The black rook is trapped.

56...Kg7 57.Kh4 Kf5/Kh5 58.b3

Now white waits patiently for black to exhaust all the pawn moves and finally forced to give up its rook with ...Rxg6

56. Bg6 +3.58
56. g4 +3.00

38. f7! is the only move to give White winning chances as the alternatives allow Black to equalize.

For example, the game move 38. Qf4?! allows 38. Qxg6! which levels out after 38... Qxg6! 39. f7 Qb1+ 40. Kh2 Rh3+! 41. gxh3 Qxb2+ 42. Kg3 Qxa3+ 43. Kxg4 Kg7 = (0.00 @ 27 depth, Deep Fritz 14).

After 38. f7!, Fritz gives best play as 38...Qf8 39. Qf4 Re1+ 40. Kh2 (+1.46 @ 27 depth), which, coincidentally, transposes to the game after 40. f7 .

Of course it's better to arrive at the game's 40th move via 38. f7! Qf8 39. Qf4 Re1+ 40. Kh2 , since 38. Qf4?! is a mistake which allows Black to equalize with 38...Qxg6! =.

I saw the Qxg3# threat unless something forcing was done, and that I could easily trade queens and then be at a B+2P vs. R material difference, which according to point count is an equal game.

At this point, I'd be guessing as to a way to promote one pawn, and win the game. But great players like Tal can quickly see that the Rook is very limited. It's safe on g5 as long as the King is on h6.

And somebody like me stops thinking about winning the rook since I can't. Heck, I might even move the g3 pawn up before I need to skip a turn on the queenside, thereby losing the game.

But Tal thinks "I've got the rook trapped. He can't move to the d-file because of his own pawn, the e-file because of my pawn, the f-file because of the Bishop, and the h-file for the same reason. He can take the Bishop at g6, but then the pawn takes and the g3 pawn is decisive. Yeah, my Bishop also can't move, so the game comes down to who can get a passed pawn on the a or b files. I have a spare tempo to spend on g4, so I'll win that little game or we'll lock up, forcing him to move the K or R."

One example is 57...Kf6 58 g4 a6 59 Bd1.

click for larger view

Now if 59...Rg8, then 60 h6, etc.

click for larger view

You need a good sense of humor to understand chess joke.

Zugzwang required a lot of calculation, this is why I didn't see the solution.

But watching the game I found something interesting.

This is the pos. white to play 21.

click for larger view

Here Tal was in hurry and played Qh6,and couple moves later the queen returned back on h4.

So, I decided that white should <lure> black queen from "f8" square.

Let's try 21.Nxd5. If white can put the knight on e7, white will be better.

After 21...Qxd5 we have forced line.
22.Qh6 Ne6 23.Bf5
if gxf5 24.Rf4 and Rh4 if Qd6 24.Bxe6 Qf8 25.Qxf8 Kxf8 26.Bxd7 and white is a piece up.

After 24...Bc6! (diagram below) in this line,

click for larger view

Andy Soltis had a column about this game back in the late 1980's, as I recall.

In other words, either move works.>

Just for laughs, I plugged the position into Houdini 4. For a while, the two moves were evaluating about the same (with a slight preference for 55.g4). But eventually, 55.Bg6 took a sizeable lead:

Houdini_4_x64B @ 35 ply:
+12.16 56.Bg6 Kg7 57.Kh4 Kf6 58.b3 a6 59.g4 Rxg6

+7.05 56.g4 Rg7 57.Kg3 Rf7 58.Kf4 Rc7 59.g5 Rc1

+0.78 56.Bg4 Rg7 57.Kh4 Rf7 58.h6 Rf2 59.Kg5 Kf7

Regardless, one of the big differences between engines and humans is that engines are equally happy with a winning position whether it has complications or not.

Well, yes and no. I found 56. g4, saw that it was winning and so stopped looking for better moves. That's laziness on my part, but I think it's also a pragmatic way to play the game.

The interesting point is defining what constitutes a "simple position that is clearly winning". For me, 56. g4 was precisely that. A rook cannot hold back two connected passed pawns with a bishop and king in close attendance. It's a simple and clear win.

It's an instance when judgement takes over from calculation. Incidentally, there is a similar instance in the 56. Bg6 line. Black can throw the rook away immediately with 56. Bg6 Rxg6 57. hxg6 Kg7

click for larger view

White's advantage is only a pawn (after black takes on g6), but any decent player will be able to see that this is an easy win. The white g pawn will keep the black king busy while the white king wanders over to the queenside and sweeps up all the pawns.

Is 56. Bg6 stronger than 56. g4? It looks more convincing because the rook trap appeals to our more basic materialistic instincts. But it isn't a true rook trap because Black can exchange his rook for pawn and bishop.

Perhaps it's purely subjective then and a minor issue, but as black, I'd resign immediately if I faced the forced king and passed-pawn game that follows 56.Bg6, e.g.

click for larger view

After 56.g4 Rg7, I still have my rook and time to deliver some checks, with barely enough complications to hope for a white blunder.

For example: 56.g4 Rg7 57.Kh4 Rc7 58.g5 Rc1 59.h6 Rh1+

click for larger view

...including a detailed account of the mixup regarding the adjourned position, which ended up requiring the game to be won "twice" on Tal's part!

Just got an email about the book, including an excerpt giving this game.

click for larger view

The next day the game had to be resumed, but one of the organizers, whose name I don't want to remember but who was an absent-minded fellow, placed the pieces in the following incorrect position, and apparently neither of the players noticed that White's Q-side pawns had been placed in different squares:

click for larger view

At 6:00 PM in the afternoon the game was resumed, and finally at about 11:30 at night Black resigned. Tal behaved very nicely and during the post-mortem analysis showed his young opponent the correct defense, and pointed a out a way where Black could have drawn this ending.

Both players were very tired, and they were about to go to sleep, when suddenly somebody noticed that the position where they started to resume the adjourned game had been wrongly placed. What to do?