Here's my computer checked breakout:

<25. a5!! Qxa5>

If 25... Nxa5, then White wins after 26. h5! Nb7 (26... Bxd6 27. exd6 Nb7 28. Nxb7 Bxb7 29. hxg6 f5 30. Be5 Qa5 31. Qh3 Qd2 32. Qxb3 Ra7 33. Rc2 Qg5 34. f4 Qxg6 35. Bxd4 Nxf4 36. Rxf4 ) 27. Ncxb7 Bxb7 28. hxg6 f5 29. exf6 Nxf6 30. Nxb7 Qxb7 31. Rc7 ;

If 25... Qc7, then White wins with 26. Nxf7 Rxf7 27. Qxg6+ Kf8 28. Nxe6+ Bxe6 29. Qxe6 ;

If 25... Qd8, then White is winning following 26. Nxf7! Rxf7 27. Nxe6 Bxe6 28. Qxg6+ Kf8 29. Rxc6 Nc7 30. Rxc7 Qxc7 Qxe6 .

<26. Nxf7!> With accurate play this move apparently wins. However, with Kasparov's follow-up it is not so clear in the game continuation.

A winning alternative here is 26. h5! Ndb4 (26... Bxd6 27. exd6 Qb5 28. d7 Qxd3 29. Bxd3 Bxd7 30. Nxd7 Ncb4 31. Be4 Rfc8 32. Be5 d3 33. Rxc8+ Rxc8 34. Rd1 d2 35. Rxd2 ) 27. Qxb3 gxh5 28. Qf3 .

<26... Rxf7 27. Qxg6+ !?> Though interesting, this appears to be inaccurate.

Stronger is 27. Nxe6! Bxe6 28. Qxg6+ Kf8 29. Rxc6 Nc7 30. Rxc7 Qxc7 31. Qxe6 Qb7 32. Qxh6+ Kg8 33. Bg6 Bf8 34. Bxf7+ Qxf7 35. Qc6 Ra7 36. Qb6 Bg7 37. f4 .

<27... Kf8?> Black overlooks a stronger defense.

Black just might be able to hold the position after 27... Rg7! 28. Qe8+ Bf8 29. Qxc6 Rb8 , when play might continue 30. Bf5 (30. Qe8 Qb5 31. Qxb5 Rxb5 32. Ne4 Bb7 33. Rfd1 Rb4 34. Nd6 Nb6 35. h5 Bd5 36. Ne8 Rb7 37. Nf6+ Kg7 38. Nxd5 exd5 39. Rd2 Nc4 40. Re2 a5 =) 30...Nc7 31. Nxe6 Bb7 32. Qxc7 Rxc7 33. Rxc7 Bd5 34. Rd7 Rb7 35. Rxb7 Bxb7 36. Nxd4 Qd8 37. Be6+ Kh8 38. Nf5 Bc8 39. Bxc8 Qxc8 =.

<28. Nxe6+ Bxe6 29. Rxc6 Bd7 30. Qxh6+> 1-0

Black resigns in lieu of 30...Ke8 (30... Rg7 31. Rg6 ; 30... Kg8 31. Rg6+ Rg7 32. Qxg7#) 31. e6 .

Kasparov vs Piket, 1997 (25.?)

White to play and win.

Material: Down a P. The Black Kg8 has 3 legal moves. White has a battery Bb1 and Qd3, which could mate if Pg6 were removed. The Black K-side has weak dark squares, which Nc5 and Nd6 can reach in 2 moves. More to the point, however, the Pf7 must protect Pg6, and 25.Nxf7 removes Pf7. It is significant that Bc8 disconnects Ra8 and Rf8, because invasion points appear on the back rank after a deflection of Rf8. Because Pf7 does not really protect Pe6, the Black Bc8 has the burden of protecting Pe6. White can remove Bc8 with 25.Nxc8, leaving Pe6 unprotected. Black therefore has a much shakier defense than first impressions indicate. In passing, the White Rc1 attacks Nc6, obscured by Nc5, so the Black Qb6 has a burden. The White Bg3 and Rf1 require activation.

Candidates (25.): Nxf7, Nxc8

25.Nxf7 (threatening 26.Qxg6+ then 27.Qh7#)

<[Rest of analysis deleted because the game variation is better.]>

With humans able to improve near the end of the full computer variation, and with the last move entered <emphasized>, Mom-and-Pop Toga II 1.3.1 analysis gives

[ply 15/47 time 01:46 value +1.77]

25.<a5> Nxa5 26.h5 Nb4 27.Qxd4 Nc2 28.Bxc2 bxc2 29.hxg6 fxg6 30.Qe4 Kg7 31.b4 Rb8 32.Qxc2 Nc6 33.Na4 Nxb4 34.Nxb6 Nxc2 35.Rxc2 Rxb6 36.Nxc8

[ply 15/56 time 04:49 value +3.09]

25.a5 <Qxa5> 26.Nxf7 Rxf7 27.Nxe6 Bxe6 28.Qxg6+ Kf8 29.Rxc6 Nc7 30.Rxc7 Qxc7 31.Qxe6 Qb7 32.Qxh6+ Kg8 33.Bg6 Rg7 34.Bf5 Rf8 35.Be6+ Rff7 36.Ra1 a5 37.Rxa5 Bb4 38.Bxf7+ Rxf7 39.Qg6+ Kh8

[ply 15/45 time 00:51 value +1.18]

25.<Nxf7> Rxf7 26.Qxg6+ Kf8 27.Qxh6+ Kg8 28.Nxe6 Bxe6 29.Qxe6 Ncb4 30.Rc8+ Rxc8 31.Qxc8+ Bf8 32.Qc4 Rg7 33.Rd1 Qc5 34.Qxb3 d3 35.Bxd3 Rxg3 36.Bh7+ Kxh7 37.Qxg3 Bh6

The game defense is much weaker than 25...Nxa5, so I do not share everyone's apparent embarrassment about 25.Nxf7. Toga indicates 25.Nxf7 is not best, but it probably still wins and is only +0.6 P worse than best play after 25.a5.

In passing, 25.a5 did occur to me (I do not write <everything> down), but I could see no advantage accruing to White in the position after 25...Nxa5. Clearly, 25...Qxa5 leaves Nc6 loose, but to me, the interest in the position is to find an explanation of why the zwischenzug 25.a5 and the resulting deflection of Nc6 under best play 25...Nxa5 leads to +0.6 P advantage.

< <Marmot PFL> wrote : I don't think black was completely lost if he played 27...Rg7 instead of the silly Kf8.>

[ply 15/48 time 01:49 value +1.20]

25.a5 Qxa5 26.Nxf7 Rxf7 27.Qxg6+ <Rg7> 28.Qe8+ Bf8 29.Nxe6 Bb7 30.Qh5 Qb6 31.Nxg7 Bxg7 32.Qf5 Nce7 33.Qh7+ Kf8 34.Rfe1 Re8 35.Rcd1 Bc8 36.Be4 Be6 37.Rd3 Bg4 38.Rc1

Hi, <Marmot PFL>.

(Your handle, like many others, makes me feel depressingly unimaginative.)

[ply 15/47 time 00:50 value -1.00]

26.<Nxf7> Rxf7 27.Na4 Qd8 28.Qxg6+ Rg7 29.Qxh6 Bf8 30.Rfd1 Raa7 31.Bg6 Rgc7 32.Qh5 Rxc1 33.Rxc1 Rg7 34.Rd1 Nc6 35.Rc1 Bb7

Instead, the candidate is 26.h5:

[ply 15/56 time 02:52 value +1.78]

26.<h5> Bxd6 27.exd6 Ra7 28.hxg6 f5 29.Qxd4 Rg7 30.Rfd1 Rxg6 31.d7 Bb7 32.Qe5 Nc6 33.Qd6 Nd8 34.Qxb6 Nxb6 35.Nxb7 Nxb7 36.Rc7 Nd8 37.Bd3

In the following, a move with <<> double emphasis> is not best play with the immediate 25.Nxf7, but might be reasonable, and a move with <<<<>>>quadruple emphasis> is prevented by the deflection 25.a5.

(1) The deflection 25.a5 Qxa5 improves, because after 25.Nxf7 Rxf7 26.Qxg6,

(1.1) 26…Kf8 27.Qxh6+ Rg7 28.Nxe6+ Bxe6 29.Qxe6 Ncb4

In the game variation, after deflection to a5, Qb6 no longer has access to the K-side.

(1.2) 26…<<>Rg7> 27.Qe8+ then 28.<<<<>>>Qxc6>

In the game variation, after deflection to a5, Qb6 no longer protects Nc6.

(2) The deflection 25.a5 Nxa5 improves, because after 25.Nxf7 Rxf7 26.Qxg6

(2.1) 26…Kf8 27.Qxh6+ Rg7 28.Ne4 <<<<>>>Nd8>

(2.2) 26…<<>Rg7> 27.Qe8+ Bf8 28.Ne4 <<<<>>>Nd8>

At first, the board looks full of possibilities for White. However, the candidate moves appear to be Nxf7, h5, Nxe6 or the quiet Qe4, with the former being the most promising. When I ran through the Nxf7 variation, the attack stalled largely because Black's queen was protecting the 6th rank.

h5 seemed decent though until I realized that Black could simply play Nxe5 followed by Bxd6 and again, the attack would stall because of Black's queen.

Nxe6 seemed bad for White after Bxe6 since it created another active defender for Black. The Qe4 had possibilities and might have been winning for White, but it was hardly conclusive. And, it didn't press home White's attack on the g6 pawn, which is the point of White's play.

Only then did I look at a5 as a move to lure Black's queen off the 6th rank. After Nxa5, this led to simplification with White having a passed pawn and active pieces (except the f-file rook). Moreover, his bishops were well placed in the center.

In contrast, Black's pieces were not at all coordinated, much less active. His bishop was rooted on c8, his rooks were on opposite sides of the board, and the knight on a5 was a spectator.

So, I checked Qxa5 and it enabled the Nxf7 variation. Either way was losing for Black, but Nxa5 might have prolonged things more. I expected 27. Qxg6+ Rg7 28. Qe8+ Bf8 29. Qxc6 followed by a resignation from Black, but was satisfied with the game moves for White (29. ... Nc7 instead of Bd7 is still losing for Black).

The move 25.h5 looks strong but after 25... Nxe5 26.Bxe5 Bxd6 27.hxg6 f5 I cannot see an immediate victory for White.

Another possibility is 25.Nxf7:

A) 25... Kxf7 26.Qxg6 mate.

B) 25... Bxc5 26.Qxg6 mate.

C) 25... Rxf7 26.Qxg6+

C.1) 26... Kh8 27.Qxf7 winning.

C.2) 26... Kf8 27.Nxe6+ Bxe6 (27... Ke8 28.Ng7+ and 29.Rxc6) 28.Qxe6 Ndb4 29.Qxh6+ and three pawns and the attack seem to be more than enough compensation for the piece.

C.3) 26... Rg7 27.Qe8+ Bf8 28.Nxe6 Bb7 (28... Bxe6 29.Qxe6+ Rf7 30.Rxc6) 29.Qh5 Re7 (29... Rf7 30.Qg6+ Bg7 31.Nxg7 Rxg7 32.Qe6+ and 33.Qxd5) 30.Qg6+ Kh8 31.Nxf8 Rxf8 32.Qxh6+ Kg8 33.Qg6+ Kh8 34.e6 threatening Rxc6 and Be5+.