From Division 2 of an on-going GM tournament:
White has a bit of difficulties, but I think this should be drawn. Various analysis spouting about seems rather faulty, based upon annotation by result perhaps.
White's 43. h4 is just asking for trouble if you ask me. A clearer path to the draw seems to me to be 43. Kb2. However, it still seems like White can draw, if he gives up the b-pawn a bit sooner. In particular, the weakening 45. f4 is not to be recommended. Instead, 45. Rc8 Rxb7 46. Rc4, and White just shuttles on the fourth rank Ra4-c4-a4-f4 (if necessary). Eventually Black plays g5 (exchanging Rooks on the b-file leads to a drawn pawn endgame), upon which White exchanges pawns and plays g3. For instance:
Black can't win here: 58... Rb3 59. Rc8 Kf6 60. Rc6+ Ke5 61. Rc5+ Ke6 62. Rc6+ Kd5 63. Rh6 Rxf3 64. Rxh5 Ke4 65. Rg5 Rf2 and draw. White has enough checking distance.
Any improvements?
How to draw rook endgames
Re: How to draw rook endgames
Maybe Black can try 46. Rc4 Kf6 47. Ra4 Ke5 48. Rc4 Rd6!?, but I think White again draws after 49. f4+ Ke6 50. Rb4 Rd1+ 51. Kxb2 Rd2+ 52. Kc3 Rxg2 53. Rb6, and even if Black wins another pawn he cannot win. Same with 49. f4+ Kf6 50. Kxb2 Rd2+ 51. Kc3 Rxg2 52. Ra6+, and Black will find his King too harassed and his Rook too off-side to co-ordinate to victory. A typical position:And Black can make no progress, as far as I can tell, with the classical draw with f- and h-pawns creeping into play.
[I meant to put a
after the subject title, but failed to do so].
[I meant to put a
after the subject title, but failed to do so].Re: How to draw rook endgames
Another try: 45. Rc8 Rxb7 46. Rc4 Kf6 47. Ra4 Ke5 48. Rc4 Rd7 49. g3 Rd3 50. Rc6 Rd6 51. Rc4 Rb6, and I think Black wins now. The push of the g-pawn weakened f3, and so Black can play Rb3 with effect in many lines. White should probably play 49. f4+ Kd5 50. Ra4 Re7 51. Kxb2 Re2+ 52. Kc2 Rxg2, and then it looks difficult, as again Black can win 2 pawns, but the end result is not clear. 53. Kd3 Rg3+ 54. Ke2 Rg4 55. Kf3 Rxh4 56. Ra7 Rg4 -- can White hold? All I can say is that I see no win for Black, which is perhaps not the greatest evidence in the world. 
I still haven't looked at whether White can do better by blocking the b-pawn one square earlier.
I still haven't looked at whether White can do better by blocking the b-pawn one square earlier.
Re: How to draw rook endgames
As a side-note, maybe White should not play Ra4 when the Black King is on e6, as Black then plays Ke5, and White is forced to play Rc4. From that configuration, Black can then play Rd6, and after f4+ will attack the Rook via Kd5, gaining a vital tempo. For example, I don't think White can hold this:But if the Rook is on a4 instead, I still can't find a win for Black.
Similarly, the Rook should not move to f4 when the Black King is on d6, as else Kd5, Ra4 Ke5, Rc4 Rd6 as above, and White is in trouble. So there is subtlety.
Similarly, the Rook should not move to f4 when the Black King is on d6, as else Kd5, Ra4 Ke5, Rc4 Rd6 as above, and White is in trouble. So there is subtlety.
-
kingliveson
- Posts: 1388
- Joined: Thu Jun 10, 2010 1:22 am
- Real Name: Franklin Titus
- Location: 28°32'1"N 81°22'33"W
Re: How to draw rook endgames
Starting from the initial position, I can't see how White lost the game, but again, my chess knowledge is not all that. The game was played out using much weaker 32-bit single core vs Houdini on quad core.
PAWN : Knight >> Bishop >> Rook >>Queen
Re: How to draw rook endgames
Same game is being discussed at Talkchess http://www.talkchess.com/forum/viewtopic.php?t=37640,BB+ wrote:From Division 2 of an on-going GM tournament:
but starting move 41.
Uri Blass argues this is won for Black, but computers are not convinced...
Re: How to draw rook endgames
The main problem with comp-comp "roll-outs" is that the whole issue is "knowledge-based" (Black needs to find a plan), and so the extra depth might not be that useful. OTOH, as the endgame manuals will point out, simply having a plan and actually executing it are two different things. This one has a number of tricks, and White has multiple ways of achieving a draw when down 2 (or even 3) pawns. The main theoretical "fact" to know is that Rook+f+h versus Rook is a book draw in many cases (certainly if the White King gets to g1). Navara gave a 40-minute "master-class" impromptu presentation about it, though the analysts will still pick apart where theory and practise diverged.The game was played out using much weaker 32-bit single core vs Houdini on quad core.
http://www.chessvibes.com/reports/rook- ... -wojtaszek
For a Wijk aan Zee example from 3 years back, see Kramnik-Aronian (Black blundered on move 103): http://www.chessgames.com/perl/chessgame?gid=1482242
-
kingliveson
- Posts: 1388
- Joined: Thu Jun 10, 2010 1:22 am
- Real Name: Franklin Titus
- Location: 28°32'1"N 81°22'33"W
Re: How to draw rook endgames
I saw that article on Chessvibe but didn't really pay much attention, so now I will schedule some time and watch the lecture. Looking at the position from your 1st post, Black has a slight advantage, but to me not enough for a win.BB+ wrote:The main problem with comp-comp "roll-outs" is that the whole issue is "knowledge-based" (Black needs to find a plan), and so the extra depth might not be that useful. OTOH, as the endgame manuals will point out, simply having a plan and actually executing it are two different things. This one has a number of tricks, and White has multiple ways of achieving a draw when down 2 (or even 3) pawns. The main theoretical "fact" to know is that Rook+f+h versus Rook is a book draw in many cases (certainly if the White King gets to g1). Navara gave a 40-minute "master-class" impromptu presentation about it, though the analysts will still pick apart where theory and practise diverged.The game was played out using much weaker 32-bit single core vs Houdini on quad core.
http://www.chessvibes.com/reports/rook- ... -wojtaszek
For a Wijk aan Zee example from 3 years back, see Kramnik-Aronian (Black blundered on move 103): http://www.chessgames.com/perl/chessgame?gid=1482242
PAWN : Knight >> Bishop >> Rook >>Queen
Re: How to draw rook endgames
Incidentally, while looking at a side-line from HIARCS-Rybka, the main variation led to Black being up two pawns, but a drawn Rook + a-pawn + c-pawn versus Rook position resulted. Not that TCEC uses 6 piece tablebases though...
Re: How to draw rook endgames
As the game progressed, more and more lines were descending into such a "draw well" (of R+ac versus R). But I don't think either had "special knowledge" of this endgame, so if a draw had occurred in this manner, it would likely have been an accident.Incidentally, while looking at a side-line from HIARCS-Rybka
Well, now it's transformed into something where White could get a fortress draw, I think:
The starting position is that of the game, the line is my PV. Is it obvious how Black can make progress? No targets for the Queen (attacking h2 just leads to Ph4, with the g3 pawn protected by the Rook), both passers blockaded, if the King supports the pawn, both the Rook/Knight guard a3, and the c-pawn remains blocked... Though again, it's not like engines know these things.