Ehrhardt Post vs Nimzowitsch 1905, draw or not ? Crazy rook

[syzygy]

Here Ehrhardt Post played Rxf3 !? (or ?!), shallow analyze computer show a very high score but no win because zillions of checks ... what's the result with an accurate play ?
[d]8/1R6/8/6K1/p3P2r/k4P2/3R4/5r2 b - - 3 60

After Rxf3, right?

My engine (6-piece TBs):

Code: Select all

  0:00.178  19   7107422    +Win13   61.Kxh4 Rf4+ 62.Kg5 Rf5+ 63.Kg6 Rf6+ 64.Kg7 Rg6+ 65.Kh8 Rg8+ 66.Kh7 Rh8+ 67.Kg6 Rh6+ 68.Kg5 Rh5+ 69.Kf4 Rf5+ 70.Ke3 Rf3+ 71.Kd4 Rf8 72.Ra7 Rd8+ 73.Ke3 Rh8 74.Rxa4+

White wins.

Stockfish also finds it quickly:

Code: Select all

info depth 26 seldepth 38 score cp 12345 nodes 5206440 nps 6235257 tbhits 252510 time 835 multipv 1 pv g5h4 f3f4 h4g5 f4f5 g5g6 f5f6 g6g7 f6g6 g7h8 g6g8 h8h7 g8h8 h7g6 h8h6 g6f5 h6h5 f5f4 h5h4 f4e3 h4h3 e3d4 h3h8 b7a7 h8d8 d4e3 d8c8 a7a4

[syzygy]

i analyzed this w/ SF 6man for about an hour. the score was 123.xx after Rxf3 & never altered. so either i didnt give it enough time to solve for mate or houdini has some nice bit of code that allowed it to recognize draw.

123.xx = TB win.
SF won't find mate starting from the position after Rxf3, as it does not search lines that enter a 6-piece position any deeper. It knows the outcome of these lines, so there is no need to search deeper.

[Vinvin]

... houdini has some nice bit of code that allowed it to recognize draw.

Yes, that's the problem (or is it a solution ?), there's so many checks that Houdini assumes it's draw until it find a more accurate eval. May be Stockfish search deeper because of its better selectivity and find winning line(s). With 6 pieces EGTB, after Kxe4 (forced), the win is clear.

[yanquis1972]

ah, its possible i forced kxe4 & then hit back w/out clearing hash. as i recall it was a pretty instant solution.

thx for the info ronald.

[syzygy]

Are you both mistyping Kxh4 as Kxe4?
Earlier Vincent wrote Kxh4 and that white would then be forced to capture the (white!) e4 pawn. Am I the only one having trouble following this?

SF+TB should find the win near instantly.

[yanquis1972]

yes, typo on my part, although i blame vincent ;D

of course looking at it now i see that TBs are essentially cheating; i sortve mindlessly popped this in (forgot i had it runnning actually) & since i know kburcham runs very high-end stuff i was assuming he was running houdini w/ full TBs [ok, looking back, turns out he did]

SF does evaluate it as mating in a couple seconds on my machine after Rxf3

[Vinvin]

Are you both mistyping Kxh4 as Kxe4?
Earlier Vincent wrote Kxh4 and that white would then be forced to capture the (white!) e4 pawn. Am I the only one having trouble following this?

SF+TB should find the win near instantly.

I don't think I mistype something

[syzygy]

Are you both mistyping Kxh4 as Kxe4?
Earlier Vincent wrote Kxh4 and that white would then be forced to capture the (white!) e4 pawn. Am I the only one having trouble following this?

SF+TB should find the win near instantly.

I don't think I mistype something

You did not?

With 6 pieces EGTB, after Kxe4 (forced), the win is clear.

Explain Kxe4.

[hcn]

[FEN "8/1R6/8/6K1/p3P2r/k4r2/3R4/8 w - - 0 61"]

61. Kxh4 {White mates in 38:} 61... Rf4+ 62. Kh5 Rf5+ 63. Kh6 Rf6+ 64. Kg7
Rg6+ 65. Kh8 Rg8+ 66. Kh7 Rh8+ 67. Kg6 Rh6+ 68. Kg5 Rg6+ 69. Kf4 Rg4+ 70. Ke5
Rxe4+ 71. Kd5 Re5+ {Lomonosov tables}{White mates in 28:} 72. Kc6 Rc5+ 73. Kd7
Rb5 74. Rc7 Rg5 75. Rd3+ Kb4 76. Ke6 Rh5 77. Rd1 Rh3 78. Rdc1 Rh6+ 79. Kd5
Rh5+ 80. Kd4 Rh4+ 81. Kd3 Rh3+ 82. Kc2 {Lomonosov tables}{White mates in 17:}
82... Rh2+ 83. Kb1 Rh3 84. Rb7+ Ka3 85. Rd1 Rc3 86. Rdd7 Rf3 87. Kc2 Ka2 88.
Rd4 Rf2+ 89. Kd1 Rf3 90. Rxa4+ Ra3 91. Rab4 Rf3 92. Ra7+ Ra3 {Lomonosov tables}

[Vinvin]

Are you both mistyping Kxh4 as Kxe4?
Earlier Vincent wrote Kxh4 and that white would then be forced to capture the (white!) e4 pawn. Am I the only one having trouble following this?

SF+TB should find the win near instantly.

I don't think I mistype something

You did not?

With 6 pieces EGTB, after Kxe4 (forced), the win is clear.

Explain Kxe4.

Ah ok, I see now, it's Rxe4 ... a mismatch sometimes happens because in French R=Roi=King. Thinking in 2 languages is sometimes not easy ...
Sorry !