2 pawns more in the Kingside. It is as if it would have only one pawn and white cannot progress. The only move to progress is to take the black pawn wit his King. I believe that if a program take the pawn with his King, it is really understanding and knows that it has a win. So after Kxg3 white can win easily and their plan is more simplest.
After Kxg3 Kh7 (Kf7 or Kg6 is a mistake, because the black Rook goes to h8 or give a check and then c8=Q winning)
Rc3 by black is also useless after Kg4, Rc5 and h4! and the f3 white pawn now is free to advance. The idea is to attack the King with the f pawn and to move it from g7 and h7. The black Rook cannot move from the C column because then the White Rook is free to move and with that also the pawn in c7. Rybka is discarding completely Kxg3. But again, after Kxg3 Kh7 and h4! a very strong move. It should not be a big problem for Rybka to see. After gxh4 Kh3 Kg7 and f4 again, the f pawn will walk and somebody have to stop it :-), and with that, white wins easily. In my opinion Rybka needs to understand this easily. Do not you believe?
Ok this is the same topic of the wrong bishop endgame. And of course 2.3.2a does not know how to save the game. I really hope that for Rybka 3 it can understand this again.
2.3.2a wants to play here 1. Ne7, saving his Knight, that seems the most natural move, but after Bf4+ 2. Kb7 Kf6 3. Kc6 Ke5! 4. Kc5 Bg5 5. Nd5 and h5, black is winning easily. Rybka should understand and let the knight to die and save the game with Kc7!
This endgame appears from time to time and it is important that Rybka knows it.
The line that saves the game is: Kc7! Bg5 2. Kd6 Kg6 3. Ke6! 3...h5 and Ke5 and it is a draw. Or 3... Kh5 Kf7! 4.Kg4 Nh6 5. Bxh6 and Kg8 with also a draw. The old Rybka version 1.2f, does not have any problems to find that.
Analysis by Rybka 1.2f 32-bit:
1.Ne7
= (-0.19) Depth: 2 00:00:00
1.Ne7
³ (-0.26) Depth: 3 00:00:00
1.Ne7 Bg5
³ (-0.26) Depth: 4 00:00:00
1.Ne7 Bg5 2.Nd5
³ (-0.34) Depth: 5 00:00:00
1.Ne7 Bg5 2.Nd5 h5
³ (-0.34) Depth: 6 00:00:00 9kN
1.Ne7 Bf4+ 2.Kb7 Kf6 3.Nd5+ Kg5 4.Kc6
³ (-0.43) Depth: 7 00:00:00 31kN
1.Ne7 Bf4+ 2.Kb7 Kf6 3.Nd5+ Kg5 4.Kc6 h5
³ (-0.63) Depth: 8 00:00:00 41kN
1.Ne7 Bf4+ 2.Kb7 Kf6 3.Nd5+ Kg5 4.Kc6 h5 5.Nc3
µ (-0.71) Depth: 9 00:00:00 63kN
1.Ne7 Bf4+ 2.Kc8 Kf6 3.Nd5+ Kg5 4.Kd7 h5 5.Nc3 h4
³ (-0.68) Depth: 10 00:00:01 138kN
1.Ne7 Bf4+ 2.Kb7 Kf6 3.Nd5+ Kg5 4.Kc6 h5 5.Nc3 h4 6.Ne4+ Kf5
µ (-0.83) Depth: 11 00:00:01 235kN
1.Ne7 Bf4+ 2.Kb7 Kf6 3.Nd5+ Kg5 4.Kc6 h5 5.Nc3 h4 6.Ne4+ Kf5 7.Nf2
µ (-0.93) Depth: 12 00:00:02 348kN
1.Ne7 Bf4+ 2.Kb7 Kf6 3.Nd5+ Kg5 4.Kc6 h5 5.Nc3 h4 6.Ne4+ Kf5 7.Nf2 Bg3
µ (-1.03) Depth: 13 00:00:02 531kN
1.Ne7 Bf4+ 2.Kb7 Kf6 3.Nd5+ Kg5 4.Kc6 h5 5.Nc3 h4 6.Ne4+ Kf5 7.Nf2 Bg3
µ (-1.19) Depth: 14 00:00:03 782kN
1.Ne7 Kf6 2.Ng8+ Kg5 3.Ne7 Bg7 4.Nd5 Be5+ 5.Kc8 h5 6.Kd7 h4 7.Ne3 h3
µ (-1.32) Depth: 15 00:00:05 2363kN
1.Ne7 Kf6 2.Ng8+ Kg5 3.Ne7 Bg7 4.Nd5 Be5+ 5.Kc8 h5 6.Kd7 h4 7.Ne3 h3
µ (-1.39) Depth: 16 00:00:06 2807kN
1.Kb7 Bf4 2.Kc6 h5 3.Ne7 Bg5 4.Nf5+ Kf6 5.Nd4 h4 6.Kd5 h3 7.Nf3 Kf5
µ (-1.06) Depth: 16 00:00:09 4195kN
1.Kb7 Bg5 2.Kc6 Kg6 3.Kd5 Kf5 4.Kd4 Kg4 5.Kd3 Kf3 6.Kd4 h5 7.Ne7 Bxe7
-+ (-3.59) Depth: 17 00:00:16 8433kN
1.Ne7 Kf6 2.Nd5+ Kf5 3.Kb7 Bg5 4.Kc6 h5 5.Nc3 h4 6.Kd5 h3 7.Nd1 h2
-+ (-2.16) Depth: 17 00:00:20 9976kN
1.Kc7 Bf4+ 2.Kd7 h5 3.Ke6 h4 4.Nf6 h3 5.Ne4 h2 6.Nf2 Kg6 7.Nh1 Kg5
-+ (-1.75) Depth: 17 00:00:25 12458kN
1.Kc7 Bf4+ 2.Kd7 h5 3.Ke6 h4 4.Nf6 h3 5.Ne4 h2 6.Nf2 Kg6 7.Nh1 Kg5
-+ (-1.97) Depth: 18 00:00:26 13189kN
1.Kc7 Bf4+ 2.Kd7 h5 3.Ke6 h4 4.Nf6 h3 5.Ne4 h2 6.Nf2 Kg6 7.Kd5 Be3
-+ (-2.12) Depth: 19 00:00:29 14197kN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 h5 6.Nf5 Ba1 7.Nh4 Kg4
³ (-0.66) Depth: 20 00:01:59 65207kN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 Kg4 6.Nd5 h5 7.Ne3+ Kf4
³ (-0.34) Depth: 21 00:02:24 81389kN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 Kg4 6.Nd5 h5 7.Ne3+ Kf3
³ (-0.34) Depth: 22 00:02:40 92148kN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 Bd4 6.Nf5 Bc3 7.Nh4 Kg5
³ (-0.60) Depth: 23 00:03:02 104mN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 Bd4 6.Nf5 Bc3 7.Nh4 Kg5
µ (-0.77) Depth: 24 00:03:25 117mN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 Bd4 6.Nf5 Bb2 7.Nh4 Kg4
³ (-0.61) Depth: 25 00:07:35 260mN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 Bd4 6.Nf5 Bh8 7.Nh4 Kg4
³ (-0.67) Depth: 26 00:09:34 322mN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 Kg4 6.Nf5 Bb2 7.Ne3+ Kg5
³ (-0.61) Depth: 27 00:26:52 873mN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 Kg4 6.Nf5 Bb2 7.Ne3+ Kg5
³ (-0.61) Depth: 28 00:27:29 891mN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 Kg4 6.Nf5 Bb2 7.Ne3+ Kg5
³ (-0.61) Depth: 29 00:28:18 916mN
(, AMD 09.09.2007)
Regards,
Gambito.
I have seen that Rybka is having some problems with these positions. It would be good, at least to take a look.
5k2/8/pppppppp/8/PPPPPPPP/8/8/5K2 w - - 0 1
In this position, white has a winning sequence. White needs to push his pawns and break the black pawn structure. In my opinion, this secuence should be not so difficult to see for a very strong program like Rybka.
Rybka is thinking this line for a long time: 1. d5! exd5 2. exd5 c5 3. a5 bxa5 4. bxa5?? The first 3 moves are good and strong enough, but 4. bxa5 is a mistake. After this, it is simply a draw.
The engine is just missing for a good time the strongest 4. a5! and after bxa5 then b5! winning.
After b5, bxa5 cxb5 and Ke7, and now white have a passed pawn and black King needs to chase the pawn. Of course, the black king finally hunts it, but then white break the black chain pawn again in the Kingside with g5! and after hxg5 then, f5 gxf5 and h5 and again white achieves to enters with the h pawn and assure the victory easily.
Equally, if black plays fxg5 rather than hxg5, then h5 gxh5 and f5, and now the f pawn enter. White is winning the game without no problems.
Rybka cannot find the winning sequence from the initial position, (Although, I do not know if with a lot of time, it can) so we can go some moves ahead.
5k2/8/pp1p1ppp/3p4/PPP2PPP/8/8/5K2 w - - 0 3
Here, the position change a bit, because Black does not plays c5 (Rybka move). It plays cxd5, but always white has a strong move to break the black pawn structure with their pawns. Rybka again has some troubles to find the winning sequence. The engine thinks for some minutes that cxd5 is the best white move here. But after cxd5, white cannot enter with a pawn in the Queenside, and the Black King it is not more in danger because if White tries to break now in the kingside it is now useless with the Black King so near of them.
It should not be so difficult for Rybka to see how to break the pawns and to pass with some of them, in this case first in the queenside and later in the other side of the board, the kingside. The engine is discarding a5 for some time and in my opinion it should understand it a bit more faster.
The winning line is: 3. a5! bxa5 4. b5 axb5 5. cxb5 Ke7 6. b6 Kd7 7. b7 Kc7, (The black king hunts the white pawn but then), 8. g5 and 1-0, because here black cannot stop the white pawn in the Kingside. It should not be so hard for Rybka
2R5/2P3k1/8/6p1/2r4p/5PPK/7P/8 b - - 0 1
I think that this is the worthiest position to rescue.
In this position, black plays hxg3 but if white plays the very normal hxg3, then white cannot progress anymore. It is a draw. After white plays hxg3, black respond with g4 and white needs to take the g4 pawn with his f pawn. If not, the f white pawn falls.
After g4 and fxg4, it does not care that here White have 2 pawns more in the Kingside. It is as if it would have only one pawn and white cannot progress. The only move to progress is to take the black pawn wit his King. I believe that if a program take the pawn with his King, it is really understanding and knows that it has a win. So after Kxg3 white can win easily and their plan is more simplest.
After Kxg3 Kh7 (Kf7 or Kg6 is a mistake, because the black Rook goes to h8 or give a check and then c8=Q winning)
Rc3 by black is also useless after Kg4, Rc5 and h4! and the f3 white pawn now is free to advance. The idea is to attack the King with the f pawn and to move it from g7 and h7. The black Rook cannot move from the C column because then the White Rook is free to move and with that also the pawn in c7. Rybka is discarding completely Kxg3. But again, after Kxg3 Kh7 and h4! a very strong move. It should not be a big problem for Rybka to see. After gxh4 Kh3 Kg7 and f4 again, the f pawn will walk and somebody have to stop it :-), and with that, white wins easily. In my opinion Rybka needs to understand this easily. Do not you believe?
1K4N1/6kp/7b/8/8/8/8/8 w - - 0 1
Ok this is the same topic of the wrong bishop endgame. And of course 2.3.2a does not know how to save the game. I really hope that for Rybka 3 it can understand this again.
2.3.2a wants to play here 1. Ne7, saving his Knight, that seems the most natural move, but after Bf4+ 2. Kb7 Kf6 3. Kc6 Ke5! 4. Kc5 Bg5 5. Nd5 and h5, black is winning easily. Rybka should understand and let the knight to die and save the game with Kc7!
This endgame appears from time to time and it is important that Rybka knows it.
The line that saves the game is: Kc7! Bg5 2. Kd6 Kg6 3. Ke6! 3...h5 and Ke5 and it is a draw. Or 3... Kh5 Kf7! 4.Kg4 Nh6 5. Bxh6 and Kg8 with also a draw. The old Rybka version 1.2f, does not have any problems to find that.
Analysis by Rybka 1.2f 32-bit:
1.Ne7
= (-0.19) Depth: 2 00:00:00
1.Ne7
³ (-0.26) Depth: 3 00:00:00
1.Ne7 Bg5
³ (-0.26) Depth: 4 00:00:00
1.Ne7 Bg5 2.Nd5
³ (-0.34) Depth: 5 00:00:00
1.Ne7 Bg5 2.Nd5 h5
³ (-0.34) Depth: 6 00:00:00 9kN
1.Ne7 Bf4+ 2.Kb7 Kf6 3.Nd5+ Kg5 4.Kc6
³ (-0.43) Depth: 7 00:00:00 31kN
1.Ne7 Bf4+ 2.Kb7 Kf6 3.Nd5+ Kg5 4.Kc6 h5
³ (-0.63) Depth: 8 00:00:00 41kN
1.Ne7 Bf4+ 2.Kb7 Kf6 3.Nd5+ Kg5 4.Kc6 h5 5.Nc3
µ (-0.71) Depth: 9 00:00:00 63kN
1.Ne7 Bf4+ 2.Kc8 Kf6 3.Nd5+ Kg5 4.Kd7 h5 5.Nc3 h4
³ (-0.68) Depth: 10 00:00:01 138kN
1.Ne7 Bf4+ 2.Kb7 Kf6 3.Nd5+ Kg5 4.Kc6 h5 5.Nc3 h4 6.Ne4+ Kf5
µ (-0.83) Depth: 11 00:00:01 235kN
1.Ne7 Bf4+ 2.Kb7 Kf6 3.Nd5+ Kg5 4.Kc6 h5 5.Nc3 h4 6.Ne4+ Kf5 7.Nf2
µ (-0.93) Depth: 12 00:00:02 348kN
1.Ne7 Bf4+ 2.Kb7 Kf6 3.Nd5+ Kg5 4.Kc6 h5 5.Nc3 h4 6.Ne4+ Kf5 7.Nf2 Bg3
µ (-1.03) Depth: 13 00:00:02 531kN
1.Ne7 Bf4+ 2.Kb7 Kf6 3.Nd5+ Kg5 4.Kc6 h5 5.Nc3 h4 6.Ne4+ Kf5 7.Nf2 Bg3
µ (-1.19) Depth: 14 00:00:03 782kN
1.Ne7 Kf6 2.Ng8+ Kg5 3.Ne7 Bg7 4.Nd5 Be5+ 5.Kc8 h5 6.Kd7 h4 7.Ne3 h3
µ (-1.32) Depth: 15 00:00:05 2363kN
1.Ne7 Kf6 2.Ng8+ Kg5 3.Ne7 Bg7 4.Nd5 Be5+ 5.Kc8 h5 6.Kd7 h4 7.Ne3 h3
µ (-1.39) Depth: 16 00:00:06 2807kN
1.Kb7 Bf4 2.Kc6 h5 3.Ne7 Bg5 4.Nf5+ Kf6 5.Nd4 h4 6.Kd5 h3 7.Nf3 Kf5
µ (-1.06) Depth: 16 00:00:09 4195kN
1.Kb7 Bg5 2.Kc6 Kg6 3.Kd5 Kf5 4.Kd4 Kg4 5.Kd3 Kf3 6.Kd4 h5 7.Ne7 Bxe7
-+ (-3.59) Depth: 17 00:00:16 8433kN
1.Ne7 Kf6 2.Nd5+ Kf5 3.Kb7 Bg5 4.Kc6 h5 5.Nc3 h4 6.Kd5 h3 7.Nd1 h2
-+ (-2.16) Depth: 17 00:00:20 9976kN
1.Kc7 Bf4+ 2.Kd7 h5 3.Ke6 h4 4.Nf6 h3 5.Ne4 h2 6.Nf2 Kg6 7.Nh1 Kg5
-+ (-1.75) Depth: 17 00:00:25 12458kN
1.Kc7 Bf4+ 2.Kd7 h5 3.Ke6 h4 4.Nf6 h3 5.Ne4 h2 6.Nf2 Kg6 7.Nh1 Kg5
-+ (-1.97) Depth: 18 00:00:26 13189kN
1.Kc7 Bf4+ 2.Kd7 h5 3.Ke6 h4 4.Nf6 h3 5.Ne4 h2 6.Nf2 Kg6 7.Kd5 Be3
-+ (-2.12) Depth: 19 00:00:29 14197kN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 h5 6.Nf5 Ba1 7.Nh4 Kg4
³ (-0.66) Depth: 20 00:01:59 65207kN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 Kg4 6.Nd5 h5 7.Ne3+ Kf4
³ (-0.34) Depth: 21 00:02:24 81389kN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 Kg4 6.Nd5 h5 7.Ne3+ Kf3
³ (-0.34) Depth: 22 00:02:40 92148kN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 Bd4 6.Nf5 Bc3 7.Nh4 Kg5
³ (-0.60) Depth: 23 00:03:02 104mN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 Bd4 6.Nf5 Bc3 7.Nh4 Kg5
µ (-0.77) Depth: 24 00:03:25 117mN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 Bd4 6.Nf5 Bb2 7.Nh4 Kg4
³ (-0.61) Depth: 25 00:07:35 260mN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 Bd4 6.Nf5 Bh8 7.Nh4 Kg4
³ (-0.67) Depth: 26 00:09:34 322mN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 Kg4 6.Nf5 Bb2 7.Ne3+ Kg5
³ (-0.61) Depth: 27 00:26:52 873mN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 Kg4 6.Nf5 Bb2 7.Ne3+ Kg5
³ (-0.61) Depth: 28 00:27:29 891mN
1.Kc7 Kg6 2.Kd6 Bg7 3.Ne7+ Kg5 4.Kd5 Kf4 5.Ke6 Kg4 6.Nf5 Bb2 7.Ne3+ Kg5
³ (-0.61) Depth: 29 00:28:18 916mN
(, AMD 09.09.2007)
Regards,
Gambito.
![[nl]](/mwf/v12/flags/nl.png "Netherlands")
I think this has to do with the continuing story of advantages and disadvantages of using Larry Kaufman's piece evaluations.
Programs that rely more on brute force calculation instead of in deep chess knowledge are able to solve these positions given enough time.
If Vas gets this right in version 3 he should be able to win another 100-200 ELO points!
Btw here is Fritz 10 20 ply analysis for your 3rd position. It prefers Kxg3 although it doesn't recognize (yet?!) that hg3x is drawn.
New game
Analysis by Fritz 10:
1. +- (5.39): 2.Kxg3 Kh7 3.h4 gxh4+ 4.Kh3 Rc3 5.Kxh4 Rc4+ 6.Kg5 Rc6 7.f4 Rg6+ 8.Kh5 Rh6+ 9.Kg4 Rg6+ 10.Kh4 Rh6+
2. +- (4.05): 2.hxg3 g4+ 3.fxg4 Rc5 4.g5 Rc4 5.g6 Rc6 6.g4 Rc1 7.g5 Rc4 8.Kg2 Rg4+ 9.Kh2 Rc4 10.Kg3
3. ± (0.86): 2.Rd8 Rxc7 3.hxg3 Rc4 4.Re8 g4+ 5.fxg4 Kf6 6.Rf8+ Kg7 7.Rd8
4. ² (0.66): 2.Kg2 gxh2 3.Kxh2 Rc2+ 4.Kh3 Rc4 5.Kg3 Kh7 6.Kf2 Kg7 7.Ke3 Kh7 8.Kd3 Rc6 9.Ke4 Kg7 10.Ke5 Rc3
5. = (0.00): 2.Rb8 gxh2 3.Kxh2 Rxc7 4.Kg3 Kg6 5.f4 gxf4+ 6.Kxf4 Rg7 7.Rh8
6. = (0.00): 2.Ra8 gxh2 3.Kxh2 Rxc7 4.Kg3 Kg6 5.f4 gxf4+ 6.Kxf4 Rg7
7. = (0.00): 2.Re8 gxh2 3.Kxh2 Rxc7 4.Kg3 Kf7 5.Re5 Kf6 6.f4 gxf4+ 7.Kxf4 Rc8
8. = (0.00): 2.f4 gxh2 3.Kxh2 gxf4 4.Kg2 Rc3 5.Re8 Rxc7 6.Kf3 Rf7 7.Re4 Kg6 8.Rxf4 Rxf4+ 9.Kxf4
9. -+ (-4.42): 2.Rg8+ Kxg8 3.c8Q+ Rxc8 4.hxg3 Kf7 5.Kg2 Rc2+ 6.Kg1 Kg7 7.Kf1 Kf6
Programs that rely more on brute force calculation instead of in deep chess knowledge are able to solve these positions given enough time.
If Vas gets this right in version 3 he should be able to win another 100-200 ELO points!
Btw here is Fritz 10 20 ply analysis for your 3rd position. It prefers Kxg3 although it doesn't recognize (yet?!) that hg3x is drawn.
New game
2R5/2P3k1/8/6p1/2r5/5PpK/7P/8 w - - 0 1
Analysis by Fritz 10:
1. +- (5.39): 2.Kxg3 Kh7 3.h4 gxh4+ 4.Kh3 Rc3 5.Kxh4 Rc4+ 6.Kg5 Rc6 7.f4 Rg6+ 8.Kh5 Rh6+ 9.Kg4 Rg6+ 10.Kh4 Rh6+
2. +- (4.05): 2.hxg3 g4+ 3.fxg4 Rc5 4.g5 Rc4 5.g6 Rc6 6.g4 Rc1 7.g5 Rc4 8.Kg2 Rg4+ 9.Kh2 Rc4 10.Kg3
3. ± (0.86): 2.Rd8 Rxc7 3.hxg3 Rc4 4.Re8 g4+ 5.fxg4 Kf6 6.Rf8+ Kg7 7.Rd8
4. ² (0.66): 2.Kg2 gxh2 3.Kxh2 Rc2+ 4.Kh3 Rc4 5.Kg3 Kh7 6.Kf2 Kg7 7.Ke3 Kh7 8.Kd3 Rc6 9.Ke4 Kg7 10.Ke5 Rc3
5. = (0.00): 2.Rb8 gxh2 3.Kxh2 Rxc7 4.Kg3 Kg6 5.f4 gxf4+ 6.Kxf4 Rg7 7.Rh8
6. = (0.00): 2.Ra8 gxh2 3.Kxh2 Rxc7 4.Kg3 Kg6 5.f4 gxf4+ 6.Kxf4 Rg7
7. = (0.00): 2.Re8 gxh2 3.Kxh2 Rxc7 4.Kg3 Kf7 5.Re5 Kf6 6.f4 gxf4+ 7.Kxf4 Rc8
8. = (0.00): 2.f4 gxh2 3.Kxh2 gxf4 4.Kg2 Rc3 5.Re8 Rxc7 6.Kf3 Rf7 7.Re4 Kg6 8.Rxf4 Rxf4+ 9.Kxf4
9. -+ (-4.42): 2.Rg8+ Kxg8 3.c8Q+ Rxc8 4.hxg3 Kf7 5.Kg2 Rc2+ 6.Kg1 Kg7 7.Kf1 Kf6
This is a very good ending to use as a test, but getting endings like this right is more likely to be worth 2 Elo points than 200; they just don't come up very often. Still it is a worthy goal.
Larry, you may have missed this question from another thread but does Rybka use dynamic piece evaluations, ie does the value of the pieces change depending on the pieces on the board?
Or is this simply classified information? ;)
Or is this simply classified information? ;)
Yes, the value of the pieces depends on what other pieces are on the board.
Ok thanks. Still puzzled than by the strange evaluation of some pawn endgames - but assume this has to do with the 'pawn endgame module' (whatever that is).
Ever thought about trying to codifying positional characteristics as well and adding those to the piece values (eg classical ...Rc3x exchange sacrifice in the Sicilian).
Ever thought about trying to codifying positional characteristics as well and adding those to the piece values (eg classical ...Rc3x exchange sacrifice in the Sicilian).
I'm not quite sure what you mean, but a move like ...Rxc3 exchange sac is evaluated based on whether the damage to the pawn structure makes up for the material loss. In a normal case where Black wins the "e" pawn and gives White a doubled isolated pawn on c3, I think the answer will be that it is a correct "sacrifice".
>This is a very good ending to use as a test, but getting endings like this right is more likely to be worth 2 Elo points than 200; they just don't >come up very often.
Performance of play in human-engine or engine-engine is not the only factor for a programmer, that wants his engine to be called top, to be considered.
Analysis strength, is the other area many other people are interested.
So 2 ELO, 3 ELO, 0.05 ELO it doesn't matter since some people buy Rybka to analyze games/positions with it and want it to be able to solve everything.
Everything is not possible for an engine to solve, but the logic of this position does worth 1 ELO so it's not a problem is not a good logic in my opinion.... Not a good logic for people that want a Chess analysis tool....
>Still it is a worthy goal.
Yes, absolutely..... :-)
![[fr]](/mwf/v12/flags/fr.png "France")
My Rybka on my machine (dual core pentium) has found the b5! key move... in d-24. The sequence is: 1.d5 cxd5 2.a5 bxa5 3.b5.
Did it find it from the initial position? How much time did it take?
Regards,
Gambito.
Regards,
Gambito.
In ten minutes, with the shootout function. It is quicker than the infinite analyse.
(If you would have more details, remove, please, your Yehova slogan... I dont beleive in your Yehova, I beleive only in Rybka! :)
(If you would have more details, remove, please, your Yehova slogan... I dont beleive in your Yehova, I beleive only in Rybka! :)
![[hu]](/mwf/v12/flags/hu.png "Hungary")
Re. these endgames:
1) Pawn endgame - this is really hard for a computer. For example, in your second diagram, any human will recognize the f4, g4, h4 vs f6, g6, h6 cluster on the kingside as being able to product a passed pawn for white. This type of 'mental chunk' makes analyzing everything much easier. Rybka has to prove this to herself with search from every single type of queenside configuration.
2) Rook endgame - yes, we need more knowledge for this. It's important, although I'm not sure we can get to it for Rybka 3. Soon for sure, though.
3) Wrong rook pawn - yes, this is easy.
Vas
1) Pawn endgame - this is really hard for a computer. For example, in your second diagram, any human will recognize the f4, g4, h4 vs f6, g6, h6 cluster on the kingside as being able to product a passed pawn for white. This type of 'mental chunk' makes analyzing everything much easier. Rybka has to prove this to herself with search from every single type of queenside configuration.
2) Rook endgame - yes, we need more knowledge for this. It's important, although I'm not sure we can get to it for Rybka 3. Soon for sure, though.
3) Wrong rook pawn - yes, this is easy.
Vas
This endgame inprovement is important for the annotation work that I do. Thank you for working on this, as the additional chess knowledge in the final stages will inevitably mean better understanding up the analytical chain.
Let me say how much I appreciate Rybka. It lets me and my readers go a lot deeper into what is an amazing game.The extraordinary chess of even a Fischer turns out to be a relative surface underneath which a much deeper level of chess occurs.
The enjoyment of the evening sky is fine in itself, but the addition of the telescope enabled Galileo to see the moons around Jupiter, an unexpected, amazing experience for him. I think of Rybka (more than any other engine) as enabling a similar qualitative leap. Version 3.0 will be a lot of fun. Thanks.
Let me say how much I appreciate Rybka. It lets me and my readers go a lot deeper into what is an amazing game.The extraordinary chess of even a Fischer turns out to be a relative surface underneath which a much deeper level of chess occurs.
The enjoyment of the evening sky is fine in itself, but the addition of the telescope enabled Galileo to see the moons around Jupiter, an unexpected, amazing experience for him. I think of Rybka (more than any other engine) as enabling a similar qualitative leap. Version 3.0 will be a lot of fun. Thanks.
Didn't that amazing experience get him excommunicated? :-)
Alan
Alan
Yes, and it was entirely to his credit.
![[us]](/mwf/v12/flags/us.png "United States")
Yes, and as you know, Vas has been similarly excommunicated by Chessbase.
I'd like to add some counterbalance to the Chessbase bashing.
True, Chessbase has for quite some time been ignoring Rybka but I think they are slowly turning around and start recognizing Vas his efforts. See http://www.chessbase.com/newsdetail.asp?newsid=3936
In addition, let's not forget that Chessbase has brought superior products to the chess community over the past 15+ years. They have been setting the standard in chess database software and been offering excellent game databases, training DVD's and engines.
Be aware that unfortunately there are very little companies that are interested in and committed to chess, so let's be carefull with the ones that do.
There's enough room for everbody!
True, Chessbase has for quite some time been ignoring Rybka but I think they are slowly turning around and start recognizing Vas his efforts. See http://www.chessbase.com/newsdetail.asp?newsid=3936
In addition, let's not forget that Chessbase has brought superior products to the chess community over the past 15+ years. They have been setting the standard in chess database software and been offering excellent game databases, training DVD's and engines.
Be aware that unfortunately there are very little companies that are interested in and committed to chess, so let's be carefull with the ones that do.
There's enough room for everbody!
![[de]](/mwf/v12/flags/de.png "Germany")
There's enough room for everbody!
No! "In the end, there can be only one." (Highlander) ;-)
No! "In the end, there can be only one." (Highlander) ;-)
![[il]](/mwf/v12/flags/il.png "Israel")
I think that rybka is missing more basic knowledge than recognizing the f4 g4 h4 vs f6 g6 h6 and I am talking about the question when passed pawns are unstoppable.
I played against rybka at depth 1 from move 9(game at the end of this post)
At move 9 it evaluated herself as completely winning when playing gxh5 that give white unstoppable passed pawns
and even one move later she did not see that white is winning and gave only small advantage for herself.
Note that the black pawn is also almost unstoppable and possible line is 10.f5 a4 11.f6 a3 12.f7 a2 13.b8Q+ Kxb8 14.f8Q+ when white win only because white can promote with check.
Note that calculation of the winner of pawn race by evaluation when the winner can win only thanks to checkin the opponent when the check is thanks to some sacrifice is extremely complex
Compare between these 2 positions
white is winning
White is losing
I doubt if there is a chess program that know both of them based on evaluation.
Distance to promotion of both sides seem to be the same.
[Event "2880'/1"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "agur"]
[Black "Rybka 2.3.2a 32-bit"]
[Result "*"]
[SetUp "1"]
[FEN "5k2/8/pppppppp/8/PPPPPPPP/8/8/5K2 w - - 0 1"]
[PlyCount "19"]
[TimeControl "1/172800:0/0:0/0"]
1. d5 exd5 2. exd5 c5 (2... cxd5 3. a5 bxa5 4. b5 axb5 5. cxb5 Ke7 6. b6 Kd7 7.
b7 Kc7 8. g5) 3. a5 bxa5 4. b5 axb5 5. cxb5 Ke7 6. b6 Kd7 7. b7 Kc7 8. g5 fxg5
9. h5 gxh5 (9... gxh5 {-10.99/1 0} 10. f5 {12} Kb8 {0.22/1 0}) 10. f5 *
I played against rybka at depth 1 from move 9(game at the end of this post)
At move 9 it evaluated herself as completely winning when playing gxh5 that give white unstoppable passed pawns
and even one move later she did not see that white is winning and gave only small advantage for herself.
Note that the black pawn is also almost unstoppable and possible line is 10.f5 a4 11.f6 a3 12.f7 a2 13.b8Q+ Kxb8 14.f8Q+ when white win only because white can promote with check.
Note that calculation of the winner of pawn race by evaluation when the winner can win only thanks to checkin the opponent when the check is thanks to some sacrifice is extremely complex
Compare between these 2 positions
white is winning
8/1Pk5/3p3p/p1pP2pp/5P2/8/8/5K2 w - - 0 10
White is losing
8/k7/3p3p/p1pP2pp/5P2/8/8/5K2 w - - 0 10
I doubt if there is a chess program that know both of them based on evaluation.
Distance to promotion of both sides seem to be the same.
[Event "2880'/1"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "agur"]
[Black "Rybka 2.3.2a 32-bit"]
[Result "*"]
[SetUp "1"]
[FEN "5k2/8/pppppppp/8/PPPPPPPP/8/8/5K2 w - - 0 1"]
[PlyCount "19"]
[TimeControl "1/172800:0/0:0/0"]
1. d5 exd5 2. exd5 c5 (2... cxd5 3. a5 bxa5 4. b5 axb5 5. cxb5 Ke7 6. b6 Kd7 7.
b7 Kc7 8. g5) 3. a5 bxa5 4. b5 axb5 5. cxb5 Ke7 6. b6 Kd7 7. b7 Kc7 8. g5 fxg5
9. h5 gxh5 (9... gxh5 {-10.99/1 0} 10. f5 {12} Kb8 {0.22/1 0}) 10. f5 *
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill