Another endgame that the Rybka family fails at

[Robert Flesher]

[d]8/5R2/8/6k1/2p5/2P3K1/2r2PP1/8 b - - 0 67


This endgame is far from simple, but any strong human would clearly say, "white wins!" There seems to be alot of engines that do not understand the power of connected passers. I am not saying black wins, but only that I believe he does with ease. I have yet to work out the variations. However, if black does win here, it shows that computers can still be lead to endgames where they are loose. But, you decide!

[Dann Corbit]

[d]8/5R2/8/6k1/2p5/2P3K1/2r2PP1/8 b - - 0 67


This endgame is far from simple, but any strong human would clearly say, "white wins!" There seems to be alot of engines that do not understand the power of connected passers. I am not saying black wins, but only that I believe he does with ease. I have yet to work out the variations. However, if black does win here, it shows that computers can still be lead to endgames where they are loose. But, you decide!

Start from this position, and show us your win against Rybka.

[Robert Flesher]

[d]8/5R2/8/6k1/2p5/2P3K1/2r2PP1/8 b - - 0 67


This endgame is far from simple, but any strong human would clearly say, "white wins!" There seems to be alot of engines that do not understand the power of connected passers. I am not saying black wins, but only that I believe he does with ease. I have yet to work out the variations. However, if black does win here, it shows that computers can still be lead to endgames where they are loose. But, you decide!

Start from this position, and show us your win against Rybka.

Perhaps you think I was implying this was a game I played against Rybka. It was not! However, I do look over hundreds of games each week to improve my own chess. I was not ego tripping if that is what you think, Rybka crushes me 99.8% of the time. But, this line seems to win the game.




68. f3 Rc1 69. f4+ Kg6 70. Re7 Kf6 71. Re5 c3 72. Rc5 Ke6 73. Kh4 Kd6
74. Rc8 Kd7 75. Rc4 Ke6 76. Kg5 Kd5 77. Rc8 Rg1 78. g4 Rc1 79. Kf6 1-0

[Robert Flesher]

I started from this position after the obvious capture.
[d]8/5R2/8/6k1/2p5/2r3K1/5PP1/8 w - - 0 68



68. f3 Rc1 69. f4+ Kg6 70. Re7 Kf6 71. Re5 c3 72. Rc5 Ke6 73. Kh4 Kd6
74. Rc8 Kd7 75. Rc4 Ke6 76. Kg5 Kd5 77. Rc8 Rg1 78. g4 Rc1 79. Kf6 1-0



So the plan seems obvious, draw the king away from the pawns and then advance. However, if black tries to keep the king in front of the passers it only serves to help white gain time as the connected passers can harass the king.

[Dann Corbit]

[d]8/5R2/8/6k1/2p5/2P3K1/2r2PP1/8 b - - 0 67


This endgame is far from simple, but any strong human would clearly say, "white wins!" There seems to be alot of engines that do not understand the power of connected passers. I am not saying black wins, but only that I believe he does with ease. I have yet to work out the variations. However, if black does win here, it shows that computers can still be lead to endgames where they are loose. But, you decide!

Start from this position, and show us your win against Rybka.

Perhaps you think I was implying this was a game I played against Rybka. It was not! However, I do look over hundreds of games each week to improve my own chess. I was not ego tripping if that is what you think, Rybka crushes me 99.8% of the time. But, this line seems to win the game.




68. f3 Rc1 69. f4+ Kg6 70. Re7 Kf6 71. Re5 c3 72. Rc5 Ke6 73. Kh4 Kd6
74. Rc8 Kd7 75. Rc4 Ke6 76. Kg5 Kd5 77. Rc8 Rg1 78. g4 Rc1 79. Kf6 1-0

You said:
"However, if black does win here, it shows that computers can still be lead to endgames where they are loose."

However, the line you propose shows white winning.

So far, every engine I tried (including Rybka) shows an immediate and building advantage for white of about one pawn. This agrees with the line that you posted, approximately.

Code: Select all

8/5R2/8/6k1/2p5/2P3K1/2r2PP1/8 b - - acd 31; acn 117463945; acs 1810; bm Rxc3+; ce -105; pv Rxc3+ f3 Rc2 f4+ Kg6 Rf8 Kg7 Rf5 Kg6 Re5; c0 "Rybka";
8/5R2/8/6k1/2p5/2P3K1/2r2PP1/8 b - - acd 30/80; acn 2218522102; acs 1941; bm Rxc3+; ce -81; pv Rxc3+ f3 Rc1 Re7 Rc2 f4+ Kf6 Re5 Ra2 Rc5 Rc2 Kf3 Rc3+ Ke4 Rg3 Rc6+ Kf7 Rxc4 Rxg2 Kf5 Ke7 Re4+ Kd6 Kf6 Kd5 Re7 Rf2 f5 Kd6 Ra7 Kd5 Ra3 Rf1 Ra5+ Ke4 Re5+ Kd3; c0 "Bright";
8/5R2/8/6k1/2p5/2P3K1/2r2PP1/8 b - - acd 33; acn 10350942402; acs 1183; bm Rxc3+; ce -87; pv Rxc3+ f3 Rc1 Rf8 c3 f4+ Kg6 Kg4; c0 "Junior";
8/5R2/8/6k1/2p5/2P3K1/2r2PP1/8 b - - acd 31/61; acn 5404709649; acs 1509; bm Rxc3+; ce -98; pv Rxc3+ f3 Rc1 Re7 Rc2 f4+ Kf6 Re5; c0 "Shredder";

Now, if Rybka is defective in assessing this position, so are Bright, Junior, and Shredder.

[Robert Flesher]

Sorry, was a long day in the sun, and a few to many scotches. I meant "white".

[Robert Flesher]

[d]8/5R2/8/6k1/2p5/2P3K1/2r2PP1/8 b - - 0 67


This endgame is far from simple, but any strong human would clearly say, "white wins!" There seems to be alot of engines that do not understand the power of connected passers. I am not saying WHITE wins, but only that I believe he does with ease. I have yet to work out the variations. However, if WHITE does win here, it shows that computers can still be lead to endgames where they are loose. But, you decide!

I guess I should proof read what I write......

My post was to illustrate that Rybka and others could not see this is winning for white. They lack some key endgame fundamentals. This knowledge if implemented would be nice for analysis, but who knows if it would slow the engine down and drop overall strength.

[Dann Corbit]

[d]8/5R2/8/6k1/2p5/2P3K1/2r2PP1/8 b - - 0 67


This endgame is far from simple, but any strong human would clearly say, "white wins!" There seems to be alot of engines that do not understand the power of connected passers. I am not saying WHITE wins, but only that I believe he does with ease. I have yet to work out the variations. However, if WHITE does win here, it shows that computers can still be lead to endgames where they are loose. But, you decide!

I guess I should proof read what I write......

My post was to illustrate that Rybka and others could not see this is winning for white. They lack some key endgame fundamentals. This knowledge if implemented would be nice for analysis, but who knows if it would slow the engine down and drop overall strength.

Every engine I tried {including Rybka 3.0} sees an advantage for white, starting at 1/2 pawn or so and growing to about a pawn. Then it holds at about one pawn for quite a while.

I suspect it might be drawn with best possible play from both sides.

4 additional evaluations:

Code: Select all

8/5R2/8/6k1/2p5/2P3K1/2r2PP1/8 b - - acd 28/61; acn 5416567067; acs 1214; bm Rxc3+; ce -101; pv Rxc3+ f3 Rc1 Re7 Rc2 f4+ Kf6 Re5 c3 Kf3 Rc1 Rc5 Ke6 Kg4 Kd6 Rc8 Kd7 Rc4 Kd6 Kg5 Kd5 Rc8 Kd4 f5 Rf1 f6 Kd3 Kg6 c2 f7 Kd2 Kg7 c1=Q Rxc1 Kxc1 g4; c0 "Deep Sjeng";
8/5R2/8/6k1/2p5/2P3K1/2r2PP1/8 b - - acd 35; acn 7760210400; acs 1335; bm Rxc3+; ce -60; pv Rxc3+ f3 Rc1 f4+ Kg6 Re7 Kf6 Re5 Rc2 Kf3 Rc3+ Kg4 Rd3 Rc5 Rd4 Kf3 Rd3+ Ke4 Rg3 Rxc4 Rxg2 Rc6+ Kg7 Kf5 Rf2 Rb6 Rf3 Rb7+ Kh6 Re7 Rf2 Re6+ Kg7 Rd6 Rf1 Rd2 Kh6 Rd5 Rf2 Rd6+ Kg7 Rd3 Ra2 Rd7+ Kh6 Rd6+ Kh5; co "Fire";
8/5R2/8/6k1/2p5/2P3K1/2r2PP1/8 b - - acd 23/41; acn 198224509; acs 1640; bm Rxc3+; ce -75; pv Rxc3+ f3 Rc1 Re7 Rc2 f4+ Kf6 Re1; co "Hiarcs";
8/5R2/8/6k1/2p5/2P3K1/2r2PP1/8 b - - acd 39; acn 8802631241; acs 1460; bm Rxc3+; ce -59; pv Rxc3+ f3 Rc1 f4+ Kg6 Re7 Kf6 Re5 Rc2 Kf3 Rc3+ Kg4 Rd3 Rc5 Rd4 Kf3 Rd3+ Ke4 Rg3 Rc6+ Kg7 Kf5 c3 Ke5 Re3+ Kd4 Rg3 Rxc3 Rxg2 Ke5 Re2+ Kf5 Rb2 Rc7+ Kh6 Re7 Rf2 Rd7 Rf3 Rd1 Kg7 Rd6 Rf2 Rd5 Rf3 Rd2 Rf1 Rd7+ Kh6 Rd6+ Kg7 Rd4 Rf2 Rd7+ Kh6 Ke4; c0  "IvanHoe";

[PauloSoare]

How about a module for endgames? GM Larry Kaufman wrote
about it.

[kingliveson]

The position is draw.

Code: Select all

kingliveson@cygnitec:~> IvanHoe
IvanHoe 9.63b x64
compiled with PREFETCH
compiled with POPCNT
Capitalist multicore mode compiled
Mode with UTILITIES compiled
RobboBases enabled by compile
TotalBaseCache is 1mb + (1mb)
position fen 8/5R2/8/6k1/2p5/2P3K1/2r2PP1/8 b - - 0 67
drawboard
      +---+---+---+---+---+---+---+---+
   8  |   | . |   | . |   | . |   | . |
      +---+---+---+---+---+---+---+---+
   7  | . |   | . |   | . | R | . |   |
      +---+---+---+---+---+---+---+---+
   6  |   | . |   | . |   | . |   | . |
      +---+---+---+---+---+---+---+---+
   5  | . |   | . |   | . |   |<K>|   |
      +---+---+---+---+---+---+---+---+
   4  |   | . |<P>| . |   | . |   | . |
      +---+---+---+---+---+---+---+---+
   3  | . |   | P |   | . |   | K |   |
      +---+---+---+---+---+---+---+---+
   2  |   | . |<R>| . |   | P | P | . |
      +---+---+---+---+---+---+---+---+
   1  | . |   | . |   | . |   | . |   |
      +---+---+---+---+---+---+---+---+
        a   b   c   d   e   f   g   h
8/5R2/8/6k1/2p5/2P3K1/2r2PP1/8 b - - 0 0
setoption name RobboTotalBaseDirectory value /mnt/sdb5/robbobases/RobboBlock/RobboTotalBase
Optional RobboTotalBaseDirectory /mnt/sdb5/robbobases/RobboBlock/RobboTotalBase
setoption name RegisterRobboTotalBases
info string Reading directory /mnt/sdb5/robbobases/RobboBlock/RobboTotalBase
info string 364 RobboTotalBases registered with 791436 indici 3mb
setoption name RobboTripleBaseDirectory value /mnt/sdb5/robbobases/RobboBlock/RobboTripleBase
Optional RobboTripleBaseDirectory /mnt/sdb5/robbobases/RobboBlock/RobboTripleBase
setoption name LoadRobboTripleBases
info string Reading directory /mnt/sdb5/robbobases/RobboBlock/RobboTripleBase
info string 364 RobboTripleBases (645mb): 615535840 data 14805806+9823864 indici
setoption name Hash value 1024
Optional TranspositionalTabularSizing 1024
setoption name PawnsHash value 128
Optional PawnsZobrist 128
go
info multipv 1 time 9 nodes 14 nps 1000 score cp -13 depth 1 pv c2c3 f2f3
info multipv 1 time 9 nodes 23 nps 2000 score cp -13 depth 1 pv c2c3 f2f3
info depth 2
info multipv 1 time 9 nodes 29 nps 3000 tbhits 1 score cp -13 depth 2 pv c2c3 f2f3
info multipv 1 time 9 nodes 41 nps 4000 tbhits 3 score cp -13 depth 2 pv c2c3 f2f3
info depth 3
info multipv 1 time 10 nodes 114 nps 11000 tbhits 7 score cp -5 depth 3 pv c2c3 f2f3 c3d3
info multipv 1 time 10 nodes 131 nps 13000 tbhits 7 score cp -5 depth 3 pv c2c3 f2f3 c3d3
info depth 4
info multipv 1 time 10 nodes 362 nps 36000 tbhits 24 score cp -12 depth 4 pv c2c3 f2f3 c3d3 g3f2
info multipv 1 time 10 nodes 392 nps 39000 tbhits 24 score cp -12 depth 4 pv c2c3 f2f3 c3d3 g3f2
info depth 5
info multipv 1 time 10 nodes 1038 nps 103000 tbhits 45 score cp 13 depth 5 pv c2c3 f2f3 c3d3 g3f2 c4c3 f3f4 g5g4
info multipv 1 time 10 nodes 1067 nps 106000 tbhits 47 score cp 13 depth 5 pv c2c3 f2f3 c3d3 g3f2 c4c3 f3f4 g5g4
info depth 6
info multipv 1 time 11 nodes 2115 nps 192000 tbhits 100 score cp 14 depth 6 pv c2c3 f2f3 c3d3 g3f2 c4c3 f7c7 d3d2 f2e3
info multipv 1 time 11 nodes 2145 nps 195000 tbhits 100 score cp 14 depth 6 pv c2c3 f2f3 c3d3 g3f2 c4c3 f7c7 d3d2 f2e3
info depth 7
info multipv 1 time 11 nodes 2860 nps 260000 tbhits 113 score cp 14 depth 7 pv c2c3 f2f3 c3d3 g3f2 c4c3 f7c7 g5f6 f2e2 d3d2 e2e3
info multipv 1 time 11 nodes 2894 nps 263000 tbhits 113 score cp 14 depth 7 pv c2c3 f2f3 c3d3 g3f2 c4c3 f7c7 g5f6 f2e2 d3d2 e2e3
info depth 8
info multipv 1 time 12 nodes 4268 nps 355000 tbhits 153 score cp 6 upperbound depth 8 pv c2c3 f2f3 c3d3 g3f2 c4c3 f7c7 g5f4 c7c4 f4e5 c4e4 e5f5
info depth 8
info multipv 1 time 13 nodes 5427 nps 417000 tbhits 158 score cp 6 depth 8 pv c2c3 f2f3 c3d3 g3f2 c4c3 g2g4 d3d5 f7c7 d5d2 f2e3
info multipv 1 time 13 nodes 5436 nps 418000 tbhits 158 score cp 6 depth 8 pv c2c3 f2f3 c3d3 g3f2 c4c3 g2g4 d3d5 f7c7 d5d2 f2e3
info depth 9
info multipv 1 time 17 nodes 14475 nps 851000 tbhits 448 score cp 0 depth 9 pv c2c3 f2f3 c3b3 f7c7 c4c3 c7c5 g5g6 g3f4 b3b2 c5c3 b2g2 c3c5
info multipv 1 time 18 nodes 14685 nps 815000 tbhits 450 score cp 0 depth 9 pv c2c3 f2f3 c3b3 f7c7 c4c3 c7c5 g5g6 g3f4 b3b2 c5c3 b2g2 c3c5
info depth 10
.
.
.
info multipv 1 time 806940 nodes 4775756172 nps 5918000 tbhits 524499694 score cp -43 depth 38 pv c2c3 f2f3 c3c1 f7e7 c1c2 f3f4 g5f6 e7e5 c4c3 g3f3 f6f7 g2g4 c2c1 f3g2 c1d1 e5c5 d1d3 g2f2 f7e6 f2e2 d3h3 g4g5 h3g3 c5e5 e6d6 e5f5 d6e6 f5c5 e6d6 c5b5 g3g4 b5b4 g4g2 e2f3 c3c2 b4d4 d6e6 d4e4 e6d6 e4c4 g2d2 f3e4 d2g2 c4c8 d6d7 c8c5 d7d6 c5c3 g2e2 e4d3 e2f2 d3e3 f2h2