|
< Earlier Kibitzing · PAGE 26 OF 29 ·
Later Kibitzing> |
| Sep-21-09 | | JamesMazur2: OK here is anaylsis for 48. ...Qxe4.
I know there are MANY possible variations but this is just a sample:49. d6 Qxf5+
50. Kg1 Rd7
51. Qxb5 Qe6
52. Qc5 Kg6
53. Rc6 e4
54. Qd4 e3
55. Qxc4 Qxc4
56. Rxc4 Rxd6
57. Kf1 Kf7
58. b5 Rb6
59. Rb4 Ke6
60. Ke2 Rb7
61. g4 f5
62. gxf5+ Kxf5
63. Kxe3 Ke5
64. c4 g5
65. Rb1 Re7
66. Kd3 Rf7
67. b6 Rf3+
68. Ke2 Rf8
69. Rd1 Rc8
70. Rd5+ Ke6
71. Rc5 Rb8
72. Rc6+ Ke5
73. c5 Kd4
74. Rc7 Ra8
75. c6 Ra2+
76. Kf3 g4+
77. Kxg4 Kc5
78. b7 Ra4+
79. Kh5 Rb4
80. Rc8 Rb1
81. b8Q Rxb8
82. Rxb8 Kxc6
83. Kxh6 Kc7
84. Rb3 Kd6
85. h4 Ke6
86. Kg6 Ke5
87. h5 Kd6
88. h6 Kd5
89. h7 Kc4
90. Rh3 Kd5
91. h8Q Kd6
92. Rc3 Ke6
93. Rd3 Ke7
94. Kf5 Kf7
95. Rd7# 1-0
This is good play though definitely not perfect...I probably will not post more analysis for a while. |
|
| Sep-26-09 | | Bartleby: After having watched the silly, but entertaining, agitprop docu-drama (with emphasis on the DRAMA, from the "sinister" background music and whispered conversations) "GAME OVER: Kasparov and the Machine" at this nifty little educational site: http://topdocumentaryfilms.com/ I've decided to throw my hat into the analytical arena, with the aid of a freshly purchased Junior 10. I'm not pretending it's going to out-perform others here aided with Shredder and Deep Rybka and what-not, but armed with "Deep Position Analysis" and the belief that Junior's unique predilections for certain dynamic/speculative moves may give my contrasting analysis a certain spice, I sallied forth. I actually entertain doubts as to Kasparov's decision to play a well-established, theoretical variation of the Smyslov Ruy Lopez, which Junior didn't even have to think about for the first twenty moves. Up to that point all one is doing is playing against a "book," assuming the programmers are worth their preparatory salt at all (and they proved it in game six's all-book crusher). Considering Garry's other openings in this match, a more anti-theoretical approach may have served him better, instead of a line that gave him a rather cramped and thankless middlegame position. A lot of ink has already been spilt over the "human-like" 37. Be4, but as others have pointed out, for today's engines it's hardly beyond their positional grasp. Junior heed and hawed over 37. Be4 vs. 37. Qb6. It liked both of them as candidate moves and depending on how long I had Junior look at the position, it would incrementally prefer one or the other by a few hundredeths of an eval. Considering Junior is supposed to be less a pawn-grubbing materialist than any other commercial chess engine out there, I presumed 37. Be4 to be almost a foregone conclusion, however it was far closer a call than I assumed it'd be. If Deep Blue was to be, say, ten years before its time in terms of strength and chess knowledge vis-a-vis today's computers, then it isn't a stretch to conclude that finding 37. Be4 is/was within its visual reach, especially if it was using the same positional "chess sense" that caused it to play 26. f4!. If you could call any move "suspicious" in this game, 26. f4! was it. For the life of me I can't get Junior to elect it to the fore of its top TEN (!) candidate moves, opting for just about every other piece maneuver imaginable, no matter how insipid: 26. Bd1, Qd1, Qd2, Qe1, Qf3 & Qf1 are all recommended over it. 26. Nf5 is more understandable, something Junior has been harping on for the last five moves over Deep Blue's choices (but it didn't really disagree with what was played either, once it was played. This happens a lot in evaluations I've observed). I was hoping Junior would recognize the necessity of applying pressure to black's center with such an intuitive (for any human playing this position, including myself) advance, aiming at advantageous line-opening when holding a monopoly on space. But I guess not! It's strange because in the time I've owned Junior it seems to understand pawn dynamics much better than previous programs I've owned. Deep Blue, with that positional knowledge programmed into it by Joel Benjamin and others, must have understood "triggers" for pawn lunges like 26. f4, or at least been more sensitive to them, than other machines. That positional knowledge (similar to Hiarcs maybe?) must have rejected similar sojourns like 28. ab ab 29. Ba7!? which Junior went bonkers for (a lamentable "computer-like" continuation if I ever saw one). However on the whole Junior concurred with Deep Blue's move choices. But would have 37. Qb6 really "ruined" white's game? 37. ...Rxa2 38. Rxa2 Ra8 39. Ra5! (and not the blunder 39. Rxa8? which allows 39. ...Qxa8 followed by invasion of the queen on the back rank and an eventual perpetual that white cannot prevent due to its critical weakness on the dark squares) when white keeps a small edge. 39. ...Rxa5 (39. ...Qb8 40. Qxb8 Rxb8 41. Ra6 Bc7 (41. ...Rd8 allows 42. Rb6) 42. d6 Bd8 43. Be4 with an eval of +0.89+, is the lessor black alternative) 40. bxa5 Qb8 41. Qxb8 Bxb8 42. Kf2 is the most accurate line, residing at a middling final eval of +0.20 or so. However, in both endgame lines, with opposite colored bishops, whether Deep Blue could have converted this into a win in the long-run is pure speculation. It's definitely a lot less convincing than the nuanced and positionally superior 37. Be4 lines, which Deep Blue's programmers should be praised for being a heuristic reality, vs. the more hatchet-like 37. Qb6. |
|
| Sep-26-09 | | Bartleby: The flip side to this "positional" vs. pure "tactical" slant (for wont of a better word) in deciding on certain moves versus others, is that Deep Blue, even at calculating tens of million of positions a second, is prone to inaccuracies. Someone mentioned such an inaccuracy that Deep Blue elected for on move forty, 40. Qb6. Not that it is bad, but apparently not the strongest, which would be 40. Qb8+. Junior confirms, and prefers that move at an Eval of +1.61, vs. 40. Qb6's "mere" +1.24. A sample line: 40. Qb8+ Rc8 41. Qb6 Rb8 42. Qc6 Qc7 (...Rd8 43. Ra5! with a bind) 43. Ra6 Qxc6 44. dxc6 Bc7 45. Kf1! with a superior endgame at a healthy +2.30 eval, which would have incidentally sidestepped all that theoretical perpetual checking nightmare (for white) that occurred later, in the post-mortem notes if not the actual game, that Kasparov resigned rather prematurely. Finally we come to the most pored over critical position, which is after 45. ...Qe3. Draw or no draw? Well, gently coaxing Junior 10 along certain paths like a shepherd with a willful goat: a: Moves such as 46. Bf3, Ra1, Ra2, are all equally drawn against. 46. Qd7+ is one of only two top contenders in Junior's mind, but after 46. ...Kg8 Junior finds the lines dead drawn (with a little coaxing along the trees) in all ensuing white continuations, : a-1: 47. Bf3 Qc1+ 48. Kf2 (48. Ke2? e4! when Junior thinks black can play on for a win, incredibly, with +1.5 evaluation approx. in both Qxd6 and Rxd6 lines) Qd2+ (49. Kg3 Qf4+, 49. Kg1 Qc1+ 50. Kh2 Qf4+, etc.) 49. Be2 Qf4+ 50. Kg1 Qe3+, etc. a-2: 47. Re7 is thwarted by the defensive undeveloping 47. ...Bf8 when white has nothing to prevent the typical perpetual check lines above. In fact, 48. Bf3? no longer works when black has 48. ...e4! and can play for endgame complications. Junior actually thinks white's best try is now liquidating the bishop with 49. Bxe4!? Qxe4 50. Qe6+ Qxe6 51. fxe6 with a couple of central passers for the minor piece. Junior still favors black in this position (51. ...Rd8 52. e7 Bxe7 53. Rxe7 Rxd5 with a 3-vs-2 king-side pawn majority), but definitely frowns on other deviations at move 49, i.e. 49. Be2 Qxc3 with a -2.00 in black's favor, with many ordinary white moves giving black even more than that (50. ...Qf7+, ...Qe6+, ...Bh5, ...d6), 50. Qc7 alone being the precise favored move. Same thing with 49. Bg4 Qxc3 and 49. Bh5 Qc1+ 50. Kf2 e3+ and 49. Bh5 Qc1+ Ke2 Qg1, all with the same rough -2.00 evaluation. b-1: 46. Qxd6 (and not faulty Qe6+? interposing check, which after 46. ...Kh7 47. Qxd6 Qxe4 white cannot prevent perpetual check around its king) is truly the only line to consider; Junior initially gave it a huge decisive advantage before settling on a solid +1.00 eval after a few minutes of calm calculation (compared to the dead 0.00 in other lines). 46. ...Re8 in necessary, but here at another branch at move 47 white must be very precise. 47. Bf3?, which Junior at first heavily favored, allows black to escape with the half-point: 47. ...Qc1+ 48. Ke2 Qc2+, etc, or 48. Kf2 Qd2+ with similar checking continuations to the above a-1 lines. .. b-2: After prolonged thought, Junior at first vouched strongly for the mysterious pawn move 47. h4, thinking it decisive, and believed that black was forced to reply with 47. ...h5 (!??) and 47. ...Qxe4 was a gross blunder. But naturally, it came around and decided that 47. ...Qxe4 forced a draw after all (so did ...h5, by-the-by): 48. Ra7+ Kh7 49. Qd7 Re7! a clever resource. Sac the rook to deflect the defender of the white f5 pawn (white is forced to take this rook, even if he tries to delay with other moves like Qd8+ or Ra8+, i.e. 50. Qc8+ Kh7 51. Ra8 Re8!), after which black can safely perpetual check the white king without interference. b-3: Obviously, the most testing continuation is what others have echoed on this page, 47. Qd7+ Re7 48. Qc6! I follow James's analysis: 48. ...Qf4+ (which Junior doesn't believe is worse than 48. ...Qxe4. I'm not going to into the ramifications of the ...Qxe4 line in any case; the amateur analysis buffs in the commentary of that YouTube vid linked on this page seem solidly on top of proving a decisive white advantage in that variation) 49. Bf3 Qc1+ 50. Kf2 Qd2+ 51. Kg1 Qc1+ 52. Kh2 Qf4+ 53. Kh1 Qc1+ 54. Bd1! Qxd1+ 55. Kh2 Qd2 56. d6, James offered 56. ...Qf4+ and 56. ...Re8, which Junior snubs as equally prospectless for black. Much to my amazement, Junior uncorked the incredible: |
|
| Sep-26-09 | | Bartleby: b-4: 56. ...e4! instead, roadblocking the long h1-a8 diagonal and preventing the queen from dropping back to g2 during perpetual checks! 57. Ra1 and 57. dxe7 both allow black to escape with the half-point, starting with 57. ...Qf4+, but white has an incredible counter-resource with 57. Ra2!!, leaving its *own* rook en prise to deflect the black queen from the white king and thwart the perpetual check threat. 57. Ra2 Qxa2 58. dxe7 Kxe7 59. Qc5+ Kf7 60. Qxb5 Qb3. Material is equal and though white's position is more active I think black has good practical chances to hold this position. Junior thinks white has at most is +1.00 eval during any point along the variation trees, and its sample line goes: 61. Qd5+ Kf8 62. Qd2 Kg8 63. Kf3 Qb1 64. Kf2! (64. Kf4 Qd3! black has neutralized white's pressure with the help of his own advanced pawns) e3! 65. Kxe3 Qxf5 66. Qd8+ Kh7 67. Qd4 Qf1, which if I were white, would find a headache to eventually convert. I'm sure it could be done with precise technique (in this position Junior is warming up to 68. b5, however black has resources too, after 68. ...f5), and I'm not saying Deep Blue or any other computer engine wouldn't be up to the task (for reasons other than fatigue, of course, as evaluations are never set in stone when a game is in transit and endgame tables don't cover every contingency). But if black doesn't have a forced draw after 46. ...Qe3, white doesn't have a forced win either, in my opinion. Against two humans after 68. b5 f5, black would have every reason to fight on for the draw (or a reversal that gives *him* chances). Computers can misstep, as we all know. Its fluid evaluation function may wrongly promote superficially attractive but insufficient lines vs. more subtly decisive ones rejected at first glance because the engine regards them as "ugly" and doesn't take dynamic counterbalancing elements enough into effect. This leads to plenty of opportunity for inaccuracy, especially in closed positions with a fixed pawn structure, such as this one. Junior, in the a-1 & a-2 lines for instance, thought 46. Qd7+ decisive at first glance, 46. Qd7+ and 46. Qxd6 equally advantageous at second glance, and needed a long hard look before realizing black can draw against all the 46. Qd7+ variation trees. It also didn't think in the b-1 lines 47. Bf3 allowed black to slip away either, and was actually the its favored choice until I prodded it into a deeper look of the position. Junior could have easily chosen a less punishing line in the b-2 losers that would have thrown away white's advantage and allowed black to draw, contrary to what snap-evaluations told it at first glance. Definitely the more time Junior, and other engines by association, spent in these Byzantine positions the more precise the variations promoted. I can presume that Deep Blue was in its own version of "Deep Position Analysis" all the time, move-by-move, taking its programmed "positional sense" into account at each juncture. While using Junior, I observed its evaluations continually refining themselves and changing, sometimes one move (!) into a variation tree, and were subject to constant re-evaluation, often down the tree to the point where it would have been too late to save half or a whole point in-game (even for an immortal computer). Only in post-mortem are we allowed to continually backtrack and chisel away at deeper truths in the position, hindsight being 20/20 for both man and machine. |
|
Sep-26-09
 | | Check It Out: <Bartleby> I clicked on that top documentary link and some sort of malware virus attacked my computer. |
|
| Sep-26-09 | | Bartleby: I've perused that site a great deal and never run into those problems. It's not a pirate or warez site. (this is as much for chessgames's benefit as yours, FYI) It doesn't ask you to run software or download stuff. It's all streaming video. You may try the direct link to the Deep Blue doc and see if it's any better: http://topdocumentaryfilms.com/game... I run SuperantiSpyware 4.27 and Avira Antivirus 9.0 and they didn't make a peep. You may be getting false positives. |
|
| Sep-26-09 | | Bartleby: Addendum: I was made aware of a couple other junctures in the game worthy of notice so I expanded one paragraph, see below: ****************
That positional knowledge (similar to Hiarcs maybe?) must have rejected similar sojourns like 28. axb5 axb5 29. Ba7!? which Junior went bonkers for (a lamentable "computer-like" continuation if I ever saw one). To Junior's credit, however, just below 28. ab on the totem pole was 28. Qf2, which Yasser Seirawan gives a ! in his analysis of this game (28. Qf2! Ne8? 29. Bd6 is the line he gives, though junior now finds 29. ...axb5! even stronger) , over the somewhat limpid 28. Qf1?! found in the text of the game, which Junior doesn't even consider in its candidate moves. However on the whole Junior concurred with Deep Blue's move choices. According to most of the contributing analysts, the real stress test is not the false decision at move 37 and whether Be4 is conceivable over Qb6, but is one move earlier, at move 36. Here, the choice between 36. axb5 or the pawn-hunting 36. Qb6 is looked upon as critical. GM Lubosh Kavalek offers the "three-pawn sac" line of 36. Qb6 Rd8 37. axb5 Rab8 38. Qxa6 e4, with counterplay on the dark squares for black. I admit, Junior prefers 36. Qb6 over 36. axb5 by about half a pawn. However, Junior shuns the gambit continuation at not one, but *two* junctures. It favors 37. Be4 over 37. axb5 (positional deja vu), and favors 38. Qc6 over pawn-snatching. For all Junior's reputation for swashbuckling bravada, it seems like its own "king-safety" algorithm is well in check here, as Deep Blue's was after Game 1. It doesn't find the position in the above Kavalek line after 38. ...e4 untenable after 39. b6 Qe5 40. Qxc4 Qh2+? 41. Kf1 with an eval of +1.46, but found the more challenging 40. ...e3!, e.g. 41.Kf1 Rbc8 42. Qg4 Qxc3 43. Ra7!, a canny minor piece sac (43. Bd1?! is frowned on in light of 43. ...Rc4 44. Qg6 Rf4+ 45. Bf3 Qd3+ when *BLACK* now holds the chances in this position, at an eval of -1.11 to -1.40, depending on how white proceeds) e2+! 44. Qxe2 Qxc2 45. Qxc2 Rxc2 46. Ra8 Bb8! which is now dead level, according to Junior's eval. Regardless, allowing this line by black in the first place would have been slipshod positional play, and I wonder if Deep Blue would have concurred with Junior's deviations had it actually played 36. Qb6 over-the-board.
******************
This is a game of virtually fathomless depths. It makes it really fascinating to study this game in its psychological context. All the parallel processors Deep Blue was running on was certainly not wasted during game 2 of the match at least. To give you an idea of its complexity, Yasser Seirawan believed the line in b-2 led to a white victory after 47. ...Qxe4, when it in fact draws. GMs are prone to plenty of oversight themselves. |
|
| Sep-26-09 | | Bartleby: Actually, I just learned that whole 56. ...e4 line is total dross. I hallucinated and somehow didn't compute (pun intended) 57. Ra1 (much stronger than Ra2, no blame to an impatient Junior, which had it lit up in its eval panel, but alas I only saw 57. Ra8 in my fallible mind's eye) which refutes the whole cheapo idea. Nevermind. James is right, the whole 48. ...Qf4+ line is prospectless for black. Now with a raging 6-hours-plus-hunkered-over-the-laptop-analyzing--
one-game migraine, I'm retreating from the cg battlefield. So many dead words in my wake. |
|
Sep-26-09
 | | OhioChessFan: Ahh, Bartleby, there, there. Be a good chap and leave the office, but don't leave the discussion. |
|
| Sep-29-09 | | JamesMazur2: <Bartleby>
First, I must say, very nice analysis. It was more work than I was willing to do (though I did have to overwrite some of Rybka's moves myself). Secondly, just a simple observation. This game alone is extremely complex. There is still debate over it over 10 years afterwards, and we have computers to analyze it, too. Imagine trying to solve the ENTIRE game of chess, as checkers was recently. Yep. It's not happening anytime soon. Thirdly, who came up with the idea of ever resigning in chess? I really wish that Kasparov and Deep Blue had played this game out until the end. Now, we will [very probably] never know if Deep Blue had 46. Qxd6 and 47. Qd7+ (or Qc7+) ready, and could pull out all the stops to win. We also will never know if Kasparov was skilled enough to draw, even if Deep Blue had found that line. I don't know who suggested the tradition of resigning as frequently as it happens in chess; this is not a common practice in most other games and sports. Fourth comment: I got into this whole mess because of "Kasparov v. Deeper Blue: The Ultimate Man v. Machine Challenge," a book by Daniel King.
It gives a thorough analysis of how every relevant move in the final position draws except for (conveniently) 47. Qd7+. The move is thrown away as irrevelant ("as the queen can not return to defend the king after 47...Qxb5"). I am a serious AMATEUR at chess (15 years old, wonder if you guessed), so how I figured out this line casually reading the book (without a computer, may I add), and a group of grandmasters missed it, I still am perplexed. This leads into my 5th comment: I think Kasparov, the Deep Blue team, and Daniel King should be notified of this, even if they don't care. The truth, in my opinion, should be common knowledge to all. And my last comment: I am basically going to retire from analyzing this game now. I plan to post one more very strong line, that shows White's good winning chances. I will then await any criticisms of Black's moves from those who still believe this position is clearly drawn, and respond accordingly with refutations. Goodbye. |
|
| Sep-30-09 | | JamesMazur2: OK, I have my final analysis :)
45. ...Qe3
46. Qxd6 Re8
47. Qd7+ Re7
48. Qc6 Qxe4
49. d6 Qd3+
50. Kg1 Re8
51. Ra1 h5
52. Qxb5 Rd8
53. Kh2 Kf8
54. Qc5 Qxd6
55. Qxd6+ Rxd6
56. Ra7 Rd3
57. b5 Rd8
58. Rb7 e4
59. Kg1 Rd6
60. b6 e3
61. Kf1 Ke8
62. g4 hxg4
63. hxg4 Kd8
64. Ke2 Kc8
65. Rxg7 Rxb6
66. Kxe3 Rd6
67. Ke4 Kd8
68. Rf7 Kc8
69. g5 fxg5
70. f6 Ra6
71. Kf5 Kd8
72. Rg7 Ra5+
73. Kg6 Ra6
74. Rg8+ Kd7
75. Kg7 Ra7
76. f7 Ke6
77. Re8+ Kf5
78. Kg8 Rxf7
79. Kxf7 g4
80. Rg8 Kf4
81. Ke6 g3
82. Kd5 Kf3
83. Kxc4 g2
84. Rxg2 Kxg2
85. Kd5 Kf3
86. c4 Kf4
87. c5 Kf5
88. c6 Kf6
89. c7 Kf5
90. c8Q+ Kf4
91. Qh3 Kg5
92. Qf3 Kh4
93. Qg2 Kh5
94. Ke5 Kh6
95. Kf6 Kh5
96. Qh3#
For anyone who still believes this game is a clear draw for Black, feel free to post which move Black made a mistake on, the correct move, and possibly a sample line towards the position in which White can make no progress, to support your claim. I would be happy to search for a refutation. |
|
| Sep-30-09 | | Keith Dow: Keith Dow: Keith Dow: Great Analysis. However I have a better move for Black. 45. ...Qe3
46. Qxd6 Re8
47. Qd7+ Re7
48. Qc6 Qxe4
49. d6 Qd3+
50. Kg1 Re8
51. Ra1
51. .......Qxc3
52. Qc7+ Kg8
53. Ra7 Qc1+
54. Kh2 Qf4+
55. Kh1 Qf1+
draw |
|
| Sep-30-09 | | newton296: < JamesMazur2: OK, I have my final analysis :) 45. ...Qe3
46. Qxd6 Re8
47. Qd7+ Re7
48. Qc6 Qxe4
49. d6 Qd3+ >
move 49 is already wrong !
try for black 49... Qxf5+ (not Qd3+? this moves throws away the draw) |
|
| Sep-30-09 | | JamesMazur2: <Keith Dow: I have a better move for Black.> Good try. It takes actually takes a few moments before Rybka finds 52. Rd1!. Not quite drawn anymore. I will post a continuation in a few hours (I would now, but I have a prior commitment to go somewhere). <newton296: 49. ...Qxf5+> I am aware of this line. To your credit, I can not guarantee it won. I have an analysis of it I posted earlier on, on this game's page, if you care to learn a little about it. However, I know it is flawed, so don't bother posting corrections. I will post myself a new and much improved analysis of 49. ...Qxf5+ in a few hours. |
|
| Sep-30-09 | | Capabal: <JamesMazur2: This leads into my 5th comment: I think Kasparov, the Deep Blue team, and Daniel King should be notified of this, even if they don't care. The truth, in my opinion, should be common knowledge to all.> The truth so far seems to be that a draw has not been found. I think they all know this. Had an inevitable draw been found, GK would certainly show it. |
|
| Sep-30-09 | | JamesMazur2: <Capabal>
I do tend to rant and rave, don't I.... |
|
| Sep-30-09 | | JamesMazur2: <Capabal: I think they all know this.>
Your argument makes good sense. My only question is then about this page: http://main.uschess.org/content/vie... Joel Benjamin mentions "the final move 45.Ra6? allowing black a perpetual check draw." This was in 2007. |
|
| Sep-30-09 | | Capabal: <JamesMazur2:>
I think Benjamin was wrong in 2007. There is no perpetual I've been able to find.<newton296: 49. ...Qxf5+> After 49... Qxf5+
Rybka:
50. Kg1 Rd7 51. Qxb5 Qe6 52. Qc5 Kg6 53. Rc6 e4 54. b5 e3 55. Qxc4 Qe5 56. Qe2
Ra7 57. Ra6 Rd7 58. c4 Kf7 59. Ra3 Qxd6 60. Qxe3 Qd1+ 61. Kh2 Qd6+ 62. Kh1 Qc7
63. Qb3 Qc5 64. b6 Rb7 65. Ra6 f5 66. Qd3 Kf6 67. g3 Ke7 68. Kg2 g6 69. Qc3 Ke6 70. Qb3 (+1.40) |
|
| Sep-30-09 | | Capabal: Another thing worth commenting is to what extent this whole problem is important and how much would it matter anyway if a forced draw was indeed found deep down. The final position does look somewhat drawish, and no doubt may GMs might have agreed to a draw at that point or later. But Kasparov resigned. And he did so because he knew that to achieve that hypothetical draw (if it existed) he was in for a long session of hard work against a machine that would take advantage of any inaccuracies in his play, and that would not get tired. So I am not sure how much it would really matter even if a forced draw was found to exist. The fact that he did not evaluate it that way would be no different from his committing an error that caused him to lose the game. And if he did think it might be a draw, but was not up to the task of trying to attain it, we are not further advanced. However you look at it, he lost the game. The whole thing seems all the more irrelevant in light of the well known fact that a few years down the road the top chess engines demonstrated uncontroversially that they can beat anyone even on a personal computer. |
|
| Oct-01-09 | | JamesMazur2: <Capabal>
Yes, I know it's irrevelant. I just did the analysis because I had nothing better to do, and to show people that Black doesn't have an "easy" draw, as was assumed for a while. |
|
| Oct-02-09 | | JamesMazur2: <newton296>
Examine this line after your suggested 49th move. Where is Black's mistake? 49. …Qxf5+
50. Kg1 Rd7
51. Qxb5 Qe6
52. Qc5 Kg6
53. Rc6 e4
54. Qd4 e3
55. Qxc4 Qe5
56. Rc5 Qxd6
57. Qe4+ Kf7
58. Qxe3 Re7
59. Qd4 Qxd4+
60. cxd4 Rd7
61. b5 Ke6
62. Rc4 Rb7
63. Rb4 Rb6
64. Kf2 Kd5
65. Ke3 f5
66. Rb1 Kd6
67. Kd3 Kd5
68. Rf1 g6
69. Rb1 f4
70. Re1 g5
71. Rb1 f3
72. gxf3 Rf6
73. b6 Rxf3+
74. Kd2 Rf8
75. b7 Rb8
76. Kd3 h5
77. Rb4 Kc6
78. Ke4 Kd6
79. Kf5 g4
80. hxg4 hxg4
81. Kxg4 Kc7
82. d5 Kd6
83. Rb5 Ke5
84. Kg3 Kd6
85. Kf4 Kc7
86. Ke5 Re8+
87. Kd4 Kb8
88. d6 Rh8
89. Kc5 Rh5+
90. Kc6 Rxb5
91. Kxb5 Kxb7
92. Kc5 Kc8
93. Kc6 Kd8
94. d7 Ke7
95. Kc7 Ke6
96. d8Q Kf5
97. Qd4 Kg5
98. Kd6 Kf5
99. Ke7 Kg5
100. Ke6 Kg6
101. Qg4+ Kh6
102. Kf6 Kh7
103. Qg7#
<Capabal>
I started by looking at your analysis.
49. ...Qxf5+
50. Kg1 Rd7
51. Qxb5 Qe6
52. Qc5 Kg6
53. Rc6 e4
54. b5 e3
55. Qxc4 Qe5
That much looked good at first. However, after 56. Qe2 you (and Rybka) missed 56. ...f5! This gives Black very good drawing chances, as I could see after examining a few lines. I don't blame you for missing it; it was a very deep threat. In fact, the threat was so deep that I could not find a good move 55 for White either! So I changed move 54 to Qd4 (against Rybka's wishes) and there is my analysis. |
|
| Oct-04-09 | | JamesMazur2: <Keith Dow>
51. ...Qxc3
52. Rd1 Qb3
53. Rf1 Qe3+
54. Kh2 Qa7
55. Qxb5 c3
56. Qc4+ Kf8
57. Qxc3 e4
58. Qc7 Qxc7
59. dxc7 Ke7
60. b5 Kd7
61. b6 e3
62. Kg1 Kc8
63. Rb1 e2
64. Kf2 Kb7
65. Ke1 Re5
66. Rb4 Re8
67. Rd4 Kc8
68. Rd2 h5
69. Rd4 Rg8
70. Kxe2 Re8+
71. Kd3 Rh8
72. Kc4 Re8
73. Kc5 Kb7
74. Rd8 Re5+
75. Kd6 Ka6
76. c8Q+ Kb5
77. b7 Ka5
78. Qc3+ Kb6
79. b8Q+ Ka6
80. Qc6+ Ka5
81. Qca8# |
|
| Jan-14-10 | | mrandersson: Can any 1 tell me how many moves does rybka concur with that of deep blues in this game? is Be4 as strong as garry said it was in the game over film? |
|
| Apr-03-10 | | MartinII: after Depth=20, Rybka don't put 45.Ra6 of Deep blue on the top 5 best moves. The best move instead of 45.Ra6 is 45.Qd7+! (see analysis below) 22 03:45 52.921.924 277.949 +1.18 Qd7+ Kg8 Ra7 Bf8 Qf7+ Kh7 Ke2 Rd8 Qb7 Rb8 Qxb6 Rxb6 Bf3 Bd6 22 03:45 54.152.194 278.200 +0.83 Qxb6 Rxb6 Ra7+ Kf8 Ke2 Bb8 Rd7 Bd6 Kd2 Ra6 Kc1 Kg8 Bf3 Kf8 22 03:45 56.518.878 278.250 +0.83 Ra1 Ke7 Qxb6 Rxb6 Ra7+ Kf8 Ke2 Bb8 Rd7 Bd6 Kd2 Ra6 Kc1 Kg8 22 03:45 57.964.410 278.148 +0.83 Ra2 Ke7 Qxb6 Rxb6 Ra7+ Kf8 Ke2 Bb8 Rd7 Bd6 Kd2 Ra6 Kc1 Kg8 22 03:45 61.240.446 278.055 +0.83 Ra5 Ke7 Qxb6 Rxb6 Ra7+ Kf8 Ke2 Bb8 Rd7 Bd6 Kd2 Ra6 Kc1 Kg8 The final position leads to draw after 45) ... Qe3!
<newton296:
move 49 is already wrong !
try for black 49... Qxf5+ (not Qd3+? this moves throws away the draw)> I don't see your argument. The topic here is that black (Kasparov) can hold a draw after a line starting with Qe3! assuming black plays the best moves. and YOU suggest Qf5 for Kas which is very weak move, hence white will naturally win. WHY do you give a weak move Qxf5 and let White (Deep Blue) win? I don't see your argument or maybe you're confused. The argument here is that Black can draw assuming he plays the best moves. Hence, final verdict, the final position of this game after 45.Ra6? is still a DRAWN position. Kas was just psychologically beaten according to journalists and GMs following this historic match. |
|
| Apr-03-10 | | MartinII: 
click for larger view 45. ... Qe3 46. Qxd6 Re8 47. Qd7+ Re7 48. Qc6 Qxe4 49. d6 Qd3+ 50. Kg1 Re8
51. Ra1 h5 52. Qxb5 Rd8 53. Kh2 h4 54. Qc6 Kg8 55. b5 Qg3+ 56. Kh1 Qxc3 57.
Rd1 Qc2 58. Re1 Qd3 59. Re4 Qxd6 60. Qxc4+ Qd5 61. Rxh4 Qxc4 62. Rxc4 Rd1+
63. Kh2 Rb1 =
click for larger view |
|
 |
|
|
< Earlier Kibitzing · PAGE 26 OF 29 ·
Later Kibitzing> |
|
|
|
Kibitzing
