< Earlier Kibitzing · PAGE 18 OF 18 ·
Later Kibitzing> 
May2614   RookFile: Maybe a year after this game was played, Spassky and Geller tested this theory in a match and Spassky won easily. Fischer would have done the same. What is the issue? Stamina, for one thing. At that age, Geller was dangerous for one game, but in a match, stamina becomes vitally important. 

May2614
  Domdaniel: <Isilimela> Eh? That's been possible for years... 

May2614
  AylerKupp: <<offramp> I'd simply say that if Geller thought 20.Qf4 would have won 15 years later, then it would've won on the day.> Consider that this was a complex game full of tactical continuations where computers excel. And Geller likely did not have computers to assist him in his analysis and, if he did, they were not all that good in 1983. So, in spite of "many years of painstaking analysis", I do not believe that anyone at that time could have analyzed the position after 20.Qf4 as thoroughly as multiple engines can do in the present day. Geller himself says that " A calculation of all the variations was impossible". Fortunately "The Application of Chess Theory" is available for download here, and I am looking forward to reading it: https://archive.org/details/The_App.... For those who are interested, here is the analysis contained in that book for this game after 19...Nxe4: After 20.Qf4!! cxb2 21.Rh5!
(a) 21...Bf6 22.Qf5 h6 23.Rxh6+!! gxh6 24.Qg6, with inevitable mate. (b) 21...Nf6 22.Rh6!, when Black can avoid the thematic Rxf6 only by going into a hopeless ending: 22...d5 23.Rxf6 Rd6 24.Rxd6 Qxd6 25.Qxd6 Bxd6 26.Rf1, and 26...Bxh2 fails to 27.Bc5, while the threat of 27.Rf6! is maintained (Murey). (c) 21...Nc3 + 22.Kxb2 Nxd1 + (22...Rxf7 23.Qxf7 Nxd1+ 24.Kb1!! Qxd4 25.Rxh7+! Kxh7 26.Qh5#) 23.Kc1 Rxf7 24.Bxf7!, with the decisive threat of 25.Rxh7 + Kxh7 26.Qf5+. White also wins after 20...Nd2+ (his task is very simple after 20...d5 21.Qe5 Nf6 22.Rxf6 Bxf6 23.Qxf6!) 21.Rxd2 cxd2 22.c3!!, and: (c1) 22...Qb7 (or 22...Qxb3), and Black succumbs to the mating combination 23.Bxg7+! Kxg7 24.Qg4 + Kh8 25.Qd4+ . (c2) 22 ...Qc5 (best, although it does not get Black out of his difficulties) 23.Kc2! Bd7 (23...Qe5!? 24.Rxe5 dxe5 25.Qxe5 Bf6, with the hope of 26.Qxf6? d1=Q+!, leads to a hopeless position after the correct 26.Qc5 Bxd4 27.cxd4 Rc8 28.Kxd2 Ba4 29.Qe7 Bxb3 30.axb3 Rcd8 31.Ke3) 24.Bxc5 Bxf5+ 25.Qxf5 dxc5 26.Kd1 Rd6 27.Qe5 Rd7 28.Qe6 Rb7 29.Bc2, and against 30.Qf5 there is no satisfactory defence. I took the liberty of relabeling Geller's second set of (a) and (b) lines as (c1) and (c2) since they are alternatives to line (c) and so I think that this labeling is more consistent. I've also eliminated some extra spaces in front of " !" and " ! !", added periods after the move numbers for White, and done some other minor formatting cleanup of the text. And, of course, I will take a look at Geller's analysis with the help of my silicon friends when time permits. 

May2614
  AylerKupp: <<Isilimela> How hard would it be to make the computer analysis live: ie so you can click on it and play through via a board interface? It's just too much to follow in your head or play through on an ordinary board.> Do you find it difficult to follow the computer analysis in your head or on an ordinary board? I find it impossible! You could copy the lines and paste them to a PGN viewer which is what I think <Domdaniel> suggested. Or, I don't know if this meets your "live" criteria, but over the last couple of years I developed a spreadsheet to record analyses generated during Team games and I recently upgraded it to incorporate a viewer. You can download the spreadsheet from here: http://www.mediafire.com/view/5724a...(Black)_Detailed_Analysis.xls It contains one tab labeled 'B020' containing all the computer analyses I posted (and some that I didn't) and one tab labeled 'Geller' which contains the analyses from Geller's book "The Application of Chess Theory" about this game. You will need Excel 2003 or later. It may work on earlier versions of Excel but I haven't tested it with anything earlier than Excel 2003. It has many features and they are boringly described in the 'Notes' tab. Or, if you wisely want to skip the reading, just go to the 'B020' tab and click on a cell (any cell) of the row containing the line that you are interested in looking at and press Ctrl+Shift+V ("V" for Viewer). A PGNlike viewer will popup and you can follow the moves like you do with the viewer on this site. When you want to look at another line, close the viewer popup, select a cell in another line, and press Ctrl+Shift+V again. 

May2714   Howard: Maybe I should have kept my big mouth shut back on May 19.... 

May2714
  AylerKupp: <Howard> What would have been the fun in that? 

May2714   RookFile: Whatever the objective merits of this game are, this is about as courageous an effort as it gets from Geller. 

May2714
  offramp: <RookFile: Whatever the objective merits of this game are, this is about as courageous an effort as it gets from Geller.> This is a fair game from Geller. He played much better games against better players. 

May2714   Petrosianic: It's a gutsy game, but for guts, I don't think Geller ever had more than in this game: Fischer vs Geller, 1967
Fischer eschewed his usual 6. Bc4, which Geller was obviously prepped for. So Geller topped Fischer's surprise with an even bigger one, playing the Poisoned Pawn Variation, a variation which Fischer specialized in, but Geller didn't, and playing it with no advance prep. And actually got away with it. 

Jun0114
  AylerKupp: Commentary on Geller's Analysis of 20.Qf4 – Summary (part 1 of 2). Given that Geller analyzed this game in his "The Application of Chess Theory", translated from Russian into English in 1984, and which can be downloaded from here: https://archive.org/details/The_App..., I felt that it was essential that any further analysis of this game needed to look at Geller's analyses and update them as necessary to try to determine if Fischer did indeed have a won game after 20.Qf4 (a move to which Geller gave two "!"s), as Geller concluded. Here is the position had White played 20.Qf4 instead of than 20.a3:
click for larger viewWhen you count variations and subvariations, Geller provided 8 analyses in his book. Here are his analyses in canonical form (all the moves after 20.Qf4 listed) and in ATSS (Analysis tree spreadsheet) Move order, with Geller’s comments and the position at the end of each of his analysis: (1) 20...cxb2 21.Rh5 Bf6 22.Qf5 h6 23.Rxh6+!! gxh6 24.Qg6, with inevitable mate.
click for larger view(2) 20...cxb2 21.Rh5 Nc3+ 22.Kxb2 Nxd1+ 23.Kcl Rxf7 24.Bxf7!, with the decisive threat of 25.Rxh7+ Kxh7 26.Qf5+.
click for larger view(3) 20...cxb2 21.Rh5 Nc3+ 22.Kxb2 Rxf7 23 Qxf7 Nxd1+ 24 Kb1 Qxd4 25.Rxh7+ Kxh7 26.Qh5#
click for larger view(4) 20...cxb2 21.Rh5 Nf6 22.Rh6!, when Black can avoid the thematic Rxf6 only by going into a hopeless ending: 22...d5 23.Rxf6 Rd6 24.Rxd6 Qxd6 25.Qxd6 Bxd6 26.Rf1, and 26...Bxh2 fails to 27.Bc5, while the threat of 27.Rf6! is maintained (Murey).
click for larger view(5) 20...d5 21.Qe5 Nf6 22.Rxf6 Bxf6 23.Qxf6! and his task is very simple.
click for larger view(6) 20...Nd2+ 21.Rxd2 cxd2 22.c3!! Qb7 23.Bxg7+! Kxg7 24.Qg4+ Kh8 25.Qd4+
. click for larger view(7) 20...Nd2+ 21.Rxd2 cxd2 22.c3!! Qc5 (best, although it does not get Black out of his difficulties) 23.Kc2! Bd7 24.Bxc5 Bxf5+ 25.Qxf5 dxc5 26.Kd l Rd6 27.Qe5 Rd7 28.Qe6 Rb7 29.Bc2, and against 30.Qf5 there is no satisfactory defence.
click for larger view(8) 20...Nd2+ 21.Rxd2 cxd2 22.c3!! Qc5 23.Kc2!! Qe5!? 24.Rxe5 dxe5 25.Qxe5 Bf6 (with the hope of 26.Qxf6? d1=Q+!) leads to a hopeless position after the correct 26.Qc5 Bxd4 27.cxd4 Rc8 28.Kxd2 Ba4 29.Qe7 Bxb3 30.axb3 Rcd8 31.Ke3.
click for larger view 

Jun0114
  AylerKupp: Commentary on Geller's Analysis of 20.Qf4 – Summary (part 2 of 2). The engine analyses validated many of Geller's conclusions, but put others in doubt since he did not always suggest what the engines considered to be Black's best moves, which would have either led to a draw or made life much more difficult for White (and hence provided more practical chances for a draw). This is not meant as a criticism of Geller; this was an incredibly complex game and so was the followup of "the game that might have been". And, as he himself said, "A calculation of all the variations was impossible" (although I think that he was referring to Fischer and not to himself!). Geller simply didn't have the tools (computers + strong chess engines) at his disposal in 1983 that we have today when he generated his analysis. So Geller's conclusions (like probably most analysis conclusions, including these), in spite of being "established after many years of painstaking analysis" are simply not "the truth" as he claimed. In particular, Geller's conclusion Black has no way of equalizing, is correct; White retains the advantage in practically all lines but it's unclear whether White has a forced win. You can decide for yourself whether Geller's conclusions or the engines' conclusions are correct. In all cases I had the engines (usually two or more) analyze both Geller's lines and alternate lines that the engines considered to be better. Four of Geller’s analyses included 20...cxb2 as Black's response to 20.Qf4, one included 20...d5, and three included 20...Nd2+. All 6 of the engines I used considered 20...cxb2 to be Black's best response to 20.Qf4, with an average evaluation for the 6 analyses of [+1.21] at an average depth of 28 ply. Clearly advantageous for White but not necessarily winning. So in my followup engine analysis I focused mostly on Geller's lines including 20...cxb2, since this was clearly Black's best response. I didn't bother to look very deeply into the four analyses which included 20...d5 and 20...Nd2+ since all 6 engines confirmed that these two continuations are inferior to 20...cxb2, with an average evaluation for the 6 analyses of each move of [+3.79] and [+3.52] respectively, with same average search depth of 28 ply. After 20.Qf4 cxb2 21.Rh5 Geller included one analysis of the inferior 21...Bf6, two analysis of the workable 21...Nc3+, and only one analysis to 21...Nf6, which the engines considered to be the best response to 21.Rh5. And in his analysis of 21...Nf6 22.Rh6 Geller did not consider 22...Rxf7 which looks might lead to a survivable game for Black. Conclusion (mine):
20.Qf4 gives White the advantage but perhaps not enough to win. Black's best approach to try to hold the game is the mostlyforced <DWINS> line and its variations where unbalanced positions are often reached which seem to provide White with a winning advantage, but the BOCs make converting White's advantage to a win very problematical, if at all possible. So, on the basis of the engine followup analysis to Geller’s analysis, I have not been able to find a forced win for White. It does not necessarily mean that the forced win is not there, just that I was not able to find it. Many of he positions still remain unclear, and the final "truth" will probably not be determined, if ever, until more powerful computers and stronger engines (not to mention analysts with more time, patience, and skill than myself) tackle this game again. And, of course, this is of theoretical interest only. The game was remarkably complex, both prior to 20.a3 and what could have happened after 20.Qf4. Many unbalanced positions could have occurred, and queen sacrifices for one or both sides were possible. As a practical matter, in an OTB situation with the clock ticking, I have my doubts that either Fischer or Geller would have found the best moves each time. So the loser of the “might have been” game (assuming that the game did not end in a draw) probably would have been, like the actual game, the player who made the last mistake. In case you are still interested, my followon analysis of Geller's 8 analyses follow. If you are not interested, this might be a good time to add me to your Ignore List. And, oh yes, these posts are (seriously) dedicated to <offramp>, who brought "The Application of Chess Theory" to my attention. A very enjoyable book from what I've read so far. 

Jun0114
  AylerKupp: Commentary on Geller's Analysis of 20.Qf4 (part 1 of 8) (1) 20...cxb2 21.Rh5 Bf6 22.Qf5 h6 23.Rxh6+!! gxh6 24.Qg6, with inevitable mate.
click for larger viewAfter 24.Qg6 (and actually much earlier, see below) Critter 1.6a agreed with Geller that mate was inevitable, evaluating the resulting position at [+M13], d=27 after 24...Qd2 (a horizon effect engine move, Critter tries to delay mate as long as possible regardless of the cost) 25.Bxf6+ Nxf6 26.Qxf6+ Kh7 27.Rxd2 Be4 28.Be6 Rb8 29.Re2 Bg6 30.Bf5 Bxf5 31.Qxf5+ Kg7 32.Qg4+ Kxf7 33.Re6 Rfe8 34.Qf5+ Kg7 35.Rg6+ Kh7 36.Qf7+ Kh8 37.Qg7#. I wonder if Geller calculated this or another mating line to its ultimate conclusion? I doubt it, even I can see that mate is likely (although, patzer that I am, I would hesitate to call it "inevitable"). But the engines did not consider 21...Bf6 to be among their top 3 moves, preferring instead either 21...Nc3+ (2 engines, with an average eval of [+1.90]) or 21...Nf6 (4 engines, with an average eval of [+0.93]. In fact, it looks like White has a won game after 21...Bf6. Here are two engines' top line at low ply, but mate is mate. And I suspect that if I had let the engines calculate longer they would have found shorter mates. <Critter 1.6a>: [+M30], d=20: 22.Qf5 h6 23.Rxh6+ gxh6 24.Qg6 Qd2 25.Bxf6+ Nxf6 26.Qxf6+ Kh7 27.Rxd2 Be4 28.Rxd6 Rxd6 29.Qxd6 Kg7 30.Qd4+ Kh7 31.Qxe4+ Kg7 32.Qd4+ Kg6 33.Kxb2 a5 34.Qg4+ Kf6 35.Qg8 Ke7 36.Qg6 Rh8 37.Qe6+ Kd8 38.Qf6+ Kc7 39.Qxh8 a4 40.f8=Q a3+
click for larger view<Komodo 6>: [+M18], d=21: 22.Qf5 h6 23.Rxh6+ gxh6 24.Qg6 Qd2 25.Bxf6+ Nxf6 26.Qxf6+ Kh7 27.Rxd2 Be4 28.Rxd6 Rxd6 29.Qxd6 Kg7 30.Qe5+ Kg6 31.Qxe4+ Kf6 32.Qe6+ Kg7 33.Qe5+ Kg6 34.h4 h5 35.Qg5+ Kh7 36.Be6 Rxf7 37.Bxf7 a6 38.Qg8+ Kh6 39.Qg6+
click for larger viewOf course, 22...h6 looks like a terrible move which weakens the kside further and almost dares White to play 23.Rxh6+. But both engines evaluated 22...h6 to be the only move that Black had available to delay the mate as long as possible. So it looks like Black is truly lost after 21...Bf6. 

Jun0114
  AylerKupp: Commentary on Geller's Analysis of 20.Qf4 (part 2 of 8) (2) 20...cxb2 21.Rh5 Nc3+ 22.Kxb2 Nxd1+ 23.Kcl Rxf7 24.Bxf7!, with the decisive threat of 25.Rxh7+ Kxh7 26.Qf5+.
click for larger viewAnd indeed 24.Bxf7 is decisive, as is the simple threat of 25.Bxg7+ and 26.Qxb4 winning the queen. But in his analyses Geller does not consider 23...Qxd4(!) 24.Qxd4 d5, when Black has R+B (after the fall of the Nd1) for the Q and White’s advanced Pf7 is ready to fall.
click for larger viewStarting from this position here are 2 engines' top line: <Gull 3.0>: [+2.33], d=25 after 25.c3 Rxf7 26.Bc2 Nf2 27.Bxh7 Be8 28.Bf5+ Kg8 29.Qxf2 Rd6 30.Qg3 Rff6 31.Rg5 Bf8 32.Bd3 Bf7 33.Kc2 a6 34.Rg4 Rde6 35.a4 g6 36.Kd1 Bd6 37.Qh3 Bf8 38.a5 Re8 39.Rg3 d4 40.c4 Re5 41.Qh4 Rd6 42.Rf3 Rxa5 43.Qf4 Rd7 44.Qb8 Ra1+ 45.Ke2 Ra3 46.Kf2 a5 47.c5 Rc3 48.Qc8 Re7
click for larger view<Houdini 1.5>: [+2.36], d=30 after 25.c3 Rxf7 26.Bc2 Nxc3 27.Qxc3 Rf1+ 28.Kd2 Rf6 29.Bxh7 Bd6 30.g3 Be8 31.Bg6+ Kg8 32.Bxe8 Rxe8 33.Rxd5 Rf2+ 34.Kd1 Rf1+ 35.Kc2 Rf2+ 36.Rd2 Rf6 37.Qc4+ Kh8 38.Rd4 Rf2+ 39.Kd1 Be7 40.Rd2 Rxd2+ 41.Kxd2 Bg5+ 42.Kd3 Re3+ 43.Kd4 Re8 44.Qf7
click for larger viewTrue, these two lines are also probably won for White, but Black could have put up much stronger resistance by 23...Qxd4 than Geller implied. And, as we will see later, 21...Nc3+ is not Black's best move. 

Jun0114
  AylerKupp: Commentary on Geller's Analysis of 20.Qf4 (part 3 of 8) (3) 20...cxb2 21.Rh5 Nc3+ 22.Kxb2 Rxf7 23 Qxf7 Nxd1+ 24 Kb1 Qxd4 25.Rxh7+ Kxh7 26.Qh5#
click for larger viewNo additional engine analysis of the final position is needed since even I can tell (most of the time, anyway) that mate is mate. But in Geller's line 22...Rxf7 is worse than superfluous since not only does it give up a rook after 23.Qxf7, but Black's king still does not have the option of ...Kg8 in response to Rxh7+. And White's DSB is no longer pinned so White can respond to a future ...Nc3(+) with Bxc3. Instead, the 2 engines that favored this line after 20.Qf4 cxb2 21.Rh5 Nc3+ 22.Kxb2, Komodo 6 and Stockfish DD, preferred 22...Nxd1+ instead of Geller's 22...Rxf7. 23.Kc1 is the engines’ preferred response to 22...Nxd1+. If instead 23.Kb1 then 23...Nc3+ leads either to a draw by repetition after 24.Kb2 Nd1+, a certain loss for White after 24.Kc1 Ne7+, or a probable loss for White after 24.Ka1 Qxd4. In the latter case Stockfish DD at d=30 evaluates the resulting position at [6.74] after 25.Qxd4 Bf6 26.Qg4 Be4 27.Ra5 d5 28.Ra6 Rxf7 29.Qf4 Na4+ 30.Kb1 Rdf8 31.Bxa4 Bd4 32.Qc1 Rf1 33.Bb5 Rxc1+ 34.Kxc1 Rf2 35.Bd3 Bxg2 36.Rc6 Be3+ 37.Kb2 g6 38.Re6 Bd4+ 39.Kb3 Kg7. White has no more checks and after 26.Qg4, 26...Be4 protects the vulnerable Ph7.
click for larger viewFrom my original analysis after 20.Qf4 cxb2 21.Rh5 Nc3+ 22.Kxb2: <Komodo 6>: [+1.07], d=26: 22...Nxd1+ 23 Kc1 Qxd4 24.Qxd4 d5 25. Qd3 h6 26.Rf5 Nc3 27. Qxc3 Bd7 28.Rxd5 Rxf7 29.Qa5 Rdf8 30.Rd1 Bg5+ 31.Kb2 Bf6+ 32.Kb1 Re7 33.a3 Rfe8 34.Ka2 Be6 35.Bxe6 Rxe6 36.Qxa7 Re2 37.Qc7 Rxg2 38.h3 Ree2 39.Kb1 Kh7 40. Qd7 Rgf2 41.Qd3+ g6 42. a4 Re7 43.Ka2
click for larger viewMy earlier comments were: "In the final position it looks like Black will be losing a pawn so its drawing chances do not look good, but these last few moves may be the result of shallow analysis so Black may have some improvements along the way." <Stockfish DD>: [+2.72], d=40: 22...Nxd1+ 23.Kc1 Qxd4 24.Qxd4 d5 25.Qd3 h6 26.Rf5 Bd7 27.Kxd1 Bxf5 28.Qxf5 Rd6 29.Bxd5 Rfd8 30.c4 Rf6 31.Qh5 Bb4 32.Qh4 Bc5 33.Kc2 Rf2+ 34.Kc3 Rf8 35.Qh5 a5 36.Qg6 Rf6 37.Qe4 Rf2 38.a4 Ra2 39.Qe8 Ra3+ 40.Kd2 Ra2+ 41.Kd3 Ra3+ 42.Kc2
click for larger viewMy earlier comments were: "It seems (that) by move 37 (Stockfish is) well on its way to building a fortress for Black with a bind on the dark squares, although it seemed to lose its way towards the end of the long line. So, in spite of its very favorable evaluation for White, I am not sure that White will have realistic winning chances. But maybe White can get something going by advancing its kside pawns." 

Jun0114
  AylerKupp: Commentary on Geller's Analysis of 20.Qf4 (part 4a of 8) (4) 20...cxb2 21.Rh5 Nf6 22.Rh6!, when Black can avoid the thematic Rxf6 only by going into a hopeless ending: 22...d5 23.Rxf6 Rd6 24.Rxd6 Qxd6 25.Qxd6 Bxd6 26.Rf1, and 26...Bxh2 fails to 27.Bc5, while the threat of 27.Rf6! is maintained (Murey).
click for larger viewGeller's comment after 22.Rh6 is that Black can avoid the thematic Rxf6 only by going into a hopeless ending might be true if Black continues with 22...d5 as Geller indicates, when the position after 26...Bxh2 is indeed hopeless after 27.Bc5. And Bouquet 1.8 agrees, evaluating the position after 27...Rxf7 (pretty much forced) 28.Rxf7 at [+9.24], d=21 since Black is a rook down. But Geller does not consider 22...Rxf7, which was the preferred move of all 4 of the engines which replied to 21.Rh5 with 21...Nf6. That line does give White an advantage, but not a clear win. 22...d5 does indeed lead to a hopeless ending, even though the 2 engines I used to analyze the position evaluated 26...Be7 to be superior to Geller's 26...Bxh2 since it prevents 27.Bc5. Here are the lines and their evaluations that the engines possibly considered to be a slightly less hopeless ending: <Bouquet 1.8>: [+2.31], d=29: 23.Rxf6 Rd6 24.Rxd6 Qxd6 25.Qxd6 Bxd6 26.Rf1 Be7 27.Rf3 h5 28.c3 Bf6 29.Rf5 Rxf7 30.Rxh5+ Kg8 31.Rxd5 Kf8 32.Rd6 Rc7 33.Bxf6 gxf6 34.Rxf6+ Kg7 35.Rf2 a5 36.Kxb2 Rb7 37.Ka3 Re7 38.Rd2 Re5 39.Kb2 Rc5 40.a4 Rh5 41.h3 Rg5
click for larger view<Critter 1.6a>: [+2.36], d=23: 23.Rxf6 Rd6 24.Rxd6 Qxd6 25.Qxd6 Bxd6 26.Rf1 Be7 27.Kxb2 h6 28.c4 dxc4 29.Bxc4 Rb8+ 30.Bb3 Bf8 31.Bxa7 Rb7 32.Rc1 Rxa7 33.Rxc6 g5 34.a4 Kg7 35.Kc3 Ra5 36.Rb6 Rf5 37.Rb8 h5 38.Bc4 Rf2 39.Bd5 Rf5 40.Rd8 Rf2
click for larger viewBut after 21...Nf6 22.Rh6 Geller does not consider 22...Rxf7 instead of 22...d5 and <IMO this would be the key continuation of the "might have been" game>. The following seems to be the key line, essentially <DWINS>'s line, which all the engines that chose 21...Nf6 suggested: 22...Rxf7 23.Bxf7 Be4 24.Bb3 d5 25.Bxf6 Bxc2+ 26.Bxc2 Qxf4 27.Rxh7+ Kg8 28.Rxg7+ Kh8.
click for larger view<This is probably the critical position of the "might have been" game>, so I had all 6 engines analyze it at somewhat deeper search depths than usual. White has chosen to give up its queen to maintain the initiative and the attack against Black's king, and is temporarily at a slight material disadvantage, R+B vs. Q, but will soon get a piece back. And this is without risk since White has a draw by repetition by 29.Rxh7+ Kg8 30.Rg7+ Kh8 31.Rh7+ etc. The answer, of course, is whether White has more than a draw. Here is what the 6 engines came up with as their top lines: <Bouquet 1.8>: [+1.10], d=32: 29.Rh7+ Kg8 30.Rh8+ Kf7 31.Rxd8 Ba3 32.Bxb2 Bxb2 33.Rd7+ Ke8 34.R7xd5 Bf6 35.Rb5 Qe3 36.Bf5 Kf8 37.Rb8+ Kg7 38.Rb7+ Kh6 39.h4 Bxh4 40.Rd6+ Kg5 41.Bc2 Qe1+ 42.Rd1 Qe5 43.a4 Bf2 44.Rbd7 Kf4 45.R7d5 Qe7 46.R1d3 Be3 47.Rb3 Kg4 48.Bf5+
click for larger viewThis looks like a draw to me in spite of White's 2pawn advantage. White's passed kside pawns are blockaded and the BOC will prevent White from either capturing Black's apawn or getting Black into zugswang. 

Jun0114
  AylerKupp: Commentary on Geller's Analysis of 20.Qf4 (part 4b of 8) <Critter 1.6a>: [+0.96], d=27: 29.Bd4 Bf6 30.Rg4 Qh6 31.Rg6 Qc1+ 32.Rxc1 bxc1Q+ 33.Kxc1 Bxd4 34.Bb3 Rd7 35.h3 Be3+ 36.Kd1 d4 37.Re6 Rd8 38.Rf6 Re8 39.Rf7 d3 40.Rf3 d2 41.h4 Bh6 42.Rf1 Rf8 43.Rxf8+ Bxf8 44.Kxd2 Be7 45.g3 Bd6 46.g4 Kg7 47.Ke3 Bg3 48.h5 Kh6 49.Kf3 Bc7 50.Bc4 Kg5 51.a4 Be5 52.Ke4 Bh8
click for larger viewRemarkably, this is the same final position arrived at by <Bouquet> 1.8 above. Same conclusion. <Gull 3.0>: [+1.33], d=27: 29.Bxe7 Rc8 30.Rh7+ Kg8 31.Rh4 Qe3 32.Rg4+ Kf7 33.Bg5 Qe2 34.Rf4+ Kg8 35.Ba4 Qxg2 36.h4 Rc1+ 37.Rxc1 bxc1Q+ 38.Kxc1 Qxa2 39.Bf6 Qa3+ 40.Kd2 Qd6 41.Bg5 Qb6 42.Kd3 Qa6+ 43.Ke3 Qb6+ 44.Kf3 Qa6 45.Bd7 Qd3+ 46.Kf2 Qd2+ 47.Kg3 Qe1+ 48.Kf3 a5 49.Bf6 Qd1+ 50.Kf2 Qd2+ 51.Kg3 Qd3+ 52.Rf3 Qa6 53.Ba4 Qc4 54.Rf4 Qd3+ 55.Kg4 Qg6+ 56.Bg5 Qd3 57.Bd7 Qg6 58.Bf5 Qa6
click for larger viewCan White win this? I don't think so. Black will be able to push his a and dpawns and force White to sacrifice at least one of his bishops for them and then the R+B+P vs. Q+P ending should be pretty even. Gull doesn't think so either; starting from this position it evaluates the resulting position at [+0.03], d=21 after 59...a4 60.Bf6 Qxf6 61.Bh7+ Kg7 62.Rxf6 Kxf6 63.Kf3 Kg7 64.Bb1 a3 65.Ba2 d4 66.Ke4 d3 67.Kxd3 Kh8 68.Bb3 Kh7 69.h5 Kh8 70.Ba2 Kg7 71.Ke4 Kh7 <Houdini 1.5a>: [+0.74], d=31: 29.Bd4 Bf6 30.Rg4 Qh6 31.Rg6 Qc1+ 32.Rxc1 bxc1Q+ 33.Kxc1 Bxd4 34.Bb3 Bb6 35.Kd2 d4 36.h4 d3 37.Rg5 Rd4 38.g3 Bd8 39.Rf5 Be7 40.Ke3 Rg4 41.Kf3 Rg6 42.Re5 Bd6 43.Re8+ Kg7 44.g4 Rf6+ 45.Ke3 Rf8 46.Rxf8
click for larger viewAfter 46...Kxf8 47.Kxd3 we'll reach a similar position as Bouquet and Critter above; White will have a 2pawn advantage but Black will be able to blockade White's kside pawns. A little bit harder since White's pawns are mobile but eventually they will need to move to light squares and then Black can establish a blockade with his king on the dark squares. 

Jun0114
  AylerKupp: Commentary on Geller's Analysis of 20.Qf4 (part 4c of 8) <Komodo 6>: [+1.14], d=31: 29.Bxe7 Rc8 30.Rh7+ Kg8 31.Rh4 Qf2 32.Bh7+ Kg7 33.Bd3 Rc1+ 34.Rxc1 bxc1Q+ 35.Kxc1 Qxg2 36.Bd8 Qxa2 37.Bg5 Qa1+ 38.Kd2 Qb2+ 39.Bc2 d4 40.Rh7+ Kg8 41.Rh3 Qb4+ 42.Kc1 a5 43.Bd2 Qc4 44.Rd3 a4 45.Be1 Qe6 46.Bb4 Qh6+ 47.Rd2 Qb6 48.Ba3 Qc6 49.Rxd4 Qh1+ 50.Kb2 Qxh2 51.Rd8+ Kf7 52.Rd7+ Ke6 53.Rd1 Qf4 54.Ka2 Qc4+ 55.Kb1 Qf4 56.Bb2 Qe3
click for larger viewCan White win this? Maybe. Maybe someone with access to 7piece tablebases could verify. Komodo is on the fence; starting from this position it evaluates the resulting position at [+0.85], d=25 after 57.Bd3 Kd5 58.Bc1 Qc5 59.Bg6+ Kc4 60.Bb2 Qf2 61.Ka1 Qf3 62.Rb1 Kb4 63.Bc2 Kc5 64.Rc1 Kb4 65.Bh7 Qh3 66.Bg8 a3 67.Bd4 Qd3 68.Rb1+ Ka5 69.Bf6 Qf3 70.Be7 <Stockfish DD>: [+2.18], d=39: 29.Bxe7 Rc8 30.Rh7+ Kg8 31.Rh3 Rxc2 32.Kxc2 Qa4+ 33.Rb3 Qe4+ 34.Rbd3 Qxe7 35.R1d2 Qb4 36.Kb1 Qc4 37.Rd1 Qe4 38.Kxb2 Qxg2+ 39.Kb3
click for larger viewCan White win this? Again, maybe and perhaps someone with access to 7piece tablebases could verify.. After 39...Qxh2 40.Rxd5 White will have 2R vs. Q and starting from that position Stockfish evaluates the resulting position at [+0.78], d=32 after 40...Qb8+ 41.Kc2 Qc7+ 42.Kb1 Qb6+ 43.Ka1 Kf7 44.R5d3 Ke6 45.R1d2 Qg1+ 46.Kb2 Qg7+ 47.Kc2 Qc7+ 48.Rc3 Qh7+ 49.Rdd3 Qe4 50.a3 a5 51.Kb2 Qb7+ 52.Ka2 a4 53.Re3+ Kd5 54.Rg3 Kd4 55.Rgd3+ Ke4 56.Rd6 Qb5 57.Rd8 Qb6 58.Rd1 Qf2+ 59.Kb1 Ke5 So in this critical line of the "might have been" game and starting the analysis from the critical position (which is almost forced), IMO the engines show four likely draws for Black and two possible wins for White, but no certain White wins. 

Jun0114
  AylerKupp: Commentary on Geller's Analysis of 20.Qf4 (part 5 of 8) (5) 20...d5 21.Qe5 Nf6 22.Rxf6 Bxf6 23.Qxf6! and his task is very simple.
click for larger viewYes, White's task is very simple. Here are the evaluations of two engines, clearly a win for White: <Houdini 1.5a>: [+9.01], d=25: 23...Rxf7 24.Qxf7 Rg8 25.bxc3 Qf8 26.Qxf8 Rxf8 27.Bxa7 Ra8 28.Bxd5 Rxa7 29.Bxc6 Kg8 30.Bd5+ Kf8 31.Rf1+ Ke8 32.c4 Re7 33.a4 Kd7 34.a5 Kd6 35.Kb2 Ra7 36.Rf7 Rxa5 37.Rxg7 h6 38.Rh7 Ke5 39.Rxh6
click for larger view<Stockfish DD>: [+6.86], d=30: 23...Bxf6 24.Bxf6 gxf6 25.bxc3 Rxf7 26.g3 Re7 27.Kc1 Kg7 28.Bc4 Re4 29.Bd3 Ra4 30.Re1 Kf7 31.Kd2 Rxa2 32.Rb1 h6 33.Be2 Ke6 34.Rf1 Rb8 35.Bg4+ Kf7 36.Bf5 Re8 37.Bd3 Rd8 38.Ke3 Ra4
click for larger viewWell, maybe Black has stalemating chances. ;)
But White's task is not so simple if instead of 22...Bxf6 Black plays 22...Qxd4(!?). Then after 23.Qxd4 Bxf6 Black has R+B vs. Q and White's Pf7 is about ready to fall. Here are two engines' evaluation of the resulting positions from the original May2014 set of analyses after 22...Qxd4: <Houdini 1.5a>: [+4.49], d=26: 23.Qxa7 Bd7 24.Bxd5 Be6 25.Bf3 Rb8 26.b3 Bxf7 27.a3 h6 28.b4 Bg8 29.Qa5 Bg5 30.b5 Bd2 31.Be4 Rbd8 32.Rh1 Rd4 33.Bf3 Rf5 34.Rd1
click for larger view<Stockfish DD>: [+4.88], d=40: 23.Qxa7 Bd7 24.Bxd5 Be6 25.Bf3 Rxd1+ 26.Bxd1 Bxf7 27. b4 Bg8 28.Bf3 Re8 29.Qf2 Bg5 30.a3 Bf4 31.b5 Bd6 32.a4 Bc5 33.Qg3 Bb6 34.Kc1 Rd8 35.Qe1 Rd4 36.Qxc3
click for larger viewSo Geller's analysis (22...Rxf7) does not show the best line for Black (22...Qxd4). OK, maybe I'm quibbling because both of these lines are also wins for White since the passed qside pawns will carry the day. But Black could have put up greater resistance. 

Jun0114
  AylerKupp: Commentary on Geller's Analysis of 20.Qf4 (part 6 of 8) (6) 20...Nd2+ 21.Rxd2 cxd2 22.c3!! Qb7 (or 22...Qxb3), and Black succumbs to the mating combination 23.Bxg7+! Kxg7 24.Qg4+ Kh8 25.Qd4+.
click for larger viewTrue, of course, in the position above it's mate in 1 after 25...Bf6 26.Qxf6#. Even I can see that. The engines in my earlier analysis indicated that 20...Nd2+ was either the 2nd or 3rd best move by Black, with an average evaluation of [+3.62] and an average search depth = 28. But all of the engines considered 22...Qc5 to be Black's best reply to 22.c3 (see line (7) below); after 22...Qc5, 23.Bxg7+ does not directly lead to mate since after 23...Kxg7 24.Qg4+ Kh8, Geller's 25.Qd4+ is not possible.
click for larger viewStockfish DD considers that the best that White has at this point is 25.Rxc5 which it evaluates at [8.28], d=32 after 25...dxc5 36.Bd1 Rxf7 27.Qc4 Rf2 28.Kc2 Bxg2 29.Qg4 Bd5 30.b4 Bf3 31.Qxf3 Rxf3 32.Bxf3 cxb4 33.c4 Rd4 34.Be2 d1R 35.Bxd1 Rxc4+ 36.Kd3
click for larger viewSo clearly Geller's mating combination does not work after 22...Qc5, but it would clearly take nerves of steel and supreme confidence in one's calculating ability to put one's queen en prise to both the opponent's rook and bishop. But White has better, see the analysis for line (7) below.
click for larger view 

Jun0114
  AylerKupp: Commentary on Geller's Analysis of 20.Qf4 (part 7 of 8) (7) 20...Nd2+ 21.Rxd2 cxd2 22.c3!! Qc5 (best, although it does not get Black out of his difficulties) 23.Kc2! Bd7 24.Bxc5 Bxf5+ 25.Qxf5 dxc5 26.Kd l Rd6 27.Qe5 Rd7 28.Qe6 Rb7 29.Bc2, and against 30.Qf5 there is no satisfactory defence.
click for larger viewAll the engines agreed with Geller that after 20...Nd2+ 21.Rxd2 cxd2 22.c3 that 22...Qc5 was Black's best reply. Starting from 20.Qf4, the engines considered both 23.Bc2 (2 engines, average evaluation [+5.02]) or 23.Qg4 (4 engines, average evaluation [+2.90]) to be better replies to 22...Qc5 that Geller's 23.Kc2, even though Geller gave it a "!". However, it makes no difference; after 20...Nd2+ Black is probably lost or, as Geller described 22...Qc5, "it does not get Black out of his difficulties". Sliding forward, here is what two engines had to say after 22...Qc5: <Critter 1.6a>: [+4.61], d=27: 23.Bc2 Qe5 24.Bxe5 d1Q+ 25.Bxd1 dxe5 26.Qxe5 Rxd1+ 27.Kc2 Rfd8 28.f8Q+ Bxf8 29.Rxf8+ Rxf8 30.Kxd1 Rd8+ 31.Kc2 Bxg2 32.Qe7 Ra8 33.c4 h6 34.c5 a6 35.Qd7 Rf8 36.c6 Be4+ 37.Kc3 Bf5 38.Qe7 Kg8 39.c7 Bc8 40.Kb4 Rf7 41.Qd8+ Rf8 42.Qd5+ Kh8 43.h4 g6 44.Qe5+ Kh7 45.b3 Kg8 46.Qe7 Rf4+ 47.Ka5 Rf8
click for larger view<Komodo 6>: [+3.22], d=28: 23.Bc2 Qe5 24.Bxe5 dxe5 25.Qxe5 d1Q+ 26.Bxd1 Rxd1+ 27.Kc2 Rfd8 28.f8Q+ Bxf8 29.Rxf8+ Rxf8 30.Kxd1 Bxg2 31.Qe7 Ra8 32.c4 h6 33.c5 Kh7 34.Kc2 Bh1 35.Kc3 a5 36.Qd7 Ra6 37.Qf5+ Rg6 38.h4 h5 39.Qxh5+ Rh6 40.Qf5+ Kg8 41.Qd7 Bc6 42.Qd8+ Kh7 43.Qd3+ Kg8 44.b3 Kh8 45.Kb2 Kg8 46.Qd8+ Kh7 47.Qc7 Bd5 48.Qe5 Bc6 49.Qf5+ Kg8
click for larger viewGame over. What I found amusing was that after Black's 22...Qc5 putting his queen en prise to two of White's pieces, that White's best reply is apparently 23.Bc2, ignoring the offered queen! It took Komodo until d=22 to realize that 23.Bc2 was better than 23.Qg4, and then eval jumped from [+2.06] at d=21 for 23.Qg4 to [+2.78] at d=22 for 23.Bc2. It took Critter slightly longer to realize the same thing; and its eval jumped from [+3.07] at d=23 for 23.Qg4 to [+4.05] at d=24 for 23.Bc2. And the evals continued to climb for both engines. 

Jun0114
  AylerKupp: Commentary on Geller's Analysis of 20.Qf4 (part 8 of 8) (8) 20...Nd2+ 21.Rxd2 cxd2 22.c3!! Qc5 23.Kc2!! Qe5!? 24.Rxe5 dxe5 25.Qxe5 Bf6 (with the hope of 26.Qxf6? d1=Q+!) leads to a hopeless position after the correct 26.Qc5 Bxd4 27.cxd4 Rc8 28.Kxd2 Ba4 29.Qe7 Bxb3 30.axb3 Rcd8 31.Ke3.
click for larger viewGeller's comment that 25...Bf6 leads to a hopeless position after 26.Qc5 is clearly correct, with Q+3P vs. 2R, and with both of Black's rooks tied to the defense of the 8th rank. To see how bad it actually is, Rybka 4.1 (just for a change of pace) evaluates the resulting position at [+13.55], d=24 after 31...h6 32.d5 g6 33.Ke4 Kg7 34.Ke5 Rxd5+ 35.Kxd5 Rxf7 36.Qe5+ Kh7 37.b4 Rf5 38.Ke6 Rxe5+ 39.Kxe5 g5 40.Kd6 Kg6 41.Kc6 Kf5 42.Kb7 Ke4 43.Kxa7 Kd3 44.b5 Kc2 45.b6 Kxb2 46.b7 Kc3. Ouch!
click for larger viewSo this is yet another way for Black to lose. 

Jun0214   Howard: So all I wanted to do was....make a "simple" inquiry back around May 20 as to whether the oftencited critical move of 20.Qf4 would have led a forced win for Fischer....Kasparov seemed to think the answer was yes, in his book MGP. So now, there's been a plethora of computergenerated analysis of that move, but unless I'm not mistaken, there's STILL no clearcut conclusion, correct ? This game is starting to shape up as being akin to KasparovTopolov, 1999 when it comes to the extensive analysis posted. Anyone besides myself who believes that this should be a Game of the Day ?! I certainly do!! 

Jun0314
  AylerKupp: <Howard> You are correct as far as I'm concerned. I think that some of the lines that Geller believed led to wins for White are not correct since he did not consider some moves that the engines assessed to be better that the moves that he considered. And some of the lines I analyzed following 20.Qf4 , and some of Geller's analysis as well, led to positions where White had a theoretically significant 2R+2P vs. Q advantage that would certainly look very promising for White but with BOC and the active Black king I didn't see how White could make progress. And in some other lines at the end of the analysis White had a clear advantage but whether it was enough to win was open to question. It is difficult to conclusively "prove" that either Fischer had a won game or that Fischer did not have a won game. Exhaustive search would be one way but other than endgame tablebases this is not possible with current technology. All we can practically do at the moment is to look at enough move alternatives to reduce the probability that that whatever we're trying to prove is correct or not. So I think that this is all I have done, reduce the probability that Fischer had a won game after 20.Qf4. There are also some deficiencies in the approach I used but I think that a detailed description of these deficiencies is offtopic, besides being extremely long and extremely boring. So, if interested, see AylerKupp chessforum . Maybe some day I'll take a cue from <kingscrusher> and make a video of these deficiencies. :) And, yes, I agree that this should be GOTD someday. I'm not sure if there is a better title than Geller's "On the edge of the abyss" when referring to this game, but maybe someone can make a clever and deserving pun for the title of the game. I did find some interesting references to "abyss" such as Nietzsche's "If you gaze long into an abyss, the abyss will gaze back into you." and H.P.Lovecraft's "The process of delving into the black abyss is to me the keenest form of fascination." You can find a more comprehensive list here http://www.brainyquote.com/quotes/k.... But, boy, talk about offtopic! 

Jun0514   qqdos: <Howard> <AylerKupp> May I suggest that a fitting and topical dedication to this
magnificent game and <AK's> exhaustive analysis would be to VELIMIROVIC, the hero of the Velimirovic Attack who died recently. Fischer was not always comfortable with this opening and the the system used here by Geller is not a refutation. With 20.Qf4! played, I would put my money on White  but with White short of time and Geller or Kasparov behind the Black pieces, who knows?!? 

Aug0915   Nickster: I decided to look at this old classic a few days ago with the latest engines. After 20.a3 most analysis give this as practically losing. click for larger viewBut the latest engines say it's still drawn.
20... Qb7 21.Rh5! (Instead of 21.Qf4?)
click for larger viewBoth Komodo 9.02 and Stockfish 6 say this is about equal. I played through their lines and it is indeed drawn. So the real losing move was 21.Qf4? not 20.a3. I wonder if this has been mentioned anywhere before. 


< Earlier Kibitzing · PAGE 18 OF 18 ·
Later Kibitzing> 


Now on DVD


