Alekhine - Capablanca 1927 once more

[Arpad Rusz]

One of the previous positions of the game was the following:

[D]8/5pk1/5rpp/8/P2R3P/6P1/5PK1/8 w - - 0 51

The game went like this 51. a5 Ra6 52. Rd5 Ooops! 52...Rf6 53. Rd4 Ra6 54. Ra4 The White Rook is happy now.

[Jouni]

Can somebody with super hardware run first position overnight to see, if either move gets winning score?

Jouni

[Terry McCracken]

Can somebody with super hardware run first position overnight to see, if either move gets winning score?

Jouni

It doesn't matter. The correct move is to support your a-pawn by going behind it. Protecting it on the Rank rather than on the File is simply incorrect even if it still wins here.

Robert Houdart said that Houdini is putting too much weight on centralization in this position.

[Robert Flesher]

Can somebody with super hardware run first position overnight to see, if either move gets winning score?

Jouni

It doesn't matter. The correct move is to support your a-pawn by going behind it. Protecting it on the Rank rather than on the File is simply incorrect even if it still wins here.

Robert Houdart said that Houdini is putting too much weight on centralization in this position.

Heya Terry, I felt the same as you when I first look at this position. I understand that putting your rook behind passed pawns is a principle rule, however, it cannot always be blindly followed. Rd5!? is an interesting move, and black seems helpless to either plan. Rd5 protects the a pawn, but also controls the black kings approach by controlling the 5th rank, so this gives Rd5 some merit. Us humans will play what is natural to us, but the engines/computers will continue to find the strange wins

[Terry McCracken]

Can somebody with super hardware run first position overnight to see, if either move gets winning score?

Jouni

It doesn't matter. The correct move is to support your a-pawn by going behind it. Protecting it on the Rank rather than on the File is simply incorrect even if it still wins here.

Robert Houdart said that Houdini is putting too much weight on centralization in this position.

Heya Terry, I felt the same as you when I first look at this position. I understand that putting your rook behind passed pawns is a principle rule, however, it cannot always be blindly followed. Rd5!? is an interesting move, and black seems helpless to either plan. Rd5 protects the a pawn, but also controls the black kings approach by controlling the 5th rank, so this gives Rd5 some merit. Us humans will play what is natural to us, but the engines/computers will continue to find the strange wins

I think it's winning but clumsy. It does have the advantage of keeping the Black King in his own camp and in many cases that is the right approach.
Still I'd choose Ra4 and hogtie my opponent and win the remaining pawns with my King in due time.

[Jouni]

Stockfish 111026 is very selective and reached 39 ply in hour in dual Pentium:

1.Rd5 Rc6 2.Kf3 Kf6 3.Ke3 Ke6 4.Rb5 Rc4 5.Rb6+ Kf5 6.Rb7 Ke6 7.a6 Ra4 8.a7 g5 9.Kd3 gxh4 10.gxh4 f6 11.h5 Kd5 12.f4 Ra3+ 13.Kc2 Kc6 14.Rh7 Kd5 15.Rb7
+- (2.82) Depth: 39/67 01:02:29 9879mN

But of course SF can display +9,xx and it's still tablebase draw

Jouni

[Terry McCracken]

Stockfish 111026 is very selective and reached 39 ply in hour in dual Pentium:

1.Rd5 Rc6 2.Kf3 Kf6 3.Ke3 Ke6 4.Rb5 Rc4 5.Rb6+ Kf5 6.Rb7 Ke6 7.a6 Ra4 8.a7 g5 9.Kd3 gxh4 10.gxh4 f6 11.h5 Kd5 12.f4 Ra3+ 13.Kc2 Kc6 14.Rh7 Kd5 15.Rb7
+- (2.82) Depth: 39/67 01:02:29 9879mN

But of course SF can display +9,xx and it's still tablebase draw

Jouni

[Dan Astrachan]

For what it's worth, I pitted two fairly strong engines (Critter 1.2 SSE 4 and Ivanhoe 46f) against each other.
My machine is reasonably fast (Quadcore, i875, 12 Gb of RAM).

I gave both engines two hours + 60 seconds per move played, and let them play out until either a mate or draw was reached.

In the first two games, the starting position was:

[D] 8/5pk1/r5pp/P2R4/7P/6P1/5PK1/8 b - - 1 54
which is the position posted by the OP after White played Rd5, which is the move most engines recommend in this position.

Result: both engines won the game with White.

Game 1: Critter_1.2_64bit_SSE4 - Ivanhoe 46f x64 Analysis
53 ... Kf6 54. Rb5 Ke7 55. g4 Kd7 56. Kg3 Ke8 57. h5 Ke7 58. Rb7+ Kf8 59. hxg6 fxg6 60. Rb6 Rxa5 61. Rxg6 Ra3+ 62. f3 h5 63. g5 Ra1 64. f4 Rg1+ 65. Kf2 Rc1 66. Rh6 Kf7 67. Rxh5 Rc3 68. Rh7+ Ke6 69. Rh6+ Kf7 70. f5 Rc5 71. Rf6+ Kg7 72. Kf3 Ra5 73. Kg4 Ra4+ 74. Kh5 Ra1 75. Rc6 Rh1+ 76. Kg4 Rg1+ 77. Kf4 Rf1+ 78. Ke4 Re1+ 79. Kd5 Rd1+ 80. Ke6 Re1+ 81. Kd6 Kf7 82. Rc7+ Ke8 83. g6 Rd1+ 84. Ke5 Kd8 85. Ra7 Re1+ 86. Kf6 Ke8 87. g7 Rg1 88. Ke6 Kd8 89. f6 Kc8 90. Kf7 Kb8 91. Re7 Kc8 92. g8=Q+ Rxg8 93. Kxg8 Kd8 94. Kg7 Kc8 95. f7 Kd8 96. f8=Q# 1-0

Here is the final position of this game:

[D] 3k1Q2/4R1K1/8/8/8/8/8/8 b - - 0 97

Game 2: Ivanhoe 46f x64 Analysis - Critter_1.2_64bit_SSE4
53 ... Kf6 54. Kf3 Ke6 55. Rb5 Kd7 56. h5 Kc6 57. Rb6+ Rxb6 58. axb6 gxh5 59. Ke4 Kxb6 60. Kd5 Kc7 61. Ke5 Kc6 62. f4 h4 63. gxh4 h5 64. f5 Kd7 65. Kf6 Ke8 66. Kg5 Kf8 67. f6 Kg8 68. Kxh5 Kh8 69. Kg4 Kh7 70. Kg5 Kg8 71. Kf4 Kh7 72. Ke4 Kh6 73. Kd5 Kh5 74. Kd6 Kh6 75. Ke7 Kg6 76. h5+ Kxh5 77. Kxf7 Kg5 78. Ke6 Kf4 79. Kd5 Kg4 80. f7 Kf3 81. f8=Q+ Ke3 82. Qf1 Kd2 83. Kd4 Kc2 84. Qb5 Kd1 85. Qb2 Ke1 86. Ke3 Kd1 87. Qd2# 1-0

Here is the final position of this game:

[D] 8/8/8/8/8/4K3/3Q4/3k4 b - - 12 88

In the last two games, the starting position was:

[D] 8/5pk1/r5pp/P7/R6P/6P1/5PK1/8 b - - 1 54
which is the position posted by the OP after White played Ra4, which is the move Alekhine actually played.

Result: both engines won the game with White.

Game 3: Critter_1.2_64bit_SSE4 - Ivanhoe 46f x64 Analysis
53 ... Kf6 54. Kf3 Ke5 55. Ke3 Kd5 56. h5 Re6+ 57. Kd2 Ra6 58. Kd3 Kd6 59. Ke4 Ke6 60. Kd4 Kd7 61. hxg6 fxg6 62. f4 Kd6 63. Ke4 Ke6 64. Ra3 h5 65. Ra4 Kd7 66. Rd4+ Kc7 67. Rd5 Re6+ 68. Kf3 Rc6 69. Rb5 Rc2 70. Rg5 Rc6 71. Kg2 Rf6 72. Rc5+ Kd7 73. Rb5 Ke8 74. Kh3 Ra6 75. Kh4 Kf7 76. Rb7+ Kf8 77. Rb6 Rxa5 78. Rxg6 Kf7 79. Rh6 Ra1 80. Kxh5 Ra5+ 81. Kg4 Ra3 82. Rb6 Re3 83. f5 Ra3 84. Kf4 Ra4+ 85. Kg5 Ra7 86. g4 Rc7 87. Rb5 Rc4 88. f6 Rc8 89. Rb7+ Ke6 90. f7 Rd8 91. Kg6 Kd5 92. g5 Ra8 93. Kg7 Ke4 94. g6 Kf5 95. f8=Q+ Rxf8 96. Rb5+ Ke4 97. Kxf8 Kd4 98. g7 Kc4 99. Rf5 Kd3 100. g8=Q Ke4 101. Qg4+ Ke3 102. Rf3+ Ke2 103. Qg2+ Ke1 104. Rf1# 1-0

Here is the final position of this game:

[D] 5K2/8/8/8/8/8/6Q1/4kR2 b - - 8 105

Game 4: Ivanhoe 46f x64 Analysis - Critter_1.2_64bit_SSE4
53 ... Kf6 54. Kf3 Ke5 55. Ke3 Kd5 56. h5 Kd6 57. hxg6 fxg6 58. Ke4 Ke6 59. f4 h5 60. Kd4 Kd6 61. Ra3 Kd7 62. Ra1 Kd6 63. Ra2 Kd7 64. Kd5 Kd8 65. Ke5 Kc7 66. Rc2+ Kd7 67. Rc5 Ke7 68. Kd5 Rf6 69. Kc4 Rd6 70. Re5+ Kf7 71. Kb5 Rd3 72. a6 Rxg3 73. a7 Ra3 74. Kb6 Rxa7 75. Kxa7 h4 76. f5 gxf5 77. Rxf5+ Kg6 78. Rd5 Kh6 79. Kb6 h3 80. Rd3 h2 81. Rh3+ Kg5 82. Rxh2 Kf4 83. Kc5 Kg3 84. Rh7 Kf4 85. Rh4+ Kg5 86. Rd4 Kf6 87. Kd6 Kf5 88. Rb4 Kf6 89. Rb5 Kf7 90. Rf5+ Kg6 91. Ke6 Kg7 92. Rg5+ Kh6 93. Kf6 Kh7 94. Rh5+ Kg8 95. Rh6 Kf8 96. Rh8# 1-0

Here is the final position of this game:

[D] 5k1R/8/5K2/8/8/8/8/8 b - - 28 97

Both moves (Rd5 and Ra4) lead to a victory by White, so saying Rd5 is not a valid choice to play in the position posted by the OP appears to be wrong.

[Terry McCracken]

For what it's worth, I pitted two fairly strong engines (Critter 1.2 SSE 4 and Ivanhoe 46f) against each other.
My machine is reasonably fast (Quadcore, i875, 12 Gb of RAM).

I gave both engines two hours + 60 seconds per move played, and let them play out until either a mate or draw was reached.

In the first two games, the starting position was:

[D] 8/5pk1/r5pp/P2R4/7P/6P1/5PK1/8 b - - 1 54
which is the position posted by the OP after White played Rd5, which is the move most engines recommend in this position.

Result: both engines won the game with White.

Game 1: Critter_1.2_64bit_SSE4 - Ivanhoe 46f x64 Analysis
53 ... Kf6 54. Rb5 Ke7 55. g4 Kd7 56. Kg3 Ke8 57. h5 Ke7 58. Rb7+ Kf8 59. hxg6 fxg6 60. Rb6 Rxa5 61. Rxg6 Ra3+ 62. f3 h5 63. g5 Ra1 64. f4 Rg1+ 65. Kf2 Rc1 66. Rh6 Kf7 67. Rxh5 Rc3 68. Rh7+ Ke6 69. Rh6+ Kf7 70. f5 Rc5 71. Rf6+ Kg7 72. Kf3 Ra5 73. Kg4 Ra4+ 74. Kh5 Ra1 75. Rc6 Rh1+ 76. Kg4 Rg1+ 77. Kf4 Rf1+ 78. Ke4 Re1+ 79. Kd5 Rd1+ 80. Ke6 Re1+ 81. Kd6 Kf7 82. Rc7+ Ke8 83. g6 Rd1+ 84. Ke5 Kd8 85. Ra7 Re1+ 86. Kf6 Ke8 87. g7 Rg1 88. Ke6 Kd8 89. f6 Kc8 90. Kf7 Kb8 91. Re7 Kc8 92. g8=Q+ Rxg8 93. Kxg8 Kd8 94. Kg7 Kc8 95. f7 Kd8 96. f8=Q# 1-0

Here is the final position of this game:

[D] 3k1Q2/4R1K1/8/8/8/8/8/8 b - - 0 97

Game 2: Ivanhoe 46f x64 Analysis - Critter_1.2_64bit_SSE4
53 ... Kf6 54. Kf3 Ke6 55. Rb5 Kd7 56. h5 Kc6 57. Rb6+ Rxb6 58. axb6 gxh5 59. Ke4 Kxb6 60. Kd5 Kc7 61. Ke5 Kc6 62. f4 h4 63. gxh4 h5 64. f5 Kd7 65. Kf6 Ke8 66. Kg5 Kf8 67. f6 Kg8 68. Kxh5 Kh8 69. Kg4 Kh7 70. Kg5 Kg8 71. Kf4 Kh7 72. Ke4 Kh6 73. Kd5 Kh5 74. Kd6 Kh6 75. Ke7 Kg6 76. h5+ Kxh5 77. Kxf7 Kg5 78. Ke6 Kf4 79. Kd5 Kg4 80. f7 Kf3 81. f8=Q+ Ke3 82. Qf1 Kd2 83. Kd4 Kc2 84. Qb5 Kd1 85. Qb2 Ke1 86. Ke3 Kd1 87. Qd2# 1-0

Here is the final position of this game:

[D] 8/8/8/8/8/4K3/3Q4/3k4 b - - 12 88

In the last two games, the starting position was:

[D] 8/5pk1/r5pp/P7/R6P/6P1/5PK1/8 b - - 1 54
which is the position posted by the OP after White played Ra4, which is the move Alekhine actually played.

Result: both engines won the game with White.

Game 3: Critter_1.2_64bit_SSE4 - Ivanhoe 46f x64 Analysis
53 ... Kf6 54. Kf3 Ke5 55. Ke3 Kd5 56. h5 Re6+ 57. Kd2 Ra6 58. Kd3 Kd6 59. Ke4 Ke6 60. Kd4 Kd7 61. hxg6 fxg6 62. f4 Kd6 63. Ke4 Ke6 64. Ra3 h5 65. Ra4 Kd7 66. Rd4+ Kc7 67. Rd5 Re6+ 68. Kf3 Rc6 69. Rb5 Rc2 70. Rg5 Rc6 71. Kg2 Rf6 72. Rc5+ Kd7 73. Rb5 Ke8 74. Kh3 Ra6 75. Kh4 Kf7 76. Rb7+ Kf8 77. Rb6 Rxa5 78. Rxg6 Kf7 79. Rh6 Ra1 80. Kxh5 Ra5+ 81. Kg4 Ra3 82. Rb6 Re3 83. f5 Ra3 84. Kf4 Ra4+ 85. Kg5 Ra7 86. g4 Rc7 87. Rb5 Rc4 88. f6 Rc8 89. Rb7+ Ke6 90. f7 Rd8 91. Kg6 Kd5 92. g5 Ra8 93. Kg7 Ke4 94. g6 Kf5 95. f8=Q+ Rxf8 96. Rb5+ Ke4 97. Kxf8 Kd4 98. g7 Kc4 99. Rf5 Kd3 100. g8=Q Ke4 101. Qg4+ Ke3 102. Rf3+ Ke2 103. Qg2+ Ke1 104. Rf1# 1-0

Here is the final position of this game:

[D] 5K2/8/8/8/8/8/6Q1/4kR2 b - - 8 105

Game 4: Ivanhoe 46f x64 Analysis - Critter_1.2_64bit_SSE4
53 ... Kf6 54. Kf3 Ke5 55. Ke3 Kd5 56. h5 Kd6 57. hxg6 fxg6 58. Ke4 Ke6 59. f4 h5 60. Kd4 Kd6 61. Ra3 Kd7 62. Ra1 Kd6 63. Ra2 Kd7 64. Kd5 Kd8 65. Ke5 Kc7 66. Rc2+ Kd7 67. Rc5 Ke7 68. Kd5 Rf6 69. Kc4 Rd6 70. Re5+ Kf7 71. Kb5 Rd3 72. a6 Rxg3 73. a7 Ra3 74. Kb6 Rxa7 75. Kxa7 h4 76. f5 gxf5 77. Rxf5+ Kg6 78. Rd5 Kh6 79. Kb6 h3 80. Rd3 h2 81. Rh3+ Kg5 82. Rxh2 Kf4 83. Kc5 Kg3 84. Rh7 Kf4 85. Rh4+ Kg5 86. Rd4 Kf6 87. Kd6 Kf5 88. Rb4 Kf6 89. Rb5 Kf7 90. Rf5+ Kg6 91. Ke6 Kg7 92. Rg5+ Kh6 93. Kf6 Kh7 94. Rh5+ Kg8 95. Rh6 Kf8 96. Rh8# 1-0

Here is the final position of this game:

[D] 5k1R/8/5K2/8/8/8/8/8 b - - 28 97

Both moves (Rd5 and Ra4) lead to a victory by White, so saying Rd5 is not a valid choice to play in the position posted by the OP appears to be wrong.

Never said it lost but it breaks a basic rule but otoh it fulfills another rule and in this case it works.

Chess is a complicated, imperfect game-Which just adds to it's complexity and fascination.

[lech]

Two questions I will try to explain.
The first is: a value of evaluation means nothing (it may be +20) if there is no gain in next depths. A joke about Stockfish is unfounded.
The second is: a static evaluation is not able to give a correct estimate for both the moves: 1.Ra4 and 1.Rd5, because some added bonuses works for 51% (or more) cases mainly.
An infinity search (a much larger depth) should verified legitimacy of such bonuses, though.
An interesting simple example:
[d] 8/6kP/1p6/p7/Pr6/K2B4/8/8 b - -
All engines give a bonus for a proximity of Kings to passed pawns.
But now solution is completely other.
1…Rh4 and next Kg7-f6-e5-d4 with win.
Try your liked engine whether is able to break this bonus.

Now, I hope you know why engines have a problem to choose 1.Ra4