Szabo should have accepted the pawn sacrifice with 16.Bxd6 Nxd6 17.exd6. Instead, he decided to keep some vague attacking chamces with 16.Bc3 and 17.Qh4.

<1. c4 Nf6 2. d4 e6 3. Nc3 Bb4 4. e3 O-O 5. Bd3 c5 6. a3 Bxc3+ 7. bxc3 Nc6>

The position is about equal. White has the two bishops, but most of his pieces sit on the back rank, he isn?t castled, and he has a doubled pawn on the c-file. White?s chances lie in completing development, castling, and eventually opening up lines for his bishops. White would like to play e4 at some point, taking over the center and opening up lines. Black?s chances lie in keeping the position closed, making the c-pawn a liability, making it difficult for white to play e4, and possibly exchanging one of the bishops.

<8. e4?!>

Ludek Pachman sees this as an error. White has not completed development, nor is his king safe, so Pachman sees this attack as premature. In the game, we find that black has a way to exploit white?s unsafe king, which will make it difficult for white to complete development.

Once white?s development is difficult to complete, we also find that white has a hard time coordinating his pieces for any sort of attack, whereas black?s pieces end up well coordinated for both attack and defense.

<8? cxd4 9. cxd4 Nxd4 10. e5>

We see the point of white?s 8. e4. With 10. e5, white intends something like 10? Ne8 11. Bxh7+ Kxh7 12. Qxd4, winning back the pawn. The ideas do make sense. With these moves, white would create a hole in black?s kingside, and he would expose black?s backward d-pawn. If black tried to develop his bishop by playing ?b6, then c5 would chip away at black?s queenside.

The problem is that white fails to account for black?s intermezzo, which will make it difficult for white to find kingside safety and to complete development.

<10? Qa5+! 11. Kf1>

The point is that 11. Bd2? is met with 11? Qxe5+, winning a second pawn.

But note that now, black would do badly by playing 11? Qxe5, because after 12. Bb2! black is in trouble. The knight is going to get captured, and white will open up the a1-h8 diagonal, which may prove difficult for black to deal with.

<11? Ne8 12. Bd2 Qd8!>

And not 12? Qc7. By playing his queen to d8, black deprives white of an important counterattacking resource. You can see it if you calculate what white is intending in this position.

<13. Bb4 d6 14. Bxh7+?! Kxh7 15. Qxd4 a5! 16. Bc3>

The point of 12? Qd8! If the black queen were on c7, then white would have the resource 16. exd6, hitting the black queen, and allowing the bishop to remain on the annoying a3-f8 diagonal. As it stands, the bishop is forced to retreat. The a1-h8 diagonal is blocked, thanks to white?s pawn on e5, and the c1-h6 diagonal doesn?t grant white any immediate strategic advantage.

Now is a good time to reassess what is going on, strategically. White has lost the two bishops and remains with a ?bad? one. His king has lost castling rights, and his pieces remain undeveloped. The knight can come out to f3 at some point, but it?s hard to see how to develop the h-rook. His pawns in the center are disconnected, so they may become long-term liabilities, especially the c-pawn on the half-open file. Usually, white would want compensation for these weaknesses, something like heightened piece activity. But, again, it is difficult for white to develop his h-rook in this position. He is clearly worse.

Logical. The idea is to remove white?s influential e-pawn, which is currently keeping black?s knight hemmed in. If 17. exd6 now, the f-pawn is keeping the dark-squared bishop from looking directly at the g7-pawn. If 17. exf6, black has some ideas of maybe 17? Qxf6 18. Nf3 Bd7 14. Qxf6 Nxf6, and black is better in the endgame because his pawn structure is better, and with both his rooks still on the board, he can generate threats against white?s king.

The move also turns out to be useful, because it clears the f7-square for white?s king, which, if white commences an attack on the h-file, will turn out to be useful. Note that the light-squared weakness around black?s king will be difficult for white to exploit: white does not have his light-squared bishop, and white?s knight will not be able to post up on g5, the ideal square for a knight to exploit light-squared weaknesses. The best white can hope for is to post his knight on f4, but it looks like black will be able to play ?dxe5 soon, controlling the f4-square and locking the knight out of the attack.

<17. Qh4+>

White begins an attack on the h-file, his only advantage in this position. If black can manage to defend, then he will likely win.

<17? Kg8 18. Rd1>

Preparing the rook lift Rd1-d3-h3.

<18? Qc7>

Obviously, 18? dxe5?? is not immediately playable, due to the pin.

<19. Rd3>

On 19. exf6, Ludek Pachman gives 19? Nxf6 followed by 20? e5, when white?s bishop is blocked from attacking black?s king, and black will be able to generate an attack quickly (his bishop would no longer be hemmed in).

<19? dxe5 20. Rh3>

Let?s assess the position again. White has managed to put his queen and rook on the h-file. But his bishop is completely blocked out of the attack, and redirecting the bishop via Bd2-h6 does not seem to hold much promise. White?s kingside remains undeveloped, he is down a pawn, and it?s not clear how white will manage to keep his king secure.

Meanwhile, black has managed to secure a central pawn majority. The e-pawn is doubled, but black?s main strategy is not to push central pawns ? they?re there to keep black?s minor pieces at bay ? so the fact that they?re immobile is not such a big weakness. It?s hard for white to meaningfully attack the e6-pawn, so black?s pawn structure can only be weakened by moves like f4 and g4-g5, which are going to further weaken white?s king position. In addition, black?s knight is no longer hemmed in by a white e-pawn, black can continue to improve his position by playing ?b6 and ?Ba6/?Bb7, and moving rooks to the c- or (newly opened) d-file. Black?s king is not the safest, but it does have the f7-square to retreat to. Further, and probably most important, black?s pieces are becoming coordinated. It?s hard to see how white ever gets his rook on h1 to coordinate with the rest of his pieces.

This position was a pretty natural fallout after Smyslov found 10? Qa5. White?s king was going to be stuck on f1, which was going to make it difficult for white to develop his h-rook. White could have tried, perhaps, pushing his h-pawn to h4 and h5. While white was trying to coordinate an attack on the kingside, he lost his light-squared bishop, which will turn out to be very advantageous for black, because he managed to retain his own light-squared bishop. The big thorn in black?s side was the pawn on e5, but while white was maneuvering his rook via Ra1-d1-d3-h3, black managed to remove the pawn from e5. White, in essence, ignored all of black?s plans, while black made moves to blunt white?s attack (which black identified was white?s only advantage in this position). Black managed to build an e5-e6-f6-g7 pawn structure that completely locks both white?s bishop and white?s knight out of the attack.

If the c-pawn falls, it will be almost impossible for white to defend. The knight is forced on the defense, further uncoordinating white?s pieces.

<22. Nd2 Rc8>

The c-pawn must fall now. Note that black can afford to go into an endgame in this position ? he?s going to be two pawns up, both rooks will still be on the board, and his bishop will be better than white?s bishop.

<23. Qh7+ Kf7 24. g4>

Necessary, or else white has absolutely no chances.

<24? Bxc4+ 25. Ke1>

25. Kg1 or 25. Kg2 is met with 25? Bd5(+). White cannot play a move like f3 or Nf3 to block the bishop without leaving his own Bc3 en prise.

Because black did not launch a premature attack, he was able to permanently weaken white?s position early (forcing Kf1). Black could then continue defending, because his pieces were closer to home, and because he was able to strengthen the squares that he needed to strengthen (?a5, ?d6, ?f6, ?dxe5). Black was then in a good position to counterattack, because he managed to coordinate his pieces properly. He won two pawns, and now he is on a king hunt. Black wins by staying in the middlegame, but he also wins by transitioning into an endgame.

White launched a premature attack, so he found his pieces being batted away, he created several weaknesses in his own camp, and then he found it difficult to develop naturally. Because it was difficult for him to develop his pieces, his pieces could not coordinate for a meaningful attack. With all the weaknesses in white?s camp, and because his pieces remain badly coordinated, white is ultimately unsuccessful in mounting a good defense, either. Black lost his pawns because of his badly coordinated pieces, and because he could not generate meaningful counterplay (again, because of his badly coordinated pieces).

<25? Qd6 26. g5>

A last ditch effort.

<26? Qxa3>

Winning a third pawn, and now threatening ?Qc1#.

<27. Qc2 Bd5>

Hitting the rook and revealing a pin on the c3-bishop, while maneuvering the bishop to a square where it is protected (heavy pieces are bad at defending minor pieces and pawns, they more naturally are suited for attack).

<28. Rg1 Qa1+ 29. Nb1 Qa2>

Black is three pawns up and wins in an endgame.

<30. g6+ Ke7 31. Qc1>

White thinks his best chances are to stay in the middlegame.

<31? Nd6>

Eyeing e4.

<32. Rh7 Nf5>

Again, minor pieces are better suited for defense than major pieces. This is what happens when you develop properly ? your pieces can naturally find good squares for both offense and defense.

<33. Rg3 b5>

And not 33? Nxg3?? 34. Rxg7+, when white suddenly has counterplay on the kingside, and his g-pawn suddenly looks menacing. Black, not falling for the trap, intends to win by pushing his queenside majority.

<34. Rd3 b4 35. Rd2 Qc4 36. Rc2 Qe4+ White resigns.>

White?s bishop is about to fall. If 37. Re2, then 37? Qh1+, forcing an exchange of queens.