Testposition

[Spliffjiffer]

[Event "Ch World (match) (PCA)"]
[Site "New York (USA)"]
[Date "1995.09.25"]
[Eventdate "?"]
[Round "9"]
[Result "1-0"]
[White "Viswanathan Anand"]
[Black "Garry Kasparov"]
[ECO "B84"]
[WhiteElo "?"]
[BlackElo "?"]
[Plycount "69"]

1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 a6 6.Be2 e6 7.O-O Be7 8.a4 Nc6 9.Be3 O-O 10.f4 Qc7 11.Kh1 Re8 12.Bf3 Bd7 13.Nb3 Na5 14.Nxa5 Qxa5 15.Qd3 Rad8 16.Rfd1 Bc6 17.b4 Qc7 18.b5 Bd7 19.Rab1 axb5 20.Nxb5 Bxb5 21.Qxb5 Ra8 22.c4 e5 23.Bb6 Qc8 24.fxe5 dxe5 25.a5 Bf8 26.h3 Qe6 27.Rd5! Nxd5 28.exd5 Qg6 29.c5 e4 30.Be2 Re5 31.Qd7 Rg5 32.Rg1 e3 33.d6 Rg3 34.Qxb7 Qe6 35.Kh2
1-0

[d]r3rbk1/1p3ppp/1B2qn2/PQ2p3/2P1P3/5B1P/6P1/1R1R3K w - - 1 27

Here Anand played Rd5! and won quickly.
Does any engine (maybe on strong hardware) can confirm that 27.Rd5 is best here?
On my old computer SF seems a bit clueless and wanna play 27.Be3 while Komodo likes 27. c5 at some point...

[AdminX]

Even on my phone (S6) Komodo sticks with c5, displaying tbhits and giving white 0.56 eval @depth 26

[Lyudmil Tsvetkov]

Without using engine to check lines tactically, I do not like Rd5 because of g5.

Be3 prevents g5.

In such positions, where there are no knight outposts, open in nature and with potential attacks, SF seems to fare better than Komodo most of the time.

Interesting position, but of course, it is not easy for humans at all.

[Terry McCracken]

[Event "Ch World (match) (PCA)"]
[Site "New York (USA)"]
[Date "1995.09.25"]
[Eventdate "?"]
[Round "9"]
[Result "1-0"]
[White "Viswanathan Anand"]
[Black "Garry Kasparov"]
[ECO "B84"]
[WhiteElo "?"]
[BlackElo "?"]
[Plycount "69"]

1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 a6 6.Be2 e6 7.O-O Be7 8.a4 Nc6 9.Be3 O-O 10.f4 Qc7 11.Kh1 Re8 12.Bf3 Bd7 13.Nb3 Na5 14.Nxa5 Qxa5 15.Qd3 Rad8 16.Rfd1 Bc6 17.b4 Qc7 18.b5 Bd7 19.Rab1 axb5 20.Nxb5 Bxb5 21.Qxb5 Ra8 22.c4 e5 23.Bb6 Qc8 24.fxe5 dxe5 25.a5 Bf8 26.h3 Qe6 27.Rd5! Nxd5 28.exd5 Qg6 29.c5 e4 30.Be2 Re5 31.Qd7 Rg5 32.Rg1 e3 33.d6 Rg3 34.Qxb7 Qe6 35.Kh2
1-0

[d]r3rbk1/1p3ppp/1B2qn2/PQ2p3/2P1P3/5B1P/6P1/1R1R3K w - - 1 27

Here Anand played Rd5! and won quickly.
Does any engine (maybe on strong hardware) can confirm that 27.Rd5 is best here?
On my old computer SF seems a bit clueless and wanna play 27.Be3 while Komodo likes 27. c5 at some point...

Accepting the Sac was Taboo! OTOH not accepting Black still had a lost game positionally.

Interesting game, thanks!

[Eelco de Groot]

Cool! Peer does not post often, but if he posts something... Although the position looked familiar, it is old enough I suppose that games were not immediately checked with computer help. So I could not find it posted although I have only done a limitd check on 1995 for instance.

I think it is still too hard for most computers. Even giving Stockfish the first move it first thinks it is just equal, then fails low for Black but a difficult move then is 31. Qd7 {!} From a distance (agreeing at this point with what was played by both players in the game) it thinks Anand should now have played 31. Bc7 and that is just not enough. After a while it finds 31. Qc4 and that is maybe even better than what Anand played.

[d]r4bk1/1p3ppp/1B4q1/PQPPr3/4p3/7P/4B1P1/1R5K w - -

Engine: SfVerification MOD MP (512 MB)
by Tord Romstad, Marco Costalba and Joona Kiiski

31 7:55 +1.12 31.Qd7 Re7 32.Qa4 e3 33.Rf1 Ree8
34.Qb3 Qe4 35.d6 Re6 36.Qc4 Qe5
37.Rd1 h6 38.d7 Be7 39.d8Q+ Bxd8
40.Bxd8 Rc8 41.Bb6 Rf6 42.Bf3 Rxf3
43.gxf3 e2 44.Re1 (1.395.069.181) 2932

31 7:55 +1.08 31.Qc4 Rg5 32.Bf1 Qf5 33.d6 g6
34.Re1 Rc8 35.Qxe4 Qxe4 36.Rxe4 Rd5
37.Bb5 Kg7 38.Ba4 Rcxc5 39.Bxc5 Rxc5
40.d7 Rc1+ 41.Kh2 Bd6+ 42.g3 Bc7
43.Re8 Ra1 44.Bb3 (1.395.069.181) 2932

31 7:55 +0.24 31.Bc7 Rg5 32.Rg1 Rxd5 33.Qxb7 Rxc5
34.Qxa8 Rxc7 35.a6 Qg5 36.Rd1 Qe7
37.Qb8 g6 38.Rd8 Rc1+ 39.Bd1 e3
40.Kh2 Qc7+ 41.Qxc7 Rxc7 42.Kg3 Kg7
43.Kf3 Bc5 44.g4 (1.395.069.181) 2932

31 7:55 0.00 31.Bc4 Rg5 32.Rg1 e3 33.Bd3 Qh5
34.Be2 Qh4 35.Qd7 Qf2 36.Qxb7 Qxe2
37.Qxa8 Rxg2 38.Rxg2 Qf1+ 39.Kh2 e2
40.Rxe2 Qxe2+ 41.Kg1 Qe1+ 42.Kg2 Qe2+
43.Kg3 Qe3+ 44.Kh2 (1.395.069.181) 2932

__________________________________________________

32 11:22 +1.17 31.Qd7 Re7 32.Qa4 e3 33.Rf1 Ree8
34.Qb3 Qe4 35.d6 Re6 36.Qc4 Qe5
37.Rd1 h6 38.d7 Be7 39.d8Q+ Bxd8
40.Bxd8 Rc8 41.Rd5 Qe4 42.Qxe4 Rxe4
43.Bb6 Kf8 44.Rd7 (2.018.775.459) 2956

32 11:22 +1.14 31.Qc4 Rg5 32.Bf1 e3 33.Bd3 Qf6
34.Qe4 g6 35.d6 Qc3 36.Rf1 Re5
37.Qc4 Qxc4 38.Bxc4 e2 39.Re1 Re4
40.Bd5 Rd4 41.Bf3 Re8 42.Rxe2 Rxe2
43.Bxe2 Bxd6 44.cxd6 (2.018.775.459) 2956

32 11:22 +0.20 31.Bc7 Rg5 32.Rg1 Rxd5 33.Qxb7 Rxc5
34.Qxa8 Rxc7 35.a6 Qg5 36.Rd1 Qe7
37.Qb8 g6 38.Rd8 Rc1+ 39.Bd1 e3
40.Kh2 Qc7+ 41.Qxc7 Rxc7 42.Kg3 Kg7
43.Kf3 Bc5 44.g4 (2.018.775.459) 2956

32 11:22 0.00 31.Bc4 Rg5 32.Rg1 e3 33.Bd3 Qh5
34.Be2 Qh4 35.Qd7 Qf2 36.Qxb7 Qxe2
37.Qxa8 Rxg2 38.Rxg2 Qf1+ 39.Kh2 e2
40.Rxe2 Qxe2+ 41.Kg1 Qe1+ 42.Kg2 Qe2+
43.Kg3 Qe3+ 44.Kh2 (2.018.775.459) 2956

__________________________________________________

34 18:06 +1.24 31.Qc4 Rg5 32.Bf1 e3 33.Bd3 Qf6
34.Qe4 g6 35.d6 Qc3 36.Rf1 Re5
37.Qc4 Qxc4 38.Bxc4 e2 39.Re1 Re4
40.Bd5 Rd4 41.Bf3 Re8 42.Rxe2 Rxe2
43.Bxe2 Bxd6 44.cxd6 (3.187.106.293) 2933

34 18:06 +1.09 31.Qd7 Re7 32.Qa4 e3 33.Rf1 Re4
34.Qd7 Re7 35.Qg4 Qxg4 36.Bxg4 f5
37.Bf3 Rd7 38.d6 Kf7 39.Re1 Re8
40.a6 Bxd6 41.Bd5+ Kf6 42.c6 bxc6
43.Bxc6 Rde7 44.Bxe8 (3.187.106.293) 2933

34 18:06 +0.22 31.Bc7 Rg5 32.Rg1 Rxd5 33.Qxb7 Rxc5
34.Qxa8 Rxc7 35.a6 Qg5 36.Qb8 Qc5
37.Ra1 Ra7 38.Rd1 Qc7 39.Qxc7 Rxc7
40.Rd4 g6 41.Rxe4 Rc1+ 42.Kh2 Bd6+
43.g3 Kg7 44.Bb5 (3.187.106.293) 2933

34 18:06 0.00 31.Bc4 Rg5 32.Rg1 e3 33.Bd3 Qh5
34.Be2 Qh4 35.Qd7 Qf2 36.Qxb7 Qxe2
37.Qxa8 Rxg2 38.Rxg2 Qf1+ 39.Kh2 e2
40.Rxe2 Qxe2+ 41.Kg3 Qe3+ 42.Kg2 Qe2+
43.Kg3 (3.187.106.293) 2933

__________________________________________________

35 27:57 +1.38 31.Qc4 e3 32.Rd1 Rf5 33.d6 Rf2
34.Rg1 Qf5 35.Bd3 Qe5 36.Re1 g6
37.Qe4 Qc3 38.Qxe3 Rd2 39.Be4 Qxe3
40.Rxe3 Re8 41.Rd3 Rxd3 42.Bxd3 Re3
43.Bc4 Re1+ 44.Kh2 (4.934.440.515) 2941

35 27:57 +1.18 31.Qd7 Re7 32.Qa4 e3 33.Rf1 Re4
34.Qd7 Re7 35.Qg4 Qxg4 36.Bxg4 f5
37.Bf3 Rd7 38.Re1 Re8 39.a6 Bxc5
40.Bxc5 bxa6 41.d6 Re5 42.Bc6 Rd8
43.Bd4 Rxd6 44.Bxe5 (4.934.440.515) 2941

35 27:57 +0.22 31.Bc7 Rg5 32.Rg1 Rxd5 33.Qxb7 Rxc5
34.Qxa8 Rxc7 35.a6 Qg5 36.Qb8 Qc5
37.Ra1 Ra7 38.Rd1 Qc7 39.Qxc7 Rxc7
40.Rd4 g6 41.Rxe4 Rc1+ 42.Kh2 Bd6+
43.g3 Bc5 44.g4 (4.934.440.515) 2941

35 27:57 0.00 31.Bc4 Rg5 32.Rg1 e3 33.Bd3 Qh5
34.Be2 Qh4 35.Qd7 Qf2 36.Qxb7 Qxe2
37.Qxa8 Rxg2 38.Rxg2 Qf1+ 39.Kh2 e2
40.Rxe2 Qxe2+ 41.Kg3 Qe3+ 42.Kg2 Qe2+
43.Kg3 (4.934.440.515) 2941

[Edit]But at depth 36 the nearsighted computer finally sees the point of Anand's move!

r4bk1/1p3ppp/1B4q1/PQPPr3/4p3/7P/4B1P1/1R5K w - -

Engine: SfVerification_010 (not quite sure about the versionnumber yet, and it may not be the strongest) MOD MP (512 MB)
by Tord Romstad, Marco Costalba and Joona Kiiski

36 46:31 +1.50 31.Qd7 Re7 32.Qa4 e3 33.Rd1 h6 34.d6 Re6
35.d7 Be7 36.d8Q+ Bxd8 37.Bxd8 Qg3
38.Qb5 Rb8 39.Bc4 e2 40.Bxe2 Qf2
41.Bf1 Qc2 42.Qd7 Qxc5 43.Bb5 Re5
44.Bc7 (8.159.519.431) 2923

36 46:31 +1.37 31.Qc4 Rg5 32.Bf1 e3 33.Bd3 Qf6
34.Qe4 g6 35.d6 Qc3 36.Rf1 Re5
37.Qc4 Qxc4 38.Bxc4 e2 39.Re1 Re4
40.Bd5 Rd4 41.Bf3 Re8 42.Rxe2 Rxe2
43.Bxe2 Bxd6 44.cxd6 (8.159.519.431) 2923

36 46:31 +0.16 31.Bc7 Rg5 32.Rg1 Rxd5 33.Qxb7 Rxc5
34.Qxa8 Rxc7 35.a6 Qg5 36.Qb8 Qa5
37.Rd1 g6 38.Rd8 Rc1+ 39.Kh2 Qc7+
40.Qxc7 Rxc7 41.g4 Kg7 42.Kg3 g5
43.Bb5 Rc3+ 44.Kg2 (8.159.519.431) 2923

36 46:31 0.00 31.Bc4 Rg5 32.Rg1 e3 33.Bd3 Qh5
34.Be2 Qh4 35.Qd7 Qf2 36.Qxb7 Qxe2
37.Qxa8 Rxg2 38.Rxg2 Qf1+ 39.Kh2 e2
40.Rxe2 Qxe2+ 41.Kg3 Qe3+ 42.Kg2 Qe2+
43.Kg3 (8.159.519.431) 2923

[Lyudmil Tsvetkov]

[d]r3rbk1/1p3p1p/1B2qn2/PQ1Rp1p1/2P1P3/5B1P/6P1/1R5K w - g6 0 2

g5 draws, I checked with SF, it ends in a forced line after Bc7, which seems best and only move for SF, as black threatens g4; it ends quickly in an opposite coloured bishops endgame, which SF assesses with 85cps edge.

Dead drawn.
Therefore it does not pick this line.

All other moves SF prefers are more dubious:

- on Rc8 SF does not see c5, and it takes a while after that for the score to increase
- As Eelco points out, it takes a while for SF to also see Nd5 is losing, although it is not its best choice
- h5 seems to be weaker than g5, longer to achieve a draw

See what intuition is: g5 was my first and only choice, I picked it up at depth 1 and never wavered from it, actually I did not even go to depth 2.

Poor SF, it deepens iteration after iteration, and still can not see g5 is an easy draw, while Nd5 and Rc8 lose.

[Lyudmil Tsvetkov]

In the g5 line, SF is very much afraid of a line ending here:

[d]8/2r2pk1/P7/1B2pK2/3bP3/6P1/R7/8 w - - 0 31

150cps white edge, why on earth should SF choose this line?

As the white king and rook get more active, the advantage increases, but no much benefit from this.

Maybe the fact that on g5 white has only acceptable response, Bc7, should somehow ring a bell.

[Lyudmil Tsvetkov]

OK, my last word here, not to drive people crazy: as I looked at the position more closely, I think the d5 move should be followed by a question mark, instead of an exclamation mark, as g5 draws.

My first impression about the moves Komodo and SF picked was wrong, and c5 is indeed the strongest move, after which black should lose by force.

So, according to me, the position on the main diagram should be won for white after c5, but Rd5 only draws.

As I always add something concerning SF and the framework, why SF does not see c5?

Well, it does not have good space in terms of pawns.
The most urgent thing to do is to give all pawns on the 5th and 6th ranks that are not connected, as connected pawns already have their bonus, not passers, and not storming pawns, like a5 and c5 above, some bonus.

But that will certainly be after the big simplification urge is over.

[gordonr]

In the g5 line, SF is very much afraid of a line ending here

Can you give this whole line? Thanks in advance.

[MikeGL]

It seems the puzzle is an "avoid move", for 27... Nxd5

So the correct fen line could be:
r3rbk1/1p3ppp/1B2qn2/PQ1Rp3/2P1P3/5B1P/6P1/1R5K b 1 1 am Nxd5;

[d]r3rbk1/1p3ppp/1B2qn2/PQ1Rp3/2P1P3/5B1P/6P1/1R5K b 1 1

Stockfish 6 actually avoids 27... Nxd5 at Depth 27, and prefers 27...g6 or 27... h5 at -22 centipawns.