In ancient times (i.e., pre-2000), an average chess player had the hope of going one-on-one with the best in the world and winning. Not at the board, but in the pages of a book or magazine. A game would be published, often with analysis appended, and any player could look for holes in the game or, especially, in the analysis. Finding holes in analysis, or cooks in problems, was a great joy.
These days that joy is largely gone. One still sees letters to the editor in magazines pointing out occasional mistakes, but the advent of extremely powerful computer engines has largely killed the fun, especially in quality books.
But I remember some of the old joys, having found a couple of problems here and there myself. Tonight I'll share one of them. It concerns the game Tal-Tolush from the 1956 Soviet Championship, and the story features Tal himself, the Hungarian chess writer Jozsef Hajtun, me (naturally), Garry Kasparov, and the programs Fritz 7 and Stockfish 6.64.
In 1956 Mikhail Tal made his debut in the Soviet Championship Final. He finished in shared 5th-7th with Polugaevsky and Kholmov at 10.5/17, a half-point behind Korchnoi in 4th, and a full point behind Averbakh, Spassky and Taimonov, who tied for first.
In the last round, Tal, as white, faced Alexander Tolush, who is probably best known in the West as Boris Spassky's one-time trainer. Tolush was a fine attacking player in his own right, and not one to back down from a fight. The game was a Sicilian Najdorf Poisoned Pawn variation, Tolush took the b2 pawn, Tal attacked, and in Tal's own words "I nevertheless succeeded with a rather complicated combinative attack."* Tal then gives the position after Black's 14th move and the rest of the game score with some light notes.
But the first time I saw the game wasn't in Tal's book, but in a book originally published in 1961, Selected Chess Games of Mikhail Tal by Jozsef Hajtun**. This book, likely published to take advantage of the popularity of the new young World Champion, contained 64 games by Tal, ranging from the 1956 Soviet Championship to Tal's first match with Botvinnik in 1960. The notes aren't particularly heavy, especially by current standard, but they were all I had! I'm pretty sure I bought this book from Duga's Book Store on Silver Star Road in 1984. This, and the three games against Tal that Fischer included in My 60 Memorable Games (a first edition, second printing of the original hardback, also bought from Duga's) were my first real introduction to Tal. And just like most everyone else, I was blown away by what I read!
(I really don't know that younger players can possibly appreciate how hard it was to get a hold of games in the old days. These days everyone has a database with five million+ games, or can look the games up online. Back then, you had it in a book or magazine, or you didn't have it. And very few books or magazines were available in most of the US, not in book stores or libraries. Kids have it easy these days, dagnabit!)
Naturally, I started at the front and began working my way through. I had decided after reading about Emanuel Lasker that I liked complicated games, so this was right up my (imagined) alley! The subject of this post is Game Four in Hajtun's book, so let's get to it!
Like Tal, I'll pick up the position after Black's 14th move. (If anyone is interested in 1950s era Sicilian Najdorf Poisoned Pawn theory, then you've come to the wrong place!)
Tal - Tolush, 1956 Soviet Championship
Black has just played 14...Qa4
rnb1kb1r/1p1n1ppp/p3p3/4P1B1/q2NN3/1R6/2PQ1KPP/5B1R w kq - 0 15
15 Bb5
Tal and Kasparov*** make note in their books that 15 Nxe6! gives White a decisive advantage, and that this had been discovered later. I can't seem to find any other games with this position in my database, though, so I don't know when this knowledge became well known. It appears Hajtun did not know the move 15 Nxe6 - he gives Tal's move an '!' and writes, "A beautiful move introducing a whole sequence of forcing combinations. From now on Tal does not let his opponent get a word in, but continues on his course with the utmost accuracy."
15 ... axb5
16 Nxb5
Kasparov: ('hinting' and Nc7 mate)
16 ... f6
17 exf6 gxf6?
Tal preferred 17...Nxf6, which he gives an '!', as does Hajtun. Both consider 17...Qxe4, but Tal dismisses the move. Kasparov delves deeply into this position and has variations that run past move 30! He determines that Tal's suggestion is insufficient. Kasparov continues:
In my view, it was best to eliminate the centralised knight immediately, in order to then combine defense and counterattack: 17...Qxe4 18 fxg7 Bc5+ 19 Kg3.
Analysis diagram from Kasparov
rnb1k2r/1p1n2Pp/4p3/1Nb3B1/4q3/1R4K1/2PQ2PP/7R b kq - 0 19
It is not hard to guess why Tolush, who was also a brilliant master of attack, avoided this continuation: the black king faces a very dangerous journey and it appears the it will be unable to withstand the onslaught of the splendidly mobilised white army. Indeed, that was the opinion Tal and other commentators after the game. However the gloomy evaluation, given by 'man', proves incorrect when the 'machine' is switched on. In the end, Black's two(!) extra pieces and the open position of the white king outweigh White's strong, but temporary threats....Kasparov then give almost a page and a half of detailed analysis to evaluate the position and prove his point! Yikes! I'm not going to reproduce that here, but I'll suggest this might make for a fine Stoyko Training Position. Back to the game:
Position after 17...gxf6?
rnb1kb1r/1p1n3p/4pp2/1N4B1/q3N3/1R6/2PQ1KPP/7R w kq - 0 18
18 Re1! Ra6!
19 Bxf6 Nxf6
20 Nf6+ Kf7
21 Rf3
Kasparov: "Terrifying, wouldn't you agree?"
21 ... Qh4+
Hajtun: "21...Qxb5 is answered by 22 Nd5 dis. ch, and the attack can easily be forced home. Surprisingly enough after 21...Bc5+; the king would be safe of g3."
22 Kf1 e5
Kasparov gives Black's last an '!', pointing out that it improves the defensive scope of the Ra6 and the Bc8.
23 Qd5+ Be6
Kasparov: "Again, the only defense."
Position after 23...Be6
1n3b1r/1p3k1p/r3bN2/1N1Qp3/7q/5R2/2P3PP/4RK2 w - - 0 24
This position is just begging for a kill-shot. And this is about where I entered the story back in the mid-1980s.
24 Nd7+
Hajtun gives this a '!' and writes, "The complications reach their peak. Black can only escape by yielding material."
24 ... Kg6
Hajtun: "The lesser of two evils. If 24...Ke7; 25 Qc5+, Kd8 (25...Ke8, 26 Qc8+); 26 Qc7+."
But this didn't look right to me. I couldn't see how White could finish Black off in the 25...Ke8 26 Qc8+ line. It turns out I was right, as later confirmed by Kasparov and Stockfish. If 25...Ke8 holds, then some move other than 26 Qc8+ is needed. And I eventually found a great move: 26 Nf6+!
Analysis position after 26 Nf6+!
1n2kb1r/1p5p/r3bN2/1NQ1p3/7q/5R2/2P3PP/4RK2 b k - 0 26
It's brilliant! Black has three legal moves: 26...Kd8 runs into 27 Qc7#. And 26...Kf7 runs into 27 Ng8+. Ouch. The Bf8 has now been denied one of its defenders and must fall. With White's queen, knight, and both rooks in close proximity to the black king, the king must surely get mated. And that left one option, 27...Qxf6, and here I stopped my analysis, assuming White would win the black queen and thus the game.
But there's a problem. Two, in fact. I had written some of the above notes down in my book. (I almost never write in a book, so that alone tells me I was serious about this!) Sometime later, I don't know when, I looked at my analysis again and figured out that I hadn't looked at all the lines deeply enough for Black after 27 Ng8+. I saw that 27...Ke8 28 Rxf8+ led to a quick demise, and 27...Kxg8 28 Qxf8#. The line with 27...Kg7 28 Qxf8+ takes a few moves, but that's also a pretty straightforward win. However, 27...Kg6 gives Black a won game!
Analysis heartbreak
1n3bNr/1p5p/r3b1k1/1NQ1p3/7q/5R2/2P3PP/4RK2 w - - 0 28
As ridiculous as it seems, Black's king is safe on g6, and Black will have time enough to reorganize his pieces and consolidate his advantage in material. I added the following note in my book, "27 Ng8+ Kg6 is the winning move for Black but all others win in beautiful fashion for White." What spoilage!
I might have suspected other errors 30 years ago (I was, after all, only a 1500 ELO punk back then, as opposed to my much more distinguished 1700 ELO status now), but it wasn't until some time back when I put my analysis into Chessbase that I confirmed other flaws. The most important one is that 27...Qxf6 most likely leads to a draw with correct play. That's according to Stockfish. But the lines (which I'm not going to bother reproducing here - plug them in to your own computer if you need to see them) look like they might favor Black in a game between humans - and awful lot can go wrong for White in a hurry as he's running out of pieces.
Worse still, it seems that if Black does play 27...Kf7, White can now win with 28 Ng4+. But this conceals an interesting point about the evolution of the game. When I first checked this when I entered the game, my most powerful engine was Fritz 7. It evaluated 28 Ng4+ as a slight plus for White, about +1.00 if you let it run awhile. (In a position such as this, that's only a slight plus!) This summer, I upgraded to the then most recent build of Stockfish, version 6.64. Stockfish tells me in very short order that 28 Ng4+ is crushing, at least +4.00. This is a sign of just how powerful the engines are getting - constant re-evaluation of old positions will continue to be necessary for some time, I suspect.
Back to the game:
Position after 24 ... Kg6
1n3b1r/1p1N3p/r3b1k1/1N1Qp3/7q/5R2/2P3PP/4RK2 w - - 0 25
Kasparov in his book knew Black's last move to be losing and gave it a '?', for it turns out that 24...Ke7 does in fact draw, as we partially demonstrated above. Hajtun doesn't give much in the way of notes after this, so we can ignore him the rest of the way. Tal's notes for the game in his autobiography were characteristically brief (at least, characteristically so for his autobiography), but Kasparov does get one more interesting note in along with some other minutia. The game continued
25 Ne5+ Kg7
And here Kasparov writes, "It was worth trying to confuse matters with 25...Kh5, although I have no doubt that Tal would have found 26 Qd1!! (with the threat of discovered check, the 'queen plus rook' battery carries a nuclear charge!)," before concluding with more variations. That was a nice turn of phrase about the queen plus rook battery!
The game wrapped up with
26 Rg3+ Qxg3
27 Qxb7+ Nd7
27 Qxb7+ Nd7
28 hxg3 Rb6
29 Qc7 Bc5
30 Nxd7 Bc4+
31 Re2
1-0
1-0
Black overstepped the time limit.
So, I had figured out that Hajtun had been wrong about 24...Ke7 25 Qc5+ Ke8 26 Qc8+ winning for White, but my attempt to resurrect the position for Tal with 26 Nf6+ and 27 Ng8+ had failed. So only partial credit at best.
A few other interesting points, though.
First, the difference in evaluations of the same position by programs just a few years apart is striking. Stockfish 6.64 really blows Fritz 7 clean out of the water, not only with its assessments, but the speed in which it reaches its conclusions. Constant reassessment of old analysis may be more important now than ever.
Second, Tal's early career was a demonstration that chess was indeed a contest between two individual human beings, and this game illustrates that well. Even against another brilliant attacking player, Tal was able to keep up constant pressure by creation of one complex problem after another for his opponent to solve. Eventually the opponent crumbled, first in missing 17...Qxe4, and later by missing the drawing move 24...Ke7. The human element should never be ignored.
But it does seem a shame that Tal couldn't win the game on the merits following 18 Re1! After 24 Nd7+ Kasparov wrote, "Usually such checks kill instantly, but here by some miracle Black is still alive."
...
And that's all Kasparov wrote about White's 24th move. Just that single line.
Kasparov wrote his book in 2002/2003, and had all the latest programs at his disposal. He had a vast amount of research material available. Garry was the number one player in the world at that time. The best player in the history of the game put a lot of effort into this book, and he had his team and editors helping him out. And he didn't find anything more important to say about White's 24th move than a single line of regret.
Tal hadn't said anything about it, either, in his autobiography published in the 1970s. Hajtun had been too busy praising White's 24th move in 1960/61 to look for anything else. And obviously during the game, Tal didn't come up with anything better than 24 Nd7+.
But this is precisely where he gave away the win! Let's look at it again:
Position after 23...Be6
1n3b1r/1p3k1p/r3bN2/1N1Qp3/7q/5R2/2P3PP/4RK2 w - - 0 24
I wrote above that there just HAS to be a kill shot in this position. And there is! Tuesday evening, when I first started writing this post, I turned on Stockfush to check a couple of things, and left it running here for a while when I had to do something with my daughter. When I came back, Stockfish told me that the winning move here was
24 Nh5+!!
Yes, I'm going to give this two exclams - if the move eluded everyone from Tal to Kasparov for more than 46 years then it has EARNED those exclamation points!
The lines are very complex, and possibly beyond the powers of all but a few humans to calculate. I really couldn't say. But looking at them it seems the big difference is that Black cannot move to either the e7 or e8 squares in this line because a white rook will settle on the d-file eventually, making the d8 square extremely vulnerable. Here's a sample line: 25.Qc5+ Ke8 26.Qxe5 Qc4+ 27.Rd3 Nd7 (for the sake of example, say Black plays 27...Rb6 instead. Then the weakness on the d-file is exposed: 28.Nc7+ Kf7 29.Qf6+ Kg8 30.Nxe6 Rxe6 31.Rxe6 Qg4 32.Rg3 Qxg3 33.Re8 Qf4+ 34.Nxf4 Nd7 35.Qe6+ Kg7 36.Nh5#) 28.Qxh8 Qxb5 29.Ng7+ Kf7 30.Nxe6 Rxe6 31.Qxh7+ Kf6 32.Qxd7 Qxd7 33.Rxd7 et cetera. Here's the final position for that line:
The lines are very complex, and possibly beyond the powers of all but a few humans to calculate. I really couldn't say. But looking at them it seems the big difference is that Black cannot move to either the e7 or e8 squares in this line because a white rook will settle on the d-file eventually, making the d8 square extremely vulnerable. Here's a sample line: 25.Qc5+ Ke8 26.Qxe5 Qc4+ 27.Rd3 Nd7 (for the sake of example, say Black plays 27...Rb6 instead. Then the weakness on the d-file is exposed: 28.Nc7+ Kf7 29.Qf6+ Kg8 30.Nxe6 Rxe6 31.Rxe6 Qg4 32.Rg3 Qxg3 33.Re8 Qf4+ 34.Nxf4 Nd7 35.Qe6+ Kg7 36.Nh5#) 28.Qxh8 Qxb5 29.Ng7+ Kf7 30.Nxe6 Rxe6 31.Qxh7+ Kf6 32.Qxd7 Qxd7 33.Rxd7 et cetera. Here's the final position for that line:
Analysis after 33 Rxd7
5b2/1p1R4/4rk2/8/8/8/2P3PP/4RK2 b - - 0 33
The final position looks reasonable enough, but getting there is murder, especially since that isn't even Black's best first move! Probably the best line for Black after 24 Nh5+!! is 24...Kg6. A sample line:
24. Nh5+!! Kg6
25. Qxe5 Bg4
25. Qxe5 Bg4
26. Ng3 Qg5
27. Qe8+ Kh6
28. h4 Qxh4
Analysis position after 28...Qxh4
1n2Qb1r/1p5p/r6k/1N6/6bq/5RN1/2P3P1/4RK2 w - - 0 29
Why 28 h4 Qxh4? To put the black queen on a square vulnerable to a knight fork should the N@b5 make its way to f5 with check. Don't believe me? Check it out yourself!
The analyse fantastique continues with
29. Nd4 Rf6
30. Qe4 Rxf3+
31. Nxf3 Qxg3
32. Qe3+ Kg6
33. Qg5+ Kf7
Analysis position after 33...Kf7
1n3b1r/1p3k1p/8/6Q1/6b1/5Nq1/2P3P1/4RK2 w - - 0 34
Only 34. Ne5+ maintains an advantage for White!
34. Ne5+ Ke8
35. Nxg4+ Qxe1+
Other moves lead to Black getting mated sooner rather than later.
36. Kxe1 Nc6
37. Qb5
An incredible position! Stockfish evaluates this at nearly +4.00 for White.
Analysis position after 37 Qb5
4kb1r/1p5p/2n5/1Q6/6N1/8/2P3P1/4K3 b k - 0 37
37. ... Bb4+
38. Kd1 Rf8
39. c3 Bd6
Taking the pawn is a bad idea: 39...Bxc3 40. Qxb7 Ne5 41. Nxe5 Bxe5 42. Qb5+ and White reaches an endgame with Q+P vs R+P.
40. Qxb7 Ne7
41. Qa8+ Kf7
42. Qe4 ...
And the h-pawn must fall. Allegedly best is:
42. ... h5
43. Qh7+ Ke8
44. Qxh5+ Kd7
Final analysis position after 44...Kd7
5r2/3kn3/3b4/7Q/6N1/2P5/6P1/3K4 w - - 0 45
I'll stop here. The point of showing this bit of fantasy is that this is one possible line, with lots of non-forcing moves by White (e.g., White's 26, 30 & 31 moves) and White's 28th would be a tough find, too. And White would have to really analyze four different moves after 24 Nh5+! Given the nature of the position, a lot of concrete analysis would be required of players of that caliber to make their moves with certainty. Lacking time, they'd have to go on instinct. It took decades to figure all of this out, the development of incredibly powerful computer programs (with many errors likely to be revealed by the computer programs of 2018, for example), and even after 46 years the greatest player of all time still couldn't get to the bottom of it. He didn't even suspect! It took more than a decade after Kasparov's once definitive analysis to see truth laid bare-ish.
So if you understood all that analysis, then go get your grandmaster title: not tomorrow, not after breakfast, NOW!
When I found this the other day I was pumped, because once again the joy of discovery was mine! And I do mean discovery, as Stockfish running on my computer had done the actual inventing. Still, I'll take what I can get, and being midwife to an engine of discovery isn't without some honor. Right? Right?!
Except it turns out I don't even get this consolation prize. GM Lubomir Kavalek had written a column about this very game back on 2014 June 27, and had identified 24 Nh5+!! as the winning move. Not only that, but Nh5+ appears to win on move 23, too. Curses and consternation! No satisfaction whatsoever, though I do think Kavalek's computer-assisted line is probably slightly inferior to what Stockfish provides above.
No, that's not entirely true about the lack of satisfaction. This HAS been fun. So I guess there's still some joy in cooking, even if one is just following someone else's old recipe....
...
NOTE: I will try to put all of this into the re-playable format used in an earlier post, but that will have to wait until another day. Or maybe I'll see if Paul can do it with Chessbase. I just want to get this monster posted before something else spoils my efforts!
...
* The Life and Games of Mikhail Tal, by Mikhail Tal, 1976 (I have the 1997 Cadogan edition.)
** Selected Chess Games of Mikhail Tal, by J. Hajtun (trans. by Robert Ejuri), first published in English by Sir Isaac Pitman & Sons, Ltd in 1961 (I've got the 1975 Dover edition of this one.)
*** Garry Kasparov On My Great Predessors, Part II, by Garry Kasparov, 2003 Everyman Chess
No comments:
Post a Comment