|
< Earlier Kibitzing · PAGE 3 OF 3 ·
Later Kibitzing> |
| Jul-15-09 | | visayanbraindoctor: <timhortons> It seems that the White rook goes after the Black pawns. How about exchanging rooks and going for the Black pawns?
click for larger view |
|
| Jul-15-09 | | visayanbraindoctor: <timhortons> Ah OK. Thanks. When I 'meet' him here in CG again, I will ask him. (",) |
|
| Jul-15-09 | | timhortons: ok doc, zanshin is the man on these:)
hes very willing to help im sure...we all run on him on problems like these... |
|
| Jul-15-09 | | visayanbraindoctor: <timhortons> I think your analysis says 29...e3 would draw. It's not that obvious though. If so, White should have tried this line, exchanging rooks. I thought that when Granda went after the Queenside pawns with his Rook, the game became more drawish. In a pure King ending, White has more chances to win. |
|
| Jul-15-09 | | timhortons: <visayanbraindoctor > its goin to end game position now where an engine without table base is useless. i might be right on that.
but let see how zanshin go forward where my engine stop, zanshin has stronger computer than me. i hope percyblaney could check on the position were talkin...his shredder has better power in evaluating endgames... |
|
| Jul-15-09 | | timhortons: vogonprotestnik(IM) whispers: naka really hold the endgame yesterday? miracle man! i pick this up at icc now.... |
|
| Jul-15-09 | | visayanbraindoctor: Since I don't have an engine, I just thought that the position above (in my head) was winning for White. In a real game, it would probably win for White. Even after 29...e3 White could move 30. f4, f5 exchange pawns and he would have a Kingside majority consisting of his g and h pawns against Black's h lone pawn. Black in a real game would be under severe pressure, even if in the case that it's objectively drawish. |
|
| Jul-15-09 | | visayanbraindoctor: <timhortons> I have been running the position above in my head after 30. f4, pretending I am in a tournament with my clock ticking away. Boy but if I had Black in a tournament with this, I would probably lose! White in effect is a pawn up in the Kingside. There are variations were White can eat the e3 pawn at leisure, and still remain a pawn up in the Kingside with a moving pawn majority. There are also variations were White ignores the Black doubled e-pawns and just keeps on pushing his own g ang h pawns. Black sooner or later will have to move his king to the g or h file. White then gobbles up the e-pawns, and then goes after the Black a pawns. The Black King can't defend them because it will have to get rid of the White e pawn first. If Granda tried this line Naka would have been in real trouble. Maybe engines see the White advantage as minimal because in most cases, Black technically remains a pawn up, but those pawns are doubled a-pawns that are in effect like one pawn blocked by White's lone a-pawn. |
|
| Jul-15-09 | | sharkw: timhortons last diagram - just a random (long) variation: 29...e3 30. fxe3 Kd6 31. Kg5 Kc5 32. Kh6 Kb4 33. Kxh7 Ka3 34. Kxg6 Kxa2 35. h4 Kb1 36. h5 a4 37. h6 a3 38. h7 a2 39. h8=Q a1=Q 40. Qxa1+ Kxa1 41. g4 a5 42. Kf7 a4 43. g5 a3 44. g6 a2 45. g7 Kb1 (Black can't avoid the exchange of queens even on b2 because of the check on the long diagonal) 46. g8=Q a1=Q 47. Qg1+ and White is winning the pawn ending. As they say, long variation - wrong variation, but the "dumb" pawn race after 29...e3 appears to be winning for White. Of course Black must have improvements. |
|
| Jul-15-09 | | visayanbraindoctor: <timhortons> Thanks. The lines you showed above were quite helpful. |
|
| Jul-15-09 | | timhortons: < sharkw >
i hope they could soon provide thorough analysis of this game, as mig said in his blog mistake could have happen as both side make a move.... |
|
| Jul-15-09 | | ontocaustic: you don't need an engine for this one to see that white's mistake was bringing the pawn all the way to a7. the way to win that kind of position is to bring it to a6, get king to a7, Rb7, Kb8, a8 queen. |
|
| Jul-15-09 | | acirce: Never pretended to be an endgame wizard, but I really don't understand 35.a7 either. Mig suggests that it was still won by 40.Kf4!, though, in a study-like manner, h4 being a mistake because it gets the pawn too close to Black's king when it turns to a pawn endgame: http://www.chessninja.com/dailydirt... If you <do> play a7 there must be a concrete winning idea behind it, of course (such as this one!) Maybe White thought he had one, but it didn't work. |
|
| Jul-15-09 | | zanshin: <tim> I'm on travel so won't be able to do the analysis. Maybe when I get back. |
|
| Jul-16-09 | | MostlyAverageJoe: < acirce: Never pretended to be an endgame wizard, but I really don't understand 35.a7 either. Mig suggests that it was still won by 40.Kf4!, though> Indeed, it should win. Engine analysis confirms this fairly easily. Below are some principal lines; feeding them into Hiarcs and backsliding (letting the engine run at least 22 plies at each step back) produced the valuations below. 40. Kf4 Rxf2+ 41. Ke5 Ra2 42. Ke6 Ra3 43. Kxe7 <+8.35> Ra6 44. Kd7 Ra1 45. Kc6 Rc1+ 46. Kd5 Ra1 47. Kxe4 The <+8.35> indicated where I stopped the backslide from move 47; the valuation was +8.35. The position is: 
click for larger view40. Kf4 Rxf2+ 41. Ke5 <+7.38> Ra2 42. Ke6 g5 43. h3 Ra6+ 44. Kxe7 Kh7 45. Kd7 Ra2 46. Kc7 Kg7 47. Kb6 Rb2+ 48. Kc5 Ra2 49. Kd4 Ra4+ 50. Ke5 Kh7 51. Kf5 Kg7 52. Kxg5 Ra3 The position after 41. Ke5 is: 
click for larger viewThe above lines refute 40...Rxf2. Some other possibilities are more complex and require much more effort, but invariably lead to white's win. For example: 40. Kf4 Ra4 41. h3 Kh7 42. Ke5 <+4.61> Kg7 43. Ke6 <+9.70> Ra2 44. Kxe7 Ra3 45. Kd6 Ra2 46. Ke5 Ra4 47. Kd5 Kh7 48. Kc5 Ra6 49. Kb5 Ra2 50. Kb6 Rb2+ 51. Kc6 Rc2+ 52. Kd5 Ra2 53. Kxe4 The position after 43.Ke6 is: 
click for larger viewThe basic game plan for the white is to win one or both of the e-pawn(s), exploiting the fact that black K cannot move from 7th rank (otherwise a check with a tempo wins immediately), and that the black R has to guard the a-file and the e4 pawn. If the backward e-pawn is won, the white needs to arrive at this kind of position: 
click for larger view The above evaluates as mate in 25, with the threat of Rc8 (tablebases are needed for the engine to see the mate). If the forward pawn is won, then the pair of white pawns becomes very dangerous. Sample line below: 40. Kf4 Ra4 41. h3 Kh7 42. Ke5 Ra6 43. Kxe4 <+5.18> Kg7 44. f4 Ra5 45. Kd3 Ra3+ 46. Kd4 Ra4+ 47. Ke5 Kf7 48. e4 Ra3 49. Kd5 Ra5+ 50. Kc6 Ra1 51. f5 gxf5 52. exf5 Kg7 53. h4 Kf7 54. h5 Rc1+ 55. Kd5 Ra1 56. h6 Ra5+ 57. Kd4 Ra4+ 58. Kc |
|
| Jul-16-09 | | twinlark: <MAJ>
I use the back sliding technique as well. When I feed the game moves into the engine, I usually allow a couple of minutes between each move, even say a 14 or 15 ply evaluation per forward move (even obvious moves), that add extremely useful hash tables that bolster reverse slide analysis. |
|
| Jul-16-09 | | visayanbraindoctor: <MostlyAverageJoe> Thanks for the analysis. This endgame bothered me because when I was watching it, I knew I would do what Granda initially did, which is to transpose into this ending. I fully expected a 1 - 0, and must first reaction when it turned into a draw was that White made mistakes because of time pressure. If the ending could not be won, then I suspect there is something seriously wrong with my understanding of Rook endings and King endings. |
|
| Jul-16-09 | | percyblakeney: So there was a win there after all, I can understand that Granda missed it though, by then he was basically playing with only the increments, and the winning method does look rather study-like. |
|
| Jul-16-09 | | MostlyAverageJoe: <twinlark> For the forward lines I do something similar to what you described, because with the engines I use, the infinity analysis gives reliable moves up to 25% or so of the claimed depth (I trust no more than 5 plies out of 20-ply forward analysis). I usually do 18 plies on obvious moves and 20 or more on non-obvious (the additional criterion for deciding on a non-obvious move is that the top choice should be stable and not change with increasing depth). |
|
| Jul-16-09 | | twinlark: <MAJ>
<(I trust no more than 5 plies out of 20-ply forward analysis)> I'm less trusting of my engine; I tend to slide ply by ply, especially when there're two or more lines that are closely rated. I've come across a few games where the top dozen options for the ply are separated by less than 40 centis evaluation. <I usually do 18 plies on obvious moves and 20 or more on non-obvious > Is this during the initial phase when you're feeding the machine with the game moves, or is it part of a forward re-slide, so to speak? Are you using a single core or a multi-core processor? |
|
| Jul-16-09 | | MostlyAverageJoe: < twinlark: ... Is this during the initial phase when you're feeding the machine with the game moves ... > Starting from the position about 3 moves before the one to analyze, on every move. < Are you using a single core or a multi-core processor?> 8-core, 3GHz. Alas, 32-bit engine, so only 4GB of RAM is usable. |
|
| Jul-22-09 | | zanshin: <visayanbraindoctor: In the position after 24. Rxc4 Rc8 .. could not White have gone straight away for the Black Kingside pawns> <VBD> I might be late for this party, but here's my run. Let me know if you want a follow-up. 
click for larger view [+1.39] d=25 25.c7 Ke5 26.Rc6 a4 27.Kg2 a3 28.Rxa6 Rxc7 29.Rxa3 Kd5 30.e3 e6 31.Ra5 Kc4 32.Re5 Kd3 33.a4 Rc6 34.a5 Ra6 35.g4 Ra7 36.Kg3 Ra6 (0:07.43) 52374kN |
|
| Aug-03-09 | | visayanbraindoctor: <zanshin, MostlyAverageJoe> Thanks for the analysis. |
|
| Nov-30-12 | | wordfunph: "I was considering playing ...a6 on move 9 and I ended up playing ...b6. But if I had played ...a6, which was the correct move, I would have been simply better in the game." - Hikaru Nakamura
Source: NIC 2009 #6 |
|
| Nov-30-12 | | fokers13: 34..g5 for black would be a big improvement over the text apparently(perhaps with the objective of locking down the kingside after an eventual g4). |
|
 |
|
|
< Earlier Kibitzing · PAGE 3 OF 3 ·
Later Kibitzing> |
Kibitzing
