< Earlier Kibitzing · PAGE 6 OF 7 ·
Later Kibitzing> 
Jan2313   morfishine: Bakh to the drawing board 

Mar3113   zebekias: The Qg7 "refutation" seems to complicate things, but white can still force the win. Houdini 3 can "see" all the winning variations for white on my 4 core pc after about half an hour of analysing 22 billion positions! Respect to Gusev for calculating this beautiful attack back in 1946! 

Apr0813
  technical draw: Great game. First time I've seen it. 

Apr2113   qqdos: <zebekias> thanks. I think there must have been a sizeable dollop of intuition and nerve in Gusev's play. A breathtaking game and the greatest admiration for Gusev! How much does the mighty Houdini "see" before and after the Qxe5 sacrifice?? 

Jun1513
  hedgeh0g: Games like this are the reason I play chess. 

Nov2413   Karpova: As Dan Scoones says in C.N. 8405, Black was not Averbach but <E. Auerbach>. Link: http://www.chesshistory.com/winter/... 

Feb1814
  AylerKupp: A great game and a magnificent conception by Gusev but ... (part 1 of 4) This is a great game and a great conception by Gusev. But 24.Qxe5, regardless of how spectacular it looks, appears to lead to no more than equality and a likely draw.
click for larger viewAfter 24...fxe5 25.Rf1, Stockfish DD at d=45 gives the following line, evaluating the resulting position at [0.73]: 25...a6 (an interesting move, discouraging the immediate 26.Bd5 and maybe gaining a tempo, but Stockfish evaluates 25...Rc7 and 25...Rc8 at [0.00]) 26.Bd1 Rc7 27.e7 27...Rxe7 28.Bb3+ Rf7 29.a4 Qd7 30.h3 Qa7 31.Kg2 Qe7 32.Kh2 Qc7 33.Kg2 Qb7 34.Kg3 a5 35.Kh2 Qe7 36.Kg2 Qa7 37.Bc4 Qb7 38.Kh2 Qe7 39.Bb3 Qd7 40.Kg2 Qe7 41.Kg3 Qc7 42.Kg2 Qa7
click for larger viewBlack can continue the blockade on the 7th rank forever and White can't break through. And, of course, any attempt to force the issue by Bxf7+, ...Qxf7, Rxf7, ...Kxf7 probably loses for White. If instead 27.Bb3, 27...Re7 and White again can't break through the blockade on the 7th rank. At d=42 Stockfish gives 28.Kg2 a5 29.Bd5 a4 30.Bc6 Qb8 31.Bd5 Qe8 and a draw by repetition. And 27...Re7 28.Rf7 Rxf7 29.exf7+ Qxf7 30.Bxf7+ Kxf7 leads to a probably lost position for White since Black's Rh8 gets out so Black is the exchange up although the win is not simple.
click for larger viewBut not 27...Qe7 since then 28.Rf8+ Qxf8 29.e7+ Rc4 30.exf8Q#. Or 29...Qf7 30.e8=Q# (or 30.e8=R# if you want to be perverse) Houdini 1.5a gives the following line at d=32, evaluating the resulting position at [0.00] showing what could happen if 25.Bxa6: 25...a6 26.Bxa6 Qe7 27.Bb7 Rd6 28.Bd5 Rxd5 29.exd5 Qc5+ 30.Kg2 Qxd5+ 31.Kg3 Qd3+ 32.Rf3 Qd6 33.Rf7 e4+ 34.Kg2 Qd8 35.Rg7+ Kf8 36.Rxh7+ Kg8 37.Rg7+ Kf8 38.Rh7+ Kg8 and a draw by repetition.
click for larger viewAnd Houdini also evaluated 25...Rc7 and 25...Qe7 at [0.00], also draws by repetition. 

Feb1814
  AylerKupp: A great game and a magnificent conception by Gusev but ... (part 2 of 4) Black's last mistake was apparently 29...a5 instead of 29...cxb3, allowing 30.bxc4. Houdini evaluated 29...a5 at [+3.17], d=31 after 30.bxc4 Qe7 31.Kg2 Qg7 (this strange looking move has its points, particularly when compared with the alternatives, since it picks up the bishop and frees up the Rh8) 32.Bxg7 Kxg7 33.Rf7+ Kh6 34.e7 Re8 35.Kf3 (the immediate 35.c5 seems quicker) 35...g5 36.Rf6+ Kg7 37.Re6 Kf7 38.Rxe5 Kf6 39.Rf5+ Kg6 40.Rxa5 Rxe7 (with a 3pawn advantage White should win, regardless of how weak the pawns look) 41.a4 h5 42.Rb5 Rf7+ 43.Ke3 h4 44.a5 g4 45.a6 g3 46.hxg3 h3 47.c5 h2 48.Rb1 Ra7 49.Rh1 Rxa6 50.Rxh2 (ditto for a 4pawn advantage) 50...Kf7 51.Rh7+ Ke6 52.Rh6+ Ke5 (one last trick, 53.Rxa6 stalemate!) 53.Rh5+ Ke6 54.g4 Ra3 55.Kd4
click for larger viewInstead, 29...cxb3 allows 30.axb3 a6 31.Kg2 a5 32.Rf2 Qe7 33.Rf3 Qe8 34.Rf1 g5 (this seems unnecessarily risky unless Black is trying for a win, but in some lines preventing the bishop from attacking d2 and allowing ...Qd2+ is required (see below). Still, if Black is was satisfied with a draw then after 34...Qe7 (still defending f7) is more straightforward since it's difficult to see how White could make progress) 35.Rf5 Qd8 36.Rf7 Qd2+ 37.Rf2 Qd6 38.Rf7 Qd2+ 39.Rf2 and another draw by repetition. (Houdini, d=30). Stockfish gives a similar line at d=45, and indicates that 30...a5 and 30...Qe7 also draw. But what about 30.axb3 a6 31.Kg2 a5 32.c4 Qe7 33.Rf2? Isn't Black in zugswang? Protecting f8 by 33...Qe8 allows 34.c5 and the advance of the cpawn. Protecting f8 and staying in the a3f8 diagonal to block the advance of the c and epawns by 33...Qd6, 33...Qc5, or 33...Qb4 allows 34.Rf7 with 35.Rg7+ threatened so that 35...Kf8 a rook move attacking the queen will pick it up since Black's king is in check. But after 33...Qa3 34.Rf7, Black has 34...Qb2+ (34...Qa2+ works just as well) and White cannot escape the either the draw by repetition by 35.Rf2 Qa3 36.Rf7 Qb2+ etc. or the perpetual check: 35.Kg1 Qd4+ 36.Kf1 Qd1+ 37.Kf2 Qc2+ 38.Kg1 Qd1+ 39.Kf2 Qc2+ etc. 35.Kf1 results in a similar sequence (Houdini, d=28). Stockfish at d=44 has similar lines, except starting with 35...Qa1+. 

Feb1814
  AylerKupp: A great game and a magnificent conception by Gusev but ... (part 3 of 4) I frankly don't understand why 29...cxb3 30.axb3 allowed Black to draw while 29...a5 30.bxc4 allowed White to win, and try as I can I still don't know.. True, after 30.bx4 White has an extra pawn, but his passed pawns are doubled and isolated, while after 30.axb3 White's passed pawn is (potentially) protected. And in either case Black has control of the a3f8 diagonal to block both the epawn and the cpawn(s) from advancing. And having an extra pawn in this position doesn't seem to me that it would be the deciding factor. Position after 29...a5 30.bxc4
click for larger viewPosition after 29...cxb3 30.axb3
click for larger viewOne difference is that after 29...a5 30.bxc4 Qe7 31.Kg2 Qa3 32.Rf7 Qb2+ 33.Kh3 Qxc3+ 34.Kg4 Black is out of checks, and attempting to get back to the a3f8 diagonal by 34...Qa3 fails to 35.Rc7, and now 36.Rc8+ is only stoppable by 36...Qf8, and Houdini at d=26 announces mate in 27 after 35.Rc7. Stockfish at d=37 has similar lines with 29...a5 30.bxc4 Qe7 31.Kg2 Qa3 32.Ra7. And if 29...a5 30.bxc4 Qe7 31.Kg2 Qe8 (maintaining contact with f7) then 32.c4 seals Black's fate since he can't stop all 3 of White's threats; c4c5c6c7c8=Q+, e6e7e8=Q+, and Rf8#. But there are apparently no general principles which can be applied, it just depends on the specific lines. For example, per Stockfish: (a) 30.axb3 Qe7 31.Kg2 Qe8 32.Kg3 Qe7 33.c4 Qe8 34.c5 and White wins. (b) 30.axb3 Qe7 31.Kg2 Qe8 32.Kg3 Qe7 33.c4 and White wins. (c) 30.axb3 Qe7 31.Kg2 Qe8 32.Kg3 and White wins after 32...Qe7 or 32...g5, but only draws after 32...a5. 

Feb1814
  AylerKupp: A great game and a magnificent conception by Gusev but ... (part 4 of 4) Still, there were several points that both sides needed to remember: 1. After 29...cxb3 30.axb3 Black's queen is tied to preventing Rf8# as well as the advance of the c and epawns. But White's pieces are equally tied up in threatening mate and preventing Black's Rhh8 from escaping since, if it does, Black's material superiority (Q vs. B) will win. So White's Bh6 is effectively just at immobile as Black's Rh8 (particularly if Black plays ...g5) and White's Rf1 is restricted to moving along the ffile, otherwise ...Kg7 (if the bishop moves) or ...Qxe6 (if the rook moves from the ffile). 2. When Black's queen is on e7 or e8, Rf7 is not an option since the simple ...Qxf7, exf7+, ...Kxf7 will result in a likely won position for Black, and certainly no worse than a draw. 3. In some lines Black must play ...g5 at the appropriate time to block the bishop from defending d2. This allows Black's queen to penetrate by ...Qe7d6d2+ or ...Qe7b4d2+. And once Black's queen penetrates, then White has to either forego Rf7 and allow a draw by repetition or allow a perpetual check after a king move. So, as much as I admire Gusev's concept, 24.Qxe5 can perhaps be considered a mild swindle attempt (nothing wrong with that!) since after either 24.Qa3 or 24.Qd4 both Stockfish (d=33) and Houdini (d=28) both consider the position even. But it is not a forced win by White. Still, 24.Qxe5 represents White's best practical chance and does not involve any risk. The odds were that Black would make a mistake somewhere down the line and he did. I wish someone had taken a picture of Black's face after White played 24.Qxe5! 

Mar3014   zebekias: AylerKupp, I had spent days on this position and I have no doubt white forces the win. I have the main variations at: http://www.mockbites.com/articles/p... The engines can not "see" the forced win immediately after Qxe5. You have to progress a bit down the line, and eventually the evaluation changes from 0 to winning for white. 

Apr1414   pikuels: <AylerKupp>, apparently the difference between 29...cxb3 30.axb3 and 29...a5 30.bxc4 is that the latter keeps a white pawn on a2, which prevents the black queen from checking from b3, which seems to be the only check to force a draw. 

Apr1414   pikuels: Besides, in your first post (1 of 4) if the game continues: 24... fxe5
25.Rf1 a6
26.Bd1 Rc7
27.Bb3 Re7
white wins. All he has to do is get a queenside passed pawn. When it reaches the 7th rank black can't capture it, and then white plays Rf7Rxe7. I still have to go through your other posts, thanks for your input anyway. 

Apr1414   pikuels: Sorry, in my first post I meant:
"apparently the difference between 29...cxb3 30.axb3 and 29...a5 30.bxc4 is that the latter keeps a white pawn on c3, which prevents the black queen from checking from b3, which seems to be the only check to force a draw." So the only way for black to draw is to exchange his rook for white's bishop, then place his queen on e7, play cxb3 and move his queen from e7 to a3 and back. It's a narrow road, but black gets a perpetual. All other lines win for white. A spectacular sacrifice, anyway, and the best practical chance white had to get the win. One of my all time favourite games. There's also a lot of analysis, by myself and some other users, in previous pages. 

Dec0414
  whiteshark: Edward Winter's chess note, mentioned by <Karpova>, can be found here now: http://www.chesshistory.com/winter/... > <8405. Gusev v Auerbach> 

Mar1215   FairyPromotion: This is an immortal above immortals. One of the greatest games of all time, underrated mostly because of the name of the winner. I remember when I first saw this game, I had thought that 33... g5 was the losing move. Later, when I read the analysis for this game, I was amazed to find out that 24. Qxe5 was winning in almost all (some claim all) variations, and the drawing lines suggested in some analysis are extremely difficult to find. Keep in mind this is a game where white sacrificed his queen, when he was already down a whole rook, without any immediate mate or material gain in the cards. But with rook on h8 prisoned, white was able to overload the black queen, and eventually win the game. Amazing! 

Nov1815   NeverAgain: Worry not, folks: the brilliant queen sacrifice was sound and White wins in <every> variation. The jumpoff position arises after 24.Qxe5!! fxe5 25.Rf1!
click for larger viewBlack has several plausible replies  <25...Rc8>, <25...Rc7>, <25...a6>, <25...Qe7> (we can dismiss <25...Rxe6> at once as it leads to mate in three at most after <26.Bc4 any move 27.Bxe6>)  and there are many variations that can spring from them. However, I'm going to demonstrate that whatever Black plays White can always follow a simple winning plan that doesn't require calculating endless variations or even using an engine. Nevertheless, I will provide a few variations backed up by engine evals to point out where AylerKupp's "refutations" fail. The key idea of this plan involves the transfer of the bishop to b3. Once it's there the threat of <e7> with check or checkmate will always hang over Black's head. Therefore the black rook will have to either blockade the e6 pawn on the e7 square or block the bishop's diagonal (b3g8). The queen is a poor blockader to begin with, and here it also has to guard the f8 square against <Rf8> with check or checkmate. In the former case White wins with a pawn breakthrough on the queenside. In the latter he waits until Black runs out of noncommital queenside pawn moves (he may choose to blockade them to speed up the process), plays his king to a different square to avoid checks by the black queen, and then as soon as the queen leaves the e8 square <Rf7> will decide. There are a couple of additional interesting twists there that I'll cover as we go. I cannot claim the authorship here, most of these ideas were outlined years ago on earlier pages of this thread, in the TalkChess thread linked there http://talkchess.com/forum/viewtopi... and in Vass' post on ChessPub here http://www.chesspub.com/cgibin/che... 

Nov1815   NeverAgain: click for larger viewThe two main points of AK's analysis are that Black draws after <25...a6> and in the game continuation after <29...cxb3>. <AK: Stockfish DD at d=45 gives the following line, evaluating the resulting position at [0.73]: 25...a6 (an interesting move, discouraging the immediate 26.Bd5 and maybe gaining a tempo, but Stockfish evaluates 25...Rc7 and 25...Rc8 at [0.00]) 26.Bd1 Rc7 27.e7> Two errors here already. First, it is pointless to rely on engines' eval this early into the analysis. Even the last nonLazy Stockfish dev build (Oct 16) with 6men Syzygy tablebases (which AK could not take advantage of in 2014) cannot see a win at this point. No, the correct strategy is to do oldfashioned manual analysis yourself, find a win and then work your way backwards with the engine of your choice. Since AK is well familiar with the "forward sliding" technique, I'm surprised he didn't use it here. Anyway, around move 30 the tablebases start to provide a major boost, with engines quickly reaching search depths of 60+  then SF finally wakes up and smells the juicy aroma of fried chicken emanating from Black's position. All the same, I found time and time again confirmations of how dumb SF is, always seeking drawn deadends, so most of the time I stuck to Houdini 4 Pro in Tactical Mode. Second, <27.e7> is an exceedingly dumb move that throws a linchpin pawn for no good reason, so there is no point in analyzing it in detail. <AK: If instead 27.Bb3, 27...Re7 and White again can't break through the blockade on the 7th rank. At d=42 Stockfish gives 28.Kg2 a5 29.Bd5 a4 30.Bc6 Qb8 31.Bd5 Qe8 and a draw by repetition.> Well, if you see SF start playing moves like <30.Bc6> and <31.Bd5> it can mean it's out of ideas, not that the position is a draw. I've seen quite a few games where the winning side repeats moves a few times in time trouble to gain time on the clock. Guess what: after the first repetition both SF and Komodo will give the = (0.00) eval, even if one side has a crushing advantage. The winning continuation runs as follows:
25...a6 26.Bd1 Rc7 (<26...Rc8>, <26...Rc4> or <26...b5> will transpose into the game continuation) 27.Bb3 Re7 28.Bd5!
click for larger view
And now whatever Black plays, it always comes down to the same thing. Note that the e7 Rook immobile  if it moves, it's checkmate in two at most  so Black only has moves by the queenside pawns or the queen at his disposal. a) 28...a5 29.a4!
Now Black can only move the queen along the eighth rank 29...Qd8
<29...Qc8 30.Rf7> and Black has only one check. 30.Rf7 Qe8 31.Rxe7 Qxe7 32.Kh1!
This must be the move that fell beyond Stockfish's horizon when it evaluated 25...a6 as "=". Anyway, at this point SF sheepishly acknowledges that it's mate in six. b) 28...b5 29.Kg2 a5 30.a3 a4 31.Rf3
Suddenly SF sees mate in eight. Since queen moves along the back rank fail the same way as in a), it goes for broke just to postpone the inevitable 31...b4 32.cxb4 Qc8 33.Rf7 Qc2+ 34.Kg3 Qd3+ 35.Rf3
The checks come to an end, and a retreat by the queen is out of the question because of <36.Rf8#>, so b1) 35...Qb3 36.Bxb3 Re8 37.Rf8+ Rxf8 38.e7+ axb3 39.exf8Q# b2) 35...Qxf3+ 36.Kxf3 g5▢ 37.Kg4 and mate in two. c) 28...Qd8 (or 28...Qb8) 29.Rf7 Qe8 30.Rxe7 and White wins like in a) d) 28...Qc8 29.Rf7 Qc5+ 30.Kh1 is mate in four.
As you see, the same basic winning plan works for any variation where Black attempts to blockade the e7 pawn with the rook. 

Nov1815   NeverAgain: Now let's turn to AK's claim that the game continuation <25...Rc8> draws. He cites <29...cxb3> as a critical improvement and I won't argue there, because White, in his turn, can improve earlier with <28.a4!>, as Vass of ChessPub discovered in 2011. The text position after 27...b5:
click for larger view
28.a4! a6▢ 29.axb5 axb5▢ and now White can play 30.Ba2! straight away: a) 30...Qa8 31.b4 Qa7+ 32.Rf2 and SF finally arrives at the eval a1) 32...Qa8 33.e7 is mate in two
a2) 32...g5 (you can tell SF is out of ideas again) 33.Kg2 and SF sees nothing better than going into a lost ending: a21) 33...Qa8 34.Bxc4 Qxe4+ 35.Kg1 Qg4+ 36.Rg2 Qd1+ (<36...Qxg2+ 37.Kxg2 bxc4 38.b5> and White mates in three) 37.Bf1 and 38.Rxg5+ a22) 33...Qg7 34.Bxg7 Kxg7 35.Rf7+ Kg6 36.Bxc4 bxc4 37.Rf5 b) 30...Qe7 31.b4 and now
b1) 31...Qa7+ and White wins as in a)
b2) 31...Qe8 32.Bxc4 bxc4 33.b5 and mate in 13
b3) 31...Rc7 32.Rf8+ Qxf8 33.e7+ and mate next move
b4) 32...Rc8 32.Rf7 breaking the blockade of the e6 pawn and forcing mate in 10. So, following the queen sacrifice White wins in all variations. QED. 

Nov2815   NeverAgain: Also see this post http://talkchess.com/forum/viewtopi... on Talkchess and the next two. 

Jan1416   scholes: this game is work of art 

Jan1416   talwnbe4: I checked this with Stockfish.. it can't make sense of this one.. this is one of those great positions and looking at Neveragain's lines checking with Stockfish it seems to me it's a forced win after 24.Qxe5!! I'm using a Core 2 quad 2.4 GHz. My engine suggested 25.. Rc7 which runs into the Bd1 line, a move that Stockfish can't find. e.g, 25..Rc7 26. Bd1 b5 27. a4 Rg7 28. Bb3 bxa4 29. Bd5 Qe7 30. Kg2 a6 31. c4 g5 32. c5 winning or.. 25..Rc7 26. Bd1 b5 27. a4 Rg7 28. Bb3 bxa4 29. Bd5 Re7 30. Kg2 a3 31. bxa3 a6 32. c4! Qb8 33. c5 even here SF has problems spotting 32. c4 or.. 25..Rc7 26. Bd1 b5 27. a4 Rg7 28. Bb3 bxa4 29. Bd5 Re7 30. Kg2 a5 31. c4 Qb8 32. c5 (same idea) Qe8 33. c6 winning 

Jan1416   talwnbe4: after 25..Rc7 26. Bd1 b5 27. a4 Rg7 28. Bb3 bxa4 29. Bd5 Re7 30. Kg2 a5 SF5 can't find 31. c4 after 11 minutes and using 3 cores of a Core 2 quad 2.4 Ghz.. it suggests 31. Rf2? d=51 1.7 billion nodes 

Jan1416   talwnbe4: talwnbe4: Stockfish 7 fares a bit better, but again is confused by the same kind of position. 

Sep0716   clement41: delightful to watch; reminds me of the unknown crazy promotion game C Lelievre vs M Francois, 2015 



< Earlier Kibitzing · PAGE 6 OF 7 ·
Later Kibitzing> 


