Rook endgame: Wojtaszek-Navara

[Albert Silver]

I did not do a lot of analysis but
the plan for black seems simple for me.
1)put the black pawn at b2 and force the white king to go to b1
2)make waiting move until white has no pawn moves and has to move the rook so black won a pawn.
3)win the rook endgame with pawn up that should be easy when you have a rook behind the pawn because you progress with your king and support the pawns and the opponent king cannot defend because if the opponent is going to capture the pawn you trade rooks and win the pawn endgame and if the and if the opponent try to move the king and block with the rook you sacrifice the pawn by promoting it and again trade rooks and win the pawn endgame.

White may try to defend by sacrificing the pawn and blocking the black pawn with the rook but I believe that it also wins for black.

Well now, after
1. Kd2 b3 2. Kc1 b2+ 3. Kb1 Kh6 4. h4 f5 5. g4 fxg4 6. fxg4 Rb3 7. Rc8 Rxb7 8.Rc4 Kg7 9. g5 h5
we have attained stage 3
[D]8/1r4k1/6p1/6Pp/2R4P/8/1p6/1K6 w - h6 0 10
Is this so straightforward to win for Black?
Computers don't think so...

Black can execute the plan a little better than that you know. Uri's rough explanation was correct, but that doesn't mean Black has to blindly play any old move and let White give him as much trouble as possible.

As in most plans of this type, Black should try and soften the pawns up and gain space as much as possible. I played this out a bit against Houdini, and sure enough its eval spiked after a bit:

1. Kd2 b3 2. Kc1 h5 (I think Navara's choice far superior to Houdini's and Rybka's 2...b2+) 3. Kb2 Kf6 4. g4 h4 5. f4 Rb5 6. Rh8 Rxb7 7. Rxh4 Rb4 8.g5+ Kf5 9. Rh7 Rb7 10. Rh4 Ke4 11. Rg4 Kf3 (strangely, Houdini thought Ke3? is the correct way to continue, though I could not understand its line at all) 12. Rh4 Rb4 13. Rh7 Rxf4 14. Kxb3 Kg4 15 Kc2 Kxg5 16. Kd3 f5 and by now its eval went past -2.xx and climbing.

It may be there are better ways to try and defend, though I let Houdini think for about a minute a move here.

It is also worth mentioning that tablebases make a big difference here. Compare the final position using Houdini x64 without GTBs and with:

[D]8/7R/6p1/5pk1/5r2/3K4/7P/8 w - f6 0 1

Analysis by Houdini 1.5 x64 Gaviota (with GTBs):

17.Rh3 Ra4 18.Ke2 Ra2+ 19.Ke1 f4 20.Rh8 Kg4 21.Rh7 g5 22.Kf1 Rb2 23.Rh8 Kf3 24.Rh3+ Ke4 25.Ke1 f3 26.Rh6 g4 27.Kf1 Rd2 28.Rh7 Kf4 29.Rf7+ Ke5 30.Rg7 Kf5 31.Rh7 Ke5 32.Rh8 Kf4 33.Rf8+ Ke5 34.Re8+ Kf4 35.Rf8+ Ke5
-+ (-1.46) Depth: 25/46 00:00:18 64959kN, tb=19520
17.Rh3 Ra4
-+ (-1.55 --) Depth: 26/48 00:00:31 114mN, tb=38090
17.Rh3 Ra4
-+ (-1.72 --) Depth: 26/48 00:00:38 145mN, tb=58387
17.Rh3 Rh4 18.Rg3+ Rg4 19.Rh3 Rg2 20.Rh7 Ra2 21.h3 Ra4 22.Ke2 Rh4 23.Rd7 Rxh3 24.Rg7 f4 25.Kf2 Re3 26.Kg2 Kf5 27.Kf2 g5 28.Ra7 Rb3 29.Ra2 g4 30.Ra5+ Kg6 31.Ra2 g3+ 32.Kg2 Kf5 33.Rd2 Re3 34.Rd5+ Kg4
-+ (-1.88) Depth: 26/48 00:00:52 200mN, tb=90160
17.Rh3 Rh4
-+ (-1.98 --) Depth: 27/50 00:01:22 319mN, tb=155208
17.Rh3 Rh4
-+ (-2.17 --) Depth: 27/54 00:03:21 913mN, tb=291067


Analysis by Houdini 1.5 x64 (without GTBs):

17.Rh3 Ra4 18.Rg3+ Kf6 19.Rg2 Ra3+ 20.Kd4 g5 21.Re2 f4 22.Ke4 Rh3 23.Rg2 Re3+ 24.Kd4 Kf5 25.Ra2 Re4+ 26.Kd3 Re5 27.Rb2 f3 28.Rb7 g4 29.Rf7+ Kg6 30.Rf4 Kh5 31.Rf8 Kg5 32.Rg8+ Kh4 33.Rh8+ Rh5 34.Re8 Rc5 35.Rh8+ Kg5 36.Rg8+ Kh4 37.Rh8+ Kg5
-/+ (-1.33) Depth: 27/47 00:00:25 96399kN
17.Rh3 Ra4 18.Rg3+ Kf6 19.Rg2 Ra3+ 20.Kd4 g5 21.Re2 f4 22.Ke4 Rh3 23.Rg2 Re3+ 24.Kd4 Kf5 25.Ra2 Re4+ 26.Kd3 Re5 27.Rb2 g4 28.Rb7 f3 29.Rf7+ Kg6 30.Rf4 Kh5 31.Rf8 Kg5 32.Rg8+ Kh4 33.Rh8+ Rh5 34.Rg8 Kh3 35.Ke4 f2
-/+ (-1.33) Depth: 28/47 00:00:30 119mN
17.Rh3 Ra4
-+ (-1.41 --) Depth: 29/54 00:01:05 260mN
17.Rh3 Ra4
-+ (-1.58 --) Depth: 29/54 00:02:06 508mN

[Karlo Bala]

I did not do a lot of analysis but
the plan for black seems simple for me.
1)put the black pawn at b2 and force the white king to go to b1
2)make waiting move until white has no pawn moves and has to move the rook so black won a pawn.
3)win the rook endgame with pawn up that should be easy when you have a rook behind the pawn because you progress with your king and support the pawns and the opponent king cannot defend because if the opponent is going to capture the pawn you trade rooks and win the pawn endgame and if the and if the opponent try to move the king and block with the rook you sacrifice the pawn by promoting it and again trade rooks and win the pawn endgame.

White may try to defend by sacrificing the pawn and blocking the black pawn with the rook but I believe that it also wins for black.

Well now, after
1. Kd2 b3 2. Kc1 b2+ 3. Kb1 Kh6 4. h4 f5 5. g4 fxg4 6. fxg4 Rb3 7. Rc8 Rxb7 8.Rc4 Kg7 9. g5 h5
we have attained stage 3
[D]8/1r4k1/6p1/6Pp/2R4P/8/1p6/1K6 w - h6 0 10
Is this so straightforward to win for Black?
Computers don't think so...

Black can execute the plan a little better than that you know. Uri's rough explanation was correct, but that doesn't mean Black has to blindly play any old move and let White give him as much trouble as possible.

As in most plans of this type, Black should try and soften the pawns up and gain space as much as possible. I played this out a bit against Houdini, and sure enough its eval spiked after a bit:

1. Kd2 b3 2. Kc1 h5 (I think Navara's choice far superior to Houdini's and Rybka's 2...b2+) 3. Kb2 Kf6 4. g4 h4 5. f4 Rb5 6. Rh8 Rxb7 7. Rxh4 Rb4 8.g5+ Kf5 9. Rh7 Rb7 10. Rh4 Ke4 11. Rg4 Kf3 (strangely, Houdini thought Ke3? is the correct way to continue, though I could not understand its line at all) 12. Rh4 Rb4 13. Rh7 Rxf4 14. Kxb3 Kg4 15 Kc2 Kxg5 16. Kd3 f5 and by now its eval went past -2.xx and climbing.

Before playing 6. Rh8 white should play 6. h3, and for example 6.... Rb4 7. Rh8 Rxb7 8. Rxh4 Rb4 9. f5! Kg7 10. fxg6 fxg6 11. Kc1 b2 12. Kb1

[Albert Silver]

I did not do a lot of analysis but
the plan for black seems simple for me.
1)put the black pawn at b2 and force the white king to go to b1
2)make waiting move until white has no pawn moves and has to move the rook so black won a pawn.
3)win the rook endgame with pawn up that should be easy when you have a rook behind the pawn because you progress with your king and support the pawns and the opponent king cannot defend because if the opponent is going to capture the pawn you trade rooks and win the pawn endgame and if the and if the opponent try to move the king and block with the rook you sacrifice the pawn by promoting it and again trade rooks and win the pawn endgame.

White may try to defend by sacrificing the pawn and blocking the black pawn with the rook but I believe that it also wins for black.

Well now, after
1. Kd2 b3 2. Kc1 b2+ 3. Kb1 Kh6 4. h4 f5 5. g4 fxg4 6. fxg4 Rb3 7. Rc8 Rxb7 8.Rc4 Kg7 9. g5 h5
we have attained stage 3
[D]8/1r4k1/6p1/6Pp/2R4P/8/1p6/1K6 w - h6 0 10
Is this so straightforward to win for Black?
Computers don't think so...

Black can execute the plan a little better than that you know. Uri's rough explanation was correct, but that doesn't mean Black has to blindly play any old move and let White give him as much trouble as possible.

As in most plans of this type, Black should try and soften the pawns up and gain space as much as possible. I played this out a bit against Houdini, and sure enough its eval spiked after a bit:

[D]1R6/1P3pkp/1r4p1/8/1p6/5P2/4K1PP/8 w - - 0 41

1. Kd2 b3 2. Kc1 h5 (I think Navara's choice far superior to Houdini's and Rybka's 2...b2+) 3. Kb2 Kf6 4. g4 h4 5. f4 Rb5 6. Rh8 Rxb7 7. Rxh4 Rb4 8.g5+ Kf5 9. Rh7 Rb7 10. Rh4 Ke4 11. Rg4 Kf3 (strangely, Houdini thought Ke3? is the correct way to continue, though I could not understand its line at all) 12. Rh4 Rb4 13. Rh7 Rxf4 14. Kxb3 Kg4 15 Kc2 Kxg5 16. Kd3 f5 and by now its eval went past -2.xx and climbing.

Before playing 6. Rh8 white should play 6. h3, and for example 6.... Rb4 7. Rh8 Rxb7 8. Rxh4 Rb4 9. f5! Kg7 10. fxg6 fxg6 11. Kc1 b2 12. Kb1

If you play 6.h3, you allow me the resource 6...Kg7, after which play might go 7.f5 (7.g5 Rb4 8.Rc8 Rxb7 9.Rc4 Rb8 followed by Re8 and white cannot defend) gxf5 8. gxf5 Kf6 9. Rh8 Rxb7 10.Rxh4 Kxf5 and Black is won.

You'll notice the idea is unchanged though. White's king is tied down to the b-pawn, allowing Black a freehand on the kingside.

[Karlo Bala]

If you play 6.h3, you allow me the resource 6...Kg7, after which play might (7.g5 Rb4 8.Rc8 Rxb7 9.Rc4 Rb8 followed by Re8 and white cannot defend)

It is hard to defend white position. However, after 10. Rd4 Re8 11. Kxb3 Re3+ 12. Kc4
[D]8/5pk1/6p1/6P1/2KR1P1p/4r2P/8/8 b - - 2 52

I let Houdini and Stockfish to play this position but they didn't know how to win.

[playjunior]

If you play 6.h3, you allow me the resource 6...Kg7, after which play might (7.g5 Rb4 8.Rc8 Rxb7 9.Rc4 Rb8 followed by Re8 and white cannot defend)

It is hard to defend white position. However, after 10. Rd4 Re8 11. Kxb3 Re3+ 12. Kc4
[D]8/5pk1/6p1/6P1/2KR1P1p/4r2P/8/8 b - - 2 52

I let Houdini and Stockfish to play this position but they didn't know how to win.

this one looks drawish, put the rook somewhere on e3 so that the black king cannot approach and approach the king to the pawn. Black has some ideas but the king is too far away.

[Albert Silver]

If you play 6.h3, you allow me the resource 6...Kg7, after which play might (7.g5 Rb4 8.Rc8 Rxb7 9.Rc4 Rb8 followed by Re8 and white cannot defend)

It is hard to defend white position. However, after 10. Rd4 Re8 11. Kxb3 Re3+ 12. Kc4
[D]8/5pk1/6p1/6P1/2KR1P1p/4r2P/8/8 b - - 2 52

I let Houdini and Stockfish to play this position but they didn't know how to win.

this one looks drawish, put the rook somewhere on e3 so that the black king cannot approach and approach the king to the pawn. Black has some ideas but the king is too far away.

Yes, I agree, despite the hopelessly optimistic evals by the engines, I think this is drawn.

[ernest]

1. Kd2 b3 2. Kc1 h5 (I think Navara's choice far superior to Houdini's and Rybka's 2...b2+) 3. Kb2 Kf6 4. g4 h4 5. f4 Rb5 6. Rh8 Rxb7 7. Rxh4 Rb4 8.g5+ Kf5 9. Rh7 Rb7 10. Rh4 Ke4 11. Rg4 Kf3 (strangely, Houdini thought Ke3? is the correct way to continue, though I could not understand its line at all) 12. Rh4 Rb4 13. Rh7 Rxf4 14. Kxb3 Kg4 15 Kc2 Kxg5 16. Kd3 f5 and by now its eval went past -2.xx and climbing.

Hi Albert,

Zappa Maxico II doesn't like your 8.g5+ defense (eval -1.45) and prefers 8.h3 (eval -0.71)
FWIW...

I think that's a really interesting game/position, which challenges the endgame capabilities of today's best engines.

[Albert Silver]

1. Kd2 b3 2. Kc1 h5 (I think Navara's choice far superior to Houdini's and Rybka's 2...b2+) 3. Kb2 Kf6 4. g4 h4 5. f4 Rb5 6. Rh8 Rxb7 7. Rxh4 Rb4 8.g5+ Kf5 9. Rh7 Rb7 10. Rh4 Ke4 11. Rg4 Kf3 (strangely, Houdini thought Ke3? is the correct way to continue, though I could not understand its line at all) 12. Rh4 Rb4 13. Rh7 Rxf4 14. Kxb3 Kg4 15 Kc2 Kxg5 16. Kd3 f5 and by now its eval went past -2.xx and climbing.

Hi Albert,

Zappa Maxico II doesn't like your 8.g5+ defense (eval -1.45) and prefers 8.h3 (eval -0.71)
FWIW...

I think that's a really interesting game/position, which challenges the endgame capabilities of today's best engines.

Not *my* defense. It was Houdini's choice:

Analysis by Houdini 1.5 x64 Gaviota:

8.g5+ Kf5 9.Rh7 Rb7 10.Rh4 Ke4 11.Rg4 Rb5 12.Rh4 Kf3 13.h3 Kg3 14.Rh7 Kxf4 15.Rxf7+ Kxg5 16.Rh7 Kf4 17.Rh4+ Kf5 18.Rh7 Kf6 19.Rh4 Rh5 20.Rf4+ Rf5 21.Rc4 Rf3 22.Rc6+ Kg7 23.Rc3 Rxc3 24.Kxc3 Kf6 25.Kxb3 Kf5 26.Kc4 Ke5 27.Kd3 g5 28.Ke3 Kd5 29.Kd3
= (-0.16) Depth: 31/49 00:00:43 135mN, tb=32314
8.g5+ Kf5 9.Rh7 Rb7 10.Rh4 Ke4 11.Rg4 Rb5 12.Rh4 Kf3 13.h3 Kg3 14.Rh7 Kxf4 15.Rxf7+ Kxg5 16.Rh7 Kf4 17.Rh4+ Kf5 18.Rh7 Kf6 19.Rh4 Rh5 20.Rf4+ Rf5 21.Rc4 Rf3 22.Rc6+ Kg7 23.Rc3 Rxc3 24.Kxc3 Kf6 25.Kxb3 Kf5 26.Kc4
= (-0.16) Depth: 32/50 00:00:59 183mN, tb=49529

In any case, the endgame does serve to show that there is a lot of progress yet to be made, no matter how strong overall. The top engines still make a number of important mistakes that can be the difference between a draw or a win/loss.