Like so many things in life, we will never really know what would have happened if things were done differently.

Black threatens 25... Rxc2.

The bishop on b4 controls d6, a good square for the knight on e4 to reinforce the attack. This suggests 25.Rxb4:

A) 25... axb4 26.Nd6

A.1) 26... Rxc2 27.Ngxf7

A.1.a) 27... g5 28.hxg5 with many threats: Nxb7, Nxd8, g6 and Qh6+, Qf6+, etc.

A.1.b) 27... Rh8 28.Nxh8 Kxh8 29.Nxb7 + /- [P].

A.2) 26... Rf8 27.Ra7 Rxc2 (27... Bxg7 28.Nxc8 Rxc8 29.Kxg2 Rxc2 30.Qd4 wins the pinned knight) 28.Rxb7 Rxb2 (avoids Qxb4 and Nf5+) 29.Qd4 with the threat 30.Ndxf7 Rxf7 31.Nxf7, etc.

B) 25... Rxc2 26.Rxb7 + - [N].

C) 25... Bxe4 26.Rxe4 Rxc2 27.Rxa5 Rxb2 28.Ra7 and White has eliminated Black's bishop pair and has improved his own pieces but it's unclear whether the advantage is enough to win.

-----

I tried this line:

25.Bb3 Nc5
26.Qf6+ Qxf6
27.exf6+ Kf8
28.Rxd8+ Rxd8
29.Nxc5 Bxc5
30.Rxa5 Bb6
31.Rb6 Rd6

Probably with good reason?

Actually Gurevich in his notes in Informator 59/547 says 25...Be4 is the only move. He was in time trouble when he played 25...ab4? 26. Nd6 . He gives the line 25. Rb4 Be4 26. Re4 Rc2 27. Ra5 Nb6

needless to say Gurevich - a Kasparov second and once elite GM - noted 27...Rb2 28. Ra7 was pretty hopeless for black.

28. Qf6+ Qxf6 29. exf6+ Kg8 30. Rb4 Nd5 31. Rb7 Nxf6 32. Rxf7 Ng4 33. g3 Rf8 34. Rxf8+ Kxf8 35. Nxe6+ Kf7 (35...Ke7 36. Nf4 Rxf2 37. Nxg6+ Kf7 38. Nf4 Rxb2 39. Nxh5 <+2.38 Depth of 15>) 36. Ng5+ Ke8 37. Ra8+ Ke7 38. Ra7+ <+1.65 Depth of 25>

I think you are most definitely correct in saying that Black has holding chances here, in this better endgame for White. This stated, without a doubt, going in for 25. Rxb4 was the best decision from GM Nikolic as it set strong practical problems on GM Gurevich.

An important thing to note at times when solving these tougher puzzles is that it is better to play the combination that leads to a a sensible and practically sound game, than trying to overpress the idea that the stellar win is there somewhere.

Of course there is nothing wrong with looking for that stellar win; but if it isn't there, then opting for a sensible game where one has the advantage is most definitely a strong/excellent course to take.

LTJ

PS. It should be noted that in order for an engine evaluation to be winning (i.e., or ), the side with the advantage must be "up" by at least 1.75.

(This is why I prefer when the evaluation is in the <2.25> range, like the 27...Rxb2 line; although sometimes even evaluations like this can be drawn).

(Part 1 of 6)

"Sometimes," as we used to say in Santa Cruz, "the exchange isn't all it's cracked up to be."

In this case, eliminating Black's dark-squared bishop is worth a sacrifice because it allows access to a succession of crucial squares — or forces Black to return the material to prevent it. The key invasion point is f7, but at present White hasn't enough attackers on it. This suggests the key move <25. ♖xb4!<>>. This leaves Black four plausible replies:

(1) 25. ...♕xb4??
26. ♕xf7†, ♔h6/♔h8
27. ♕h7#

Plausible at first glance but obviously wrong because it allows White everything he wants without a fight.

(2) 25. ...♖xc2?
26. ♖xb7

Black has lost a piece.

(3) 25. ...♗xe4!?
26. ♖xe4, ♖xc2
27. ♖xa5, ♖xb2
28. ♖a7

Although material is even, White's pieces are ideally placed for a continuing attack. Black may have an improvement on move 27, but nothing clearly suggests itself.

This brings us back to the crucial test:

(4) <25. ...axb4
26. ♘d6...<>>

Again, Black has four choices:

(4.1) 26. ...♖xc2
27. ♘gxf7...

White threatens mate in two, starting with 28. ♕h6†. There are three possible defenses:

(4.1.1) 27. ...♘xe5
28. ♕xe5†, ♔h7
29. ♘g5†, ♔g8

<(If 29. ...♕xg5?
30. ♕xg5, and Black can't capture on d6 because of the fork 31. ♕e7†.)>

30. ♘df7

To avoid mate, Black will have to give up queen for two knights, and his loose position assures that he will remain under attack.

(Part 2 of 6)

(4.1.2) 27. ...g5
28. hxg5, ♖a8

This appears to be Black's best bid to eliminate some of White's threats.

29. ♖d1...

Here is what makes an "insane" puzzle: Black, under no immediate compulsion, has four plausible ways to save his bishop:

(4.1.2.1) 29. ...♗d5?!
30. g6...

White's mate threat is renewed, forcing one of two choices:

(4.1.2.1.1) 30. ...♘xe5
31. ♕xe5†, ♔xg6
32. ♖xd5 or

(4.1.2.1.2) 30. ...♖h8
31. ♘xh8, ♔xh8

Black doesn't have to take this knight, and perhaps shouldn't; but it's hard to find a better defense.

32. ♕h6†, ♔g8
33. ♘f7

Black will have to give up queen for knight to avoid mate. This brings us back to

(4.1.2.2) 29. ...♘c5
30. g6...

White is threatening mate again, so Black's reply is forced:

30. ...♖h8
31. ♘xh8, ♔xh8

Again, this is not forced, but no better defense suggests itself.

32. ♕h6†, ♔g8
33. ♘f7

Here, not even the queen sac will defend: 33. ...♕xf7; 34. ♖d8† and mate next move.

(4.1.2.3.) 29. ...♘f8
30. g6, ♘xg6
31. ♕h6†, ♔g8
32. ♕xg6†, ♔f8
33. ♕h6†, ♔g8
34. ♕h8#

(4.1.2.4) 29. ...♖c7
30. g6...

(Part 3 of 6)

Again, Black's choices are two:

(4.1.2.4.1) 30. ...♘xe5
31. ♕xe5†, ♔xg6
32. ♖d3, h4
33. ♖d4, ♗f3
34. gxf3, ♖g8
35. ♖g4† or

(4.1.2.4.2) 30. ...♖h8
31. ♘xh8, ♔xh8
32. ♕h6†, ♔g8
33. ♘f7

Again, Black will have to give up queen for knight to prevent mate. This returns us to move 27:

(4.1.3) 27. ...♖h8
28. ♘xh8...

And Black has two relevant alternatives:

(4.1.3.1) 28. ...♔xh8
29. ♘xb7

White retains an edge thanks to his more active pieces and threat to invade with his rook on the eighth, but there is still plenty of play left.

(4.1.3.2) 28. ...g5}

<(Here White can head straight into the endgame with 29. ♕f7†, ♕xf7; 30. ♘hxf7

After Black saves his bishop, White can pick up the g-pawn and perhaps retain a slight edge with his extra pawn, although this is by no means an easy endgame to win after a later ...♖xb2, when Black's passed b-pawn may become dangerous.)>

29. hxg5, ♔xh8
30. ♘xb7

And again we see an ending where White is slightly ahead on material, but there is no obvious win in sight.

We now return to move 26:

(4.2) 26. ...♖f8
27. ♖a7...

This looks more aggressive than taking on c8, when Black recaptures with apparent parity. Black now has three logical alternatives (assuming he doesn't want to open lines with ...f6):

(4.2.1) 27. ...♗d5
28. ♘xc8...

<Tempting but unsound is 28. ♖xd7?, ♕xd7; 29. ♕f6†, ♔g8; 30. ♗xg6?, fxg6; 31. ♕xg6†, ♕g7 >

(Part 4 of 6)

28. ...♖xc8
29. ♗a4, ♗c6
30. ♗xc6, ♖xc6
31. ♘xf7, ♕xf7
32. ♕xf7†, ♔xf7
33. ♖xd7†

White has an extra pawn and retains an edge, but final victory is again not in sight.

(4.2.2) 27. ...♘c5
28. ♘xc8, ♖xc8
29. ♕xb4

White has an extra pawn and retains pressure, but this does not guarantee triumph.

(4.2.3) 27. ...♖xc2
28. ♖xb7, ♖xb2
29. ♕d4, ♖b1†
30. ♔h2...

White threatens 31. ♖xd7 followed by 32. ♘f5†, so Black must move his king to a safe square.

30. ...♔g8
31. ♘dxf7!, ♖xf7
32. ♘xf7, ♕xf7
33. ♖xd7, ♕f5
34. ♖d8†, ♔g7/♔h7
35. ♕a7†

If Black is not to be mated, he must interpose the queen on f7 and lose it to 36. ♖d7. We now return to move 26:

(4.3) 26. ...f6
27. ♘gf7...

Threatens mate in two, as in some preceding lines. As before, Black has three options:

(4.3.1) 27. ...fxe5?
28. ♕h6†, ♔f6
29. ♕xg6#

(4.3.2) 27. ...g5!
28. hxg5, fxg5
29. ♘xg5...

(Part 5 of 6)

Again, Black has three logical choices:

(4.3.2.1) 29. ...♖xc2
30. ♘gf7, ♖h8
31. ♘xh8, ♔xh8

<(If 31. ...♗d5?; 32. ♘hf7 renews the threat when it cannot be parried.)>

32. ♘xb7

(4.3.2.2) 29. ...♖f8
30. ♘xc8...

Black can choose one of:

(4.3.2.2.1) 30. ...♖xf4
31. ♘xe7, ♖g4!
32. ♘xe6†!

White gets to keep his extra piece: (a) 32. ...♔f7; 33. ♘d8†, ♔xe7; 34. ♘xb7 (b) 32. ...♔h6?; 33. ♘f5†, ♔g6/♔h7; 34. ♘e3† (c) 32. ...♔h8; 33. ♘g6†, ♔g8; 34. ♘gf4: White remains a piece ahead, and Black's rook is in danger of being trapped after 35. ♗f5.

(Part 6 of 6)

Black's other option is

(4.3.2.2.2) 30. ...♖xc8
31. ♖a7, ♖xc2
32. ♖xb7

If Black tries to regain material parity with 32. ...♖xb2?, he runs into 33. ♖xd7!, ♕xd7; 34. ♕f6†, ♔g8; 35. ♕g6†, ♔f8; 36. ♘h7†, ♔e7; 37. ♕f6†, ♔e8; 38. ♕f8#.

Now we return to move 26:

(4.4) 26. ...♘xe5
27. ♕xe5†, f6
28. ♕xe6!, ♖c7

<(If 28. ...♕xe6?; 29. ♘xe6† and White will emerge with an extra piece.)>

29. ♘xb7, ♖xb7
30. ♕xe7†, ♖xe7
31. ♘f3

With two minor pieces for a rook, White should eventually win.

Conclusion: Removing an important defender, the bishop on b4, looks like White's strongest continuation. However, the ensuing complications are almost incalculably complex, and many of them seem less than convincing. Congratulations to <ChessGames.com> on picking a Sunday puzzle that really lives up to its billing.

<<>A request to <ChessGames.com>: If you're going to confront us with puzzles this complex, please allow us longer word/character limits per post.

He made a beeeeeg jump not long afterwards, because I witnessed these games the following year (P Nikolic vs Kasparov, 1994) and (Kasparov vs P Nikolic, 1994).