51 losses by SF_21061923 in NCM

331 views
Skip to first unread message

Bryan

unread,
Jun 20, 2021, 6:15:35 PMJun 20
to FishCooking
I went through all the pgns from the 20k games at NCM between SF_21061923 (with net nn-75980ca503c6) and SF7, and looked at all 62 SFdev losses. Of those 62 losses, I threw out 11 that were busted openings (greater than +/-1.00 w/o contempt). I put the remaining 51 losses in a pgn, and attached it. Of those 51 losses, 12 are with black. I have numbered the 51 games individually (1 through 51), so that we can refer to them easily.
SFdev_2061923_losses_vs_SF7.pgn

Bryan

unread,
Jun 20, 2021, 7:15:21 PMJun 20
to FishCooking
In Game 1, the mistake is at move 43, black to move, where SFdev played h7-h5. The better move is Rg5-h5. SFdev finds this at d=35. SFdev mistakenly thinks white will reply to 43...h7-h5 with 44.h3-h4, which allows black's rook to reach the fourth rank, after which white has no good options, and the white king can't get past the third rank. But 44.Ra7-a6 pins black's bishop, such that unless black moves a pawn on the a-file or b-file, white's king breaks through to g5 and h6. But if black moves a pawn on the a-file or b-file, white plays Bc2-d4, and double-attacks black's pinned bishop.

FEN: 8/R5pp/2bk4/1p3Pr1/p7/6PP/P1B2K2/8 b - - 0 43
Stockfish_21062023_x64_avx2:
Found 145 tablebases
NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 30/39    00:09     90,867k    9,424k     0.00    h7-h5 h3-h4 Rg5-g4 Bc2-d1 Rg4-c4 Bd1xh5 b5-b4 Ra7xg7 Kd6-e5 Bh5-g6 Rc4-c2+ Kf2-e3 Rc2-c3+ Ke3-d2 Rc3xg3 h4-h5 Rg3-g2+ Kd2-e3 Rg2-g3+ Ke3-d2
 31/42    00:11     103,807k    9,430k     0.00    h7-h5 h3-h4 Rg5-g4 Bc2-d1 Rg4-c4 Bd1xh5 b5-b4 Ra7xg7 Kd6-e5 Bh5-g6 Rc4-c2+ Kf2-e3 Rc2-c3+ Ke3-d2 Rc3xg3 h4-h5 Rg3-g2+ Kd2-e3 Rg2-g3+
 32/42    00:11     109,951k    9,438k     0.00    h7-h5 h3-h4 Rg5-g4 Bc2-d1 Rg4-c4 Ra7xg7 Kd6-e5 Bd1xh5 b5-b4 Bh5-g6 Rc4-c2+ Kf2-e3 Rc2-c3+ Ke3-d2 Rc3xg3 h4-h5 Rg3-g2+ Kd2-e3 Rg2-g3+ Ke3-d2
 33/41    00:14     132,570k    9,447k     0.00    h7-h5 h3-h4 Rg5-g4 Bc2-d1 Rg4-c4 Bd1xh5 b5-b4 Bh5-d1 Rc4-d4 Kf2-e1 Rd4-e4+ Ke1-d2 Re4-d4+ Kd2-e1
 34/31    00:18     171,440k    9,489k     0.00    h7-h5 h3-h4 Rg5-g4 Bc2-d1 Rg4-c4 Bd1xh5 b5-b4 Ra7xg7 Kd6-e5 Bh5-g6 Rc4-c2+ Kf2-e3 Rc2-c3+ Ke3-d2 Rc3xg3 h4-h5 Rg3-g2+ Kd2-e3 Rg2-g3+ Ke3-d2
 35/36-    00:27     261,800k    9,531k    -0.08    h7-h5 Ra7-a6
 35/36-    00:29     278,605k    9,527k    -0.16    h7-h5 Ra7-a6
 35/36-    00:49     470,814k    9,426k    -0.71    h7-h5 Ra7-a6
 35/52+    01:07     630,708k    9,389k    -0.51    Rg5-h5
 35/52    01:07     637,077k    9,389k    -0.30    Rg5-h5 h3-h4 Bc6-d7 Ra7-b7 a4-a3 Bc2-d3 Bd7xf5 Rb7xb5 g7-g6 Bd3-e2 Bf5-d7 Rb5-b6+ Kd6-c7 Rb6-a6 Rh5-f5+ Be2-f3 g6-g5 Kf2-e3 Rf5-c5 Ke3-d4 Rc5-c2 Kd4-e5 g5xh4 Ra6-a7+ Kc7-b6 Ra7-b7+ Kb6-a6 Rb7xd7 Rc2xa2 g3xh4 Ra2-h2 Rd7-d6+ Ka6-b5 h4-h5 a3-a2 Rd6-d1 Kb5-b4
 36/52    01:10     664,116k    9,392k    -0.30    Rg5-h5 h3-h4 Bc6-d7 Ra7-b7 a4-a3 Bc2-d3 Bd7xf5 Rb7xb5 g7-g6 Bd3-e2 Bf5-d7 Rb5-b6+ Kd6-c7 Rb6-a6 Rh5-f5+ Be2-f3 g6-g5 Kf2-e3 Rf5-c5 Ke3-d4 Rc5-c2 Kd4-e5 g5xh4 Ra6-a7+ Kc7-b6 Ra7-b7+ Kb6-a6 Rb7xd7 Rc2xa2 g3xh4 Ra2-h2 Rd7-d6+ Ka6-b5 h4-h5 a3-a2 Rd6-d1 Kb5-b4


Bryan

unread,
Jun 20, 2021, 9:59:55 PMJun 20
to FishCooking
In Game 2, the mistake is at move 77, where SFdev played 77...Rb1-d1. SFdev (at shallow depth) initially thinks that after 77...Rb1-d1, white white will capture with 78.Rd2xd1, but white has the winning move 78.d2-c2. After 78.Rd2-c2, every move for black is no better than a TB mate.

Move 78: black to move: hash cleared: multipv=6: contempt=0: threads=15
FEN: 8/5p2/7k/4p3/1p2P2p/5P2/P1R3QK/q2r4 b - - 3 78
Stockfish_21062023_x64_avx2:
[...]
---------------------------------------------------------------------------
 43/48    1:15:17    107,743,855k    23,852k    -152.65    Rd1-f1 Rc2-c6+ Kh6-h7 Rc6-f6 Rf1-h1+ Qg2xh1 Qa1xa2+ Qh1-g2 Qa2xg2+ Kh2xg2 Kh7-g8 Rf6-b6 Kg8-g7 Kg2-h3 b4-b3 Rb6xb3 Kg7-f6 Rb3-b7 Kf6-g6 Rb7-e7 Kg6-f6 Re7-a7 Kf6-e6 Kh3xh4 Ke6-d6 Ra7xf7 Kd6-c6 Kh4-h3 Kc6-d6 Rf7-f6+ Kd6-c5 Kh3-g4 Kc5-b5 Rf6-f5 Kb5-b4 Kg4-h4 Kb4-c3 Rf5xe5 Kc3-d3 f3-f4 Kd3-d4 Kh4-h3 Kd4-e3 f4-f5 Ke3-e2 f5-f6 Ke2-d2 f6-f7
 43/47    1:15:17    107,743,855k    23,852k    -152.55    Rd1-b1 Rc2-c8 Rb1-h1+ Qg2xh1 Qa1xa2+ Qh1-g2 Qa2xg2+ Kh2xg2 Kh6-g6 Rc8-b8 Kg6-g5 Kg2-h3 Kg5-f4 Kh3xh4 Kf4xf3 Rb8xb4 Kf3-e3 Rb4-b7 f7-f6 Rb7-b6 Ke3-f4 Rb6xf6+ Kf4-e3 Kh4-g5 Ke3xe4
 43/44    1:15:17    107,743,855k    23,852k    -152.55    b4-b3 Rc2-c8 Rd1-h1+ Qg2xh1 Qa1xa2+ Qh1-g2 Qa2xg2+ Kh2xg2 b3-b2 Rc8-b8 Kh6-g5 Rb8xb2 Kg5-f4 Rb2-b3 f7-f6 Kg2-h3 Kf4-g5 Rb3-b6 f6-f5 e4xf5 Kg5xf5 Kh3xh4 Kf5-f4 Rb6-b3 e5-e4
 43/45    1:15:17    107,743,855k    23,852k    -152.55    Qa1-b1 Rc2-c8 Rd1-h1+ Qg2xh1 Qb1xa2+ Qh1-g2 Qa2xg2+ Kh2xg2 Kh6-g6 Rc8-b8 Kg6-g5 Kg2-h3 Kg5-f4 Kh3xh4 Kf4xf3 Rb8xb4 Kf3-e3 Rb4-b7 f7-f6 Rb7-b6 Ke3-f4 Rb6xf6+ Kf4-e3 Kh4-g5 Ke3xe4
 43/45    1:15:17    107,743,855k    23,852k    -152.54    Kh6-h7 Rc2-c7 Kh7-h6 Rc7-c8 Rd1-h1+ Qg2xh1 Qa1xa2+ Qh1-g2 Qa2xg2+ Kh2xg2 Kh6-g6 Rc8-b8 Kg6-g5 Kg2-h3 Kg5-f4 Kh3xh4 Kf4xf3 Rb8xb4 Kf3-e3 Rb4-b7 f7-f6 Rb7-b6 Ke3-f4 Rb6xf6+ Kf4-e3 Kh4-g5 Ke3xe4
 43/45    1:15:17    107,743,855k    23,852k    -152.53    Qa1-c3 Rc2xc3 b4xc3 Kh2-h3 Rd1-d3 Kh3xh4 f7-f6 Qg2-f1 Rd3-d2 Qf1-c1 Kh6-g6 Qc1xc3 Rd2-h2+ Kh4-g3 Rh2xa2 f3-f4 e5xf4+ Kg3xf4 Ra2-f2+ Kf4-g3 Rf2-f1 e4-e5 Rf1-g1+ Kg3-h2 Rg1-g5 e5xf6 Rg5-h5+ Kh2-g3 Rh5-f5 f6-f7 Rf5-g5+ Kg3-f4 Rg5-f5+ Kf4-e3 Rf5xf7
 43/42    1:15:17    107,743,855k    23,852k    -152.52    Qa1-d4 Rc2-c8 Qd4-d8 Rc8xd8 Rd1xd8 Kh2-h3 b4-b3 a2xb3 Rd8-d6 Kh3xh4 Rd6-g6 Qg2xg6+ Kh6xg6 Kh4-g4 Kg6-f6 b3-b4 Kf6-e6 f3-f4 Ke6-e7 f4xe5 Ke7-d8 Kg4-f5 Kd8-e7 b4-b5 Ke7-d8 e5-e6 Kd8-e7 e6xf7 Ke7xf7
 43/57    1:15:17    107,743,855k    23,852k    -152.38    Rd1-d6 Qg2-g8 Rd6-g6 Qg8-h8+ Kh6-g5 Rc2-g2+ Kg5-f4 Qh8xh4+ Kf4xf3 Qh4-h3+ Kf3xe4 Rg2-e2+ Ke4-d5 Qh3-d7+ Rg6-d6 Re2-d2+ Qa1-d4 Rd2xd4+ e5xd4 Qd7-b5+ Kd5-e6 Qb5xb4 Ke6-e5 a2-a4 d4-d3 Qb4-c5+ Rd6-d5 Qc5-e3+ Ke5-f6 Qe3-d2 Rd5-d6 a4-a5 Kf6-g7 Kh2-g2 Kg7-h7 Qd2-f4 Rd6-g6+ Kg2-f3 Kh7-g8 Qf4-d4 d3-d2 Qd4xd2 Rg6-f6+ Kf3-g4 Rf6-g6+ Kg4-f5 Rg6-e6 Qd2-d7 Re6-e1 Qd7-c8+ Kg8-h7 Qc8-c4 Kh7-g7 Qc4-c3+ Kg7-h7
---------------------------------------------------------------------------

If black had played 77...Kh6-h7, then black could have saved the draw:

Move 78: white to move after 77...Kh6-h7.
FEN: 8/5p1k/8/4p3/1p2P2p/5P2/P2R2QK/qr6 w - - 2 78
Stockfish_21062023_x64_avx2:
[...]
---------------------------------------------------------------------------
 69/13    40:04     27,132,563k    11,284k     0.00    Rd2-c2 Rb1-c1 Rc2xc1 Qa1xc1 Qg2-g4 Qc1-d2+ Kh2-h3 Qd2xa2 Qg4-h5+ Kh7-g8 Qh5-g5+ Kg8-h7
 69/7    40:04     27,132,563k    11,284k     0.00    Rd2-d7 Rb1-b2 Rd7xf7+ Kh7-h6 Rf7-f6+ Kh6-h7
 69/26    40:04     27,132,563k    11,284k     0.00    a2-a3 b4-b3 Rd2-d7 Rb1-b2 Rd7xf7+ Kh7-h6 Rf7-f6+ Kh6-h5 Rf6-f5+ Kh5-h6
 69/27    40:04     27,132,563k    11,284k     0.00    f3-f4 Qa1-c3 Qg2-f2 Kh7-g7 Rd2-c2 Qc3-e1 Qf2-g2+ Kg7-h7 Rc2-c7 Qe1-g3+ Qg2xg3 h4xg3+ Kh2xg3 e5xf4+ Kg3-g4 Kh7-g7 e4-e5 Rb1-b2 e5-e6 Rb2xa2 Rc7xf7+ Kg7-g6 Kg4xf4 Ra2-f2+ Kf4-e3 Rf2xf7
-------------------------------------------------------------------------

The latest SFdev finds 77...Kh6-h7 at d=19:

Move 77: black to move: hash cleared: multipv=1: contempt=0: threads=15
FEN: 8/5p2/7k/4p3/1p2P2p/5P2/P2R2QK/qr6 b - - 1 77
Stockfish_21062023_x64_avx2:

NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 12/23    00:00     459k    12,085k    +0.09    Rb1-d1 Rd2xd1 Qa1xd1 Kh2-h3 Qd1-e1 Qg2-h2 Qe1-g3+ Qh2xg3 h4xg3 Kh3xg3 Kh6-g5 Kg3-g2 Kg5-h4 Kg2-h2 f7-f5 e4xf5 Kh4-g5 Kh2-g3 Kg5xf5 Kg3-f2 Kf5-f4
 13/21    00:00     607k    12,139k    +0.05    Rb1-d1 Rd2xd1 Qa1xd1 Kh2-h3 Qd1-e1 Qg2-h2 Qe1-g3+ Qh2xg3 h4xg3 Kh3xg3 Kh6-g5 Kg3-g2 Kg5-h4 Kg2-h2 Kh4-h5 Kh2-g1 Kh5-h6 Kg1-f2 Kh6-g6 Kf2-g3
 14/27    00:00     855k    12,386k     0.00    Rb1-d1 Rd2xd1 Qa1xd1 Kh2-h3 Qd1-e1 Qg2-h2 f7-f6 Kh3-g4 Kh6-g6 Qh2xh4 Qe1-g1+ Kg4-h3 Qg1-h1+ Kh3-g3 Qh1xh4+ Kg3xh4 f6-f5 e4xf5+ Kg6xf5 Kh4-g3 Kf5-g5
 15/20    00:00     897k    12,293k     0.00    Rb1-d1 Rd2xd1 Qa1xd1 Kh2-h3 Qd1-e1 Qg2-h2 f7-f6 Kh3-g4 Kh6-g6 Qh2xh4 Qe1-d2 Kg4-g3 Qd2-g5+ Qh4xg5+ Kg6xg5 Kg3-g2 Kg5-h4 Kg2-h2 Kh4-h5 Kh2-g3
 16/22    00:00     1,019k    12,281k     0.00    Rb1-d1 Rd2xd1 Qa1xd1 Kh2-h3 Qd1-e1 Qg2-h2 f7-f6 Kh3-g4 Kh6-g6 Qh2xh4 Qe1-g1+ Kg4-h3 Qg1-c5 Kh3-g3 Kg6-g7 Kg3-h3 Qc5-c1 Qh4-h5 Qc1-h1+ Kh3-g4 Qh1xh5+ Kg4xh5
 17/24    00:00     1,112k    12,357k     0.00    Rb1-d1 Rd2xd1 Qa1xd1 Kh2-h3 Qd1-e1 Qg2-h2 f7-f6 Kh3-g4 Kh6-g6 Qh2xh4 Qe1-g1+ Kg4-h3 Qg1-c5 Kh3-g3 Kg6-g7 Kg3-h3
 18/23    00:00     1,141k    12,405k     0.00    Rb1-d1 Rd2xd1 Qa1xd1 Kh2-h3 Qd1-e1 Qg2-h2 f7-f6 Kh3-g4 Kh6-g6 Qh2xh4 Qe1-d2 Kg4-g3 Qd2-d4 Qh4-g4+ Kg6-f7 Qg4-c8 Qd4-g1+ Kg3-h3 Qg1-h1+ Kh3-g3
 19/29    00:00     1,707k    11,615k     0.00    Kh6-h7 Rd2-d7 Rb1-b2 Rd7xf7+ Kh7-h8 Rf7-f8+ Kh8-h7
 20/8    00:00     1,715k    11,589k     0.00    Kh6-h7 Rd2-d7 Rb1-b2 Rd7xf7+ Kh7-h6 Rf7-f6+ Kh6-h7
 21/8    00:00     1,739k    11,592k     0.00    Kh6-h7 Rd2-d7 Rb1-b2 Rd7xf7+ Kh7-h6 Rf7-f6+ Kh6-h7
 22/8    00:00     2,002k    11,374k     0.00    Kh6-h7 Rd2-d7 Rb1-b2 Rd7xf7+ Kh7-h6 Rf7-f6+ Kh6-h7
 23/8    00:00     2,155k    11,344k     0.00    Kh6-h7 Rd2-d7 Rb1-b2 Rd7xf7+ Kh7-h6 Rf7-f6+ Kh6-h7
 24/8    00:00     2,657k    11,354k     0.00    Kh6-h7 Rd2-d7 Rb1-b2 Rd7xf7+ Kh7-h8 Rf7-f8+ Kh8-h7
 25/10    00:00     2,706k    11,368k     0.00    Kh6-h7 Rd2-d7 Rb1-b2 Rd7xf7+ Kh7-h6 Rf7-f6+ Kh6-h5 Rf6-f5+ Kh5-h6
 26/10    00:00     2,798k    11,328k     0.00    Kh6-h7 Rd2-d7 Rb1-b2 Rd7xf7+ Kh7-h6 Rf7-f6+ Kh6-h5 Rf6-f5+ Kh5-h6
 27/10    00:00     4,018k    11,129k     0.00    Kh6-h7 Rd2-d7 Rb1-b2 Rd7xf7+ Kh7-h6 Rf7-f6+ Kh6-h5 Rf6-f5+ Kh5-h6
 28/8    00:00     6,064k    10,946k     0.00    Kh6-h7 Rd2-d7 Rb1-b2 Rd7xf7+ Kh7-h6 Rf7-f6+ Kh6-h7
 29/8    00:00     6,072k    10,960k     0.00    Kh6-h7 Rd2-d7 Rb1-b2 Rd7xf7+ Kh7-h6 Rf7-f6+ Kh6-h7
 30/8    00:00     6,091k    10,956k     0.00    Kh6-h7 Rd2-d7 Rb1-b2 Rd7xf7+ Kh7-h6 Rf7-f6+ Kh6-h7
 31/8    00:00     7,352k    10,859k     0.00    Kh6-h7 Rd2-d7 Rb1-b2 Rd7xf7+ Kh7-h6 Rf7-f6+ Kh6-h7
 32/8    00:00     7,387k    10,863k     0.00    Kh6-h7 Rd2-d7 Rb1-b2 Rd7xf7+ Kh7-h6 Rf7-f6+ Kh6-h7
--------------------------------------------------------------------------

Bryan

unread,
Jun 20, 2021, 11:26:41 PMJun 20
to FishCooking
In Game 3 SFdev played 27...Re8-c8, expecting white to reply with 28.Qe2-d3 to double-defend the c3 pawn. But white has a crushing reply: 28.Rd6-d7, giving up the c3 pawn to play Qe2-h5 and put a double-attack on black's pawn on f7 pawn, forcing black to create weaknesses on the dark squares on the kingside with g7-g6. What SFdev should have played at move 27 was h7-h6. The move 27...Re8-c8 allows 28.Rd6-d7, because Re8-c8 prevents the black queen from moving to the c8 square, and hitting simultaneously both d7 and c3. So Re8-c8 gives the white rook free tempi on the seventh rank while all black's pieces are on the queenside.

At move 28, white to move, white is winning:
FEN: bqr3k1/5ppp/3Rp3/Np2P3/nB1P1P2/2P3P1/4Q2P/6K1 w - - 5 28

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
 [...]
 51/82    27:13     19,493,503k    11,935k    +5.07    Rd6-d7 Na4xc3 Qe2-h5 g7-g6 Qh5-h4 Nc3-d5 Bb4-c5 h7-h5 f4-f5 e6xf5 Qh4-g5 Nd5-b6 Bc5-d6 Nb6xd7 Bd6xb8 Rc8xb8 e5-e6 Nd7-f8 e6-e7 Nf8-e6 Qg5-d2 Rb8-e8 Qd2-b4 Ne6-g7 Qb4-d6 Ba8-e4 Qd6-d8 b5-b4 Na5-b3 f5-f4 g3xf4 Be4-c6 Nb3-a5 Ng7-e6 Qd8-d6 Bc6-e4 Na5-b3 Ne6-g7 Nb3-d2 Ng7-f5 Qd6-d8 Be4-c6 Nd2-e4 Nf5xe7 Ne4-f6+ Kg8-f8 Nf6xe8 Bc6xe8 d4-d5 b4-b3 d5-d6 Ne7-d5 d6-d7 Nd5-f6 d7xe8Q+ Nf6xe8 Qd8-b8 Kf8-e7 Qb8-b7+ Ke7-e6 Qb7xb3+ Ke6-e7 Kg1-f2 Ne8-d6 Kf2-f3 Nd6-f5 Qb3-d5 Ke7-f8 Qd5-d7 Nf5-h6 Kf3-e4 Kf8-g7 Qd7-d4+ Kg7-g8 Qd4-f6 Nh6-f5 Ke4-e5 Nf5-h6 h2-h3 Kg8-h7 Qf6-e7 Kh7-g7

But if at move 27, black had played h7-h6, black could have saved the draw.

Move 28: white to move after 27...h7-h6: hash cleared: multipv=0: contempt=0: thread=15
FEN: bq2r1k1/5pp1/3Rp2p/Np2P3/nB1P1P2/2P3P1/4Q2P/6K1 w - - 0 28

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 63/20    23:32     12,623,865k    8,940k     0.00    Na5-c6 Ba8xc6 Rd6xc6 Qb8-b7 d4-d5 Re8-d8 Rc6-d6 Rd8xd6 e5xd6 Qb7xd5 Qe2-f2 Qd5-d1+ Kg1-g2 Qd1-d3 h2-h4 Qd3-e4+ Kg2-g1 Qe4-d3

SFdev finds 27...h7-h6 at d=26:

Move 27: black to move: hash cleared: multipv=0: contempt=0: threads=15
FEN: bq2r1k1/5ppp/3Rp3/Np2P3/nB1P1P2/2P3P1/4Q2P/6K1 b - - 4 27

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
 [...]
 23/25    00:00     8,394k    9,133k     0.00    Re8-c8 Qe2-d3 Ba8-d5 f4-f5 Qb8-a8 f5-f6 Bd5-h1 d4-d5 Bh1xd5 Qd3-e3 h7-h6 f6xg7 Kg8xg7 c3-c4 b5xc4 Bb4-d2 Kg7-g8 Qe3xh6 Qa8-a7+ Bd2-e3 Qa7xa5 Qh6-g5+ Kg8-f8 Qg5-h6+
 24/35    00:01     12,354k    9,117k     0.00    Re8-c8 Qe2-d3 Ba8-d5 f4-f5 Qb8-a8 f5-f6 Bd5-h1 d4-d5 Bh1xd5 Qd3-e3 h7-h6 f6xg7 Kg8xg7 c3-c4 b5xc4 Bb4-d2 Kg7-g8 Qe3xh6 Qa8-a7+ Bd2-e3 Qa7xa5 Qh6-g5+ Kg8-f8 Qg5-h6+
 25/25    00:01     13,386k    9,112k     0.00    Re8-c8 Qe2-d3 Ba8-d5 f4-f5 Qb8-a8 f5-f6 Bd5-h1 d4-d5 Bh1xd5 Qd3-e3 h7-h6 f6xg7 Kg8xg7 c3-c4 b5xc4 Bb4-d2 Kg7-g8 Qe3xh6 Qa8-a7+ Bd2-e3 Qa7xa5 Qh6-g5+ Kg8-f8 Qg5-h6+
 26/41+    00:03     35,202k    9,017k    -0.18    h7-h6
 26/41    00:04     36,969k    9,004k    -0.22    h7-h6 Na5-c6 Ba8xc6 Rd6xc6 Qb8-b7 d4-d5 Re8-d8 Bb4-d6 Rd8-d7 d5xe6 Qb7xc6 e6xd7 Qc6xd7 c3-c4 b5xc4 Qe2xc4 Na4-b6 Qc4-e4 g7-g6 Qe4-d4 Qd7-b5 Bd6-c5 Qb5-b1+ Kg1-g2 Qb1-a2+ Kg2-h3 Qa2-e6+ g3-g4 Nb6-d5 Kh3-g3 g6-g5 f4xg5 h6xg5
 27/35+    00:05     52,132k    8,998k    -0.14    h7-h6
 27/36    00:06     57,918k    9,015k    -0.07    h7-h6 f4-f5 Qb8-c8 f5xe6 f7xe6 Qe2-c2 Ba8-d5 c3-c4 Bd5xc4 Na5xc4 Qc8xc4 Qc2xc4 b5xc4 Rd6-c6 Re8-b8 Rc6xc4 Na4-b6 Bb4-d6 Rb8-a8 Rc4-b4 Nb6-d5 Rb4-b8+ Ra8xb8 Bd6xb8 h6-h5 Kg1-g2 g7-g5 Kg2-f3 Kg8-h7 Kf3-e4 Kh7-h6 Bb8-d6 g5-g4 Bd6-b8
 28/40    00:07     66,455k    8,996k    -0.08    h7-h6 f4-f5 Qb8-c8 f5xe6 f7xe6 Qe2-c2 Ba8-d5 c3-c4 Bd5xc4 Na5xc4 Qc8xc4 Qc2xc4 b5xc4 Rd6-c6 Re8-b8 Rc6xc4 Na4-b6 Bb4-d6 Rb8-a8 Rc4-b4 Nb6-d5 Rb4-b8+ Ra8xb8 Bd6xb8 h6-h5 Kg1-f2 g7-g5 Kf2-f3 Kg8-h7 Bb8-d6 g5-g4+ Kf3-e4 Kh7-g6 Bd6-b8 Nd5-c3+ Ke4-d3 Nc3-d5 Kd3-c4 Kg6-f5 Kc4-c5
 29/42+    00:10     97,532k    8,941k     0.00    h7-h6
 29/42    00:10     97,741k    8,942k     0.00    h7-h6 f4-f5 Qb8-c8 f5xe6 f7xe6 Qe2-c2 Ba8-d5 c3-c4 Bd5xc4 Na5xc4 Qc8xc4 Qc2xc4 b5xc4 Rd6-c6 Re8-b8 Bb4-e1 c4-c3 Kg1-f2 Rb8-b1 Kf2-e2 Na4-b6 Be1xc3 Nb6-d5 Bc3-d2 Kg8-f7 Bd2-c1 Rb1-b4 Ke2-d3 Rb4-b3+ Kd3-e4 Nd5-c3+ Ke4-d3
 



Bryan

unread,
Jun 21, 2021, 1:20:52 AMJun 21
to FishCooking
In Game 4, at move 19, white to move, white (SF7) is already completely winning.

FEN: r3n1k1/pp1b1ppp/3q4/n2p2P1/P2P1N2/2PB1P2/7P/R1B1Q1K1 w - - 1 19

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 44/66    06:03     3,965,730k    10,901k    +5.55    Qe1-h4 g7-g6 Bc1-a3 Qd6-c6 Ba3-b4 Na5-c4 Ra1-e1 Ne8-g7 Nf4xd5 Qc6xd5 Bd3-e4 Ng7-f5 Qh4-f2 Qd5-e6 Be4xb7 Qe6-b6 Bb7xa8 a7-a5 Bb4-c5 Qb6-d8 Ba8-e4 Qd8xg5+ Qf2-g2 Qg5-f4 Kg1-h1 Bd7xa4 Qg2-g4 Qf4xg4 f3xg4 Nf5-e3 Be4-f3 Ba4-d7 d4-d5 Ne3xg4 Re1-e7 Ng4-e5 d5-d6 Ne5xf3 Re7xd7 Nc4-e5 Rd7-c7 f7-f6 d6-d7 Ne5xd7 Rc7xd7 a5-a4 Kh1-g2 Nf3-g5 Rd7-a7 Ng5-e4 Bc5-b4 f6-f5 c3-c4

At move 19, black played 19...g7-g6. If at move 18 black had played g7-g6, the position would have remained almost equal.

Move 19: white to move after 18...g7-g6: hash cleared: multipv=1: contempt=0: threads=15
FEN: r3n1k1/pp1b1p1p/2nq2p1/3p2P1/P2P1N2/2PB1P2/7P/R1B1Q1K1 w - - 0 19
Stockfish_21062023_x64_avx2:
Found 145 tablebases

NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 50/67    1:01:01     29,989,271k    8,191k    +0.21    Qe1-g3 Bd7-e6 h2-h4 Ne8-g7 Nf4-h5 Qd6xg3+ Nh5xg3 Ra8-c8 Bc1-a3 Be6-d7 Ng3-e2 Ng7-h5 Kg1-f2 Nc6-a5 Ba3-b4 Na5-c4 Bb4-e7 Nc4-b2 Bd3-c2 Rc8-e8 Be7-d6 a7-a6 Bd6-h2 b7-b5 a4xb5 Bd7xb5 Ne2-c1 h7-h6 Ra1-a2 Nb2-c4 Nc1-d3 Nc4-e3 Nd3-c5 h6xg5 h4xg5 Ne3xc2 Ra2xc2 f7-f6 g5xf6 Nh5xf6 Bh2-e5 Kg8-f7 Rc2-a2 Re8-c8 Kf2-g3 Nf6-d7 Nc5xd7 Bb5xd7 Ra2xa6 Bd7-f5 Ra6-f6+ Kf7-e7 Rf6-d6 Bf5-e6 Rd6-a6 Rc8xc3 Be5-d6+ Ke7-f7

So the losing mistake was 18...Nc6-a5.

SFdev abandons 18...Nc6-a5 and finds g7-g6 at d=17.

Move 18: black to move: hash cleared: multipv=1: contempt=0: threads=15
FEN: r3n1k1/pp1b1ppp/2nq4/3p2P1/P2P1N2/2PB1P2/7P/R1B1Q1K1 b - - 0 18

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 16/23    00:00     1,928k    8,530k    -0.43    Nc6-a5 Qe1-h4 g7-g6 Bc1-a3 Qd6-c6 Ba3-b4 Na5-c4 a4-a5 Ne8-g7 Ra1-e1 Bd7-f5 Bd3xc4 Qc6xc4 Re1-e5 Bf5-e6 Qh4-e1 b7-b6 Bb4-e7 b6xa5 Be7-f6 a5-a4 Bf6xg7 Kg8xg7 Nf4xe6+ f7xe6 Re5xe6
 17/31    00:00     3,336k    8,687k    -0.62    Nc6-a5 Qe1-h4 g7-g6 Bc1-a3 Qd6-c6 Ba3-b4 Na5-c4 a4-a5 Ne8-g7 Ra1-e1 Bd7-e6 Qh4-f2 b7-b6 a5-a6 Qc6-c8 Bd3xc4 Qc8xc4 Qf2-e2 Qc4-c6 Qe2-e5 Ra8-e8 Kg1-f2 Be6-d7
 18/31    00:01     10,440k    8,707k    -0.38    g7-g6 Ra1-b1 b7-b6 Nf4-g2 Bd7-e6 Bc1-f4 Qd6-d7 Rb1-b2 Ne8-g7 Bf4-g3 Ng7-h5 Rb2-e2 a7-a6 Re2-b2 Nc6-a5 Bg3-e5 Be6-f5 Bd3xf5
 19/31    00:01     12,012k    8,654k    -0.52    g7-g6 Ra1-b1 Nc6-a5 Qe1-e5 Qd6xe5 d4xe5 Bd7-c6 Kg1-f2 Ne8-g7 Bc1-a3 Ng7-e6 Nf4xe6 f7xe6 Ba3-b4 Na5-c4 Bd3xc4 d5xc4 a4-a5 h7-h6 h2-h4 h6xg5 h4xg5 b7-b6 a5xb6 a7xb6 Kf2-e3 b6-b5 Rb1-h1 Ra8-a2
 20/30    00:01     16,677k    8,628k    -0.30    g7-g6 Nf4-g2 Bd7-e6 Bc1-f4 Qd6-d7 Ng2-e3 Ne8-g7 h2-h4 Ra8-e8 Qe1-d2 a7-a6 Qd2-b2 Ng7-h5 Bf4-h2 Be6-h3 Kg1-f2 f7-f5 g5xf6/ep Nh5xf6
 21/28    00:03     27,503k    8,568k    -0.21    g7-g6 Ra1-b1 Nc6-a5 Qe1-e5 Qd6xe5 d4xe5 Bd7-c6 Kg1-f2 Ne8-c7 Bc1-a3 Nc7-e6 Nf4xe6 f7xe6 Ba3-b4 Na5-c4 Bd3xc4 d5xc4 a4-a5 Ra8-d8 Bb4-d6 h7-h6 g5xh6 Kg8-h7 Rb1-b4 Rd8-d7 f3-f4 Kh7xh6 Rb4xc4
 22/30    00:03     28,767k    8,562k    -0.24    g7-g6 Ra1-b1 Nc6-a5 Qe1-e5 Qd6xe5 d4xe5 Ne8-c7 Kg1-f2 b7-b6 Bc1-a3 Bd7xa4 Ba3-

The mistake in these early pv lines is the assumption that white will play a4-a5 (marked with bold font). That a4-a5 is not necessary for white. White has a much better line that, after Ra1-e1, includes Nf4xd5, then after Qc6xd5, Bd3-e4, then capturing with Be4xb7 and then capturing the rook on a8 with Bb7xa8. This better line does not include a4-a5.


Bryan

unread,
Jun 21, 2021, 9:23:51 AMJun 21
to FishCooking
In Game 5 SFdev made two suboptimal moves. The first was at move 11, where SFdev played g7-g6. This lost about 120 centipawns, and the better move here was Bf8-e7, which SFdev played in the game at move 17. If at move 11 SFdev had played Bf8-f7, and white had used the same subsequent line of play, then at move 16 black could play Nd7-c5 or O-O and have a significant advantage (+1.00). What g7-g6 does, among other things, is occupy the g6 square, making that square unavailable to the black queen. So if black plays 11...Bf8-e7, and white tries the same line as in the game, then after 16...Nd7-c5 17.Be2, black can play 17...Qh5-g6. And of course 11...Bf8-e7 allows the black king to castle O-O at move 16, whereas in the game the black king cannot castle at move 16, because the bishop is still on f8.

Move 11: black to move: hash from subsequent moves: multipv=1: contempt=0: threads=15
FEN: r1b1kb1r/2qn1ppp/pp1ppn2/8/2P1PP2/1NNB4/PP2Q1PP/R1B2RK1 b kq - 1 11

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 40/54    13:57     6,585,282k    7,866k    -0.67    Bf8-e7 Bc1-e3 Bc8-b7 Nb3-d2 Ra8-c8 Ra1-c1 g7-g6 Kg1-h1 O-O a2-a3 Qc7-b8 Bd3-c2 Nf6-h5 Bc2-a4 Nd7-f6 b2-b4 Nh5-g7 Kh1-g1 Bb7-a8 Ba4-c2 b6-b5 Bc2-d3 d6-d5 c4xd5 Rf8-d8 Kg1-h1 Rc8xc3 Rc1xc3 e6xd5 e4-e5 d5-d4 e5xf6 Be7xf6 Rc3-c2 d4xe3 Qe2xe3 Bf6-d4 Qe3-h3 Ng7-h5 Bd3-e4 Nh5xf4

Move 12: black to move: hash from subsequent moves: multipv=1: contempt=0: threads=15
FEN: r1b1kb1r/2qn1p1p/pp1ppnp1/4P3/2P2P2/1NNB4/PP2Q1PP/R1B2RK1 b kq - 0 12

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
 [...]
 40/67    02:56     1,450,210k    8,227k    -1.83    d6xe5 f4xe5 Qc7xe5 Qe2-f3 Ra8-a7 Bc1-f4 Qe5-h5 Qf3-e3 Bc8-b7 Ra1-d1 Bb7-a8 Qe3-d2 Bf8-c5+ Nb3xc5 Qh5xc5+ Bf4-e3 Qc5-c6 h2-h3 O-O Qd2-f2 Nd7-e5 Rd1-d2 Nf6-d7 Bd3-e4 Qc6xc4 Rd2-d4 Qc4-c7 Be3-f4 Ba8-c6 Rf1-d1 Qc7-c8 Bf4-h6 Bc6-b5 Nc3xb5 a6xb5 Bh6xf8 Nd7xf8 Rd4-d8 Qc8-c7 b2-b4 Kg8-g7 Qf2-d4 Kg7-g8 Qd4-d6 Qc7xd6 Rd1xd6 Ra7xa2

At move 11 the latest SFdev abandons g7-g6 and finds Bf8-e7 at d=20

Move 11: black to move: hash cleared: multipv=1: contempt=0: threads=15
FEN: r1b1kb1r/2qn1ppp/pp1ppn2/8/2P1PP2/1NNB4/PP2Q1PP/R1B2RK1 b kq - 1 11

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 17/23    00:00     4,354k    7,734k    -0.91    g7-g6 Rf1-e1 Bf8-e7 Bc1-e3 O-O Nb3-d2 Nf6-h5 Qe2-f2 Bc8-b7 Ra1-c1 Ra8-c8 a2-a3 f7-f5 e4xf5 g6xf5 Bd3-e2 Nh5-g7
 18/23    00:00     5,870k    7,765k    -0.66    g7-g6 Rf1-e1 Bf8-e7 Bc1-e3 O-O Nb3-d2 Bc8-b7 Ra1-c1 Nf6-h5 Qe2-f2 Nh5-g7 b2-b4 f7-f5 Nc3-a4 f5xe4 Nd2xe4 Ra8-b8 Na4xb6
 19/24    00:00     7,644k    7,706k    -0.70    g7-g6 Rf1-e1 Bf8-e7 Bc1-e3 Bc8-b7 Ra1-d1 O-O Nb3-d2 Nf6-h5 Qe2-f2 Nh5-g7 f4-f5 g6xf5 e4xf5 Nd7-e5 Bd3-b1 Ne5-g4 Qf2-g3 Ng4xe3 Qg3xe3 Qc7-c5 Qe3xc5
 20/40    00:02     21,520k    8,000k    -0.94    Bf8-e7 Bc1-e3 O-O Nb3-d2 Bc8-b7 Ra1-c1 g7-g6 a2-a3 Nf6-h5 b2-b4 Ra8-e8 Bd3-c2 Be7-f6 Bc2-a4 Bf6xc3 Rc1xc3 f7-f5 Qe2-d3 e6-e5 e4xf5 g6xf5
 21/22    00:02     22,905k    8,000k    -0.94    Bf8-e7 Bc1-e3 Bc8-b7 Nb3-d2 O-O Ra1-c1 g7-g6 b2-b4 Rf8-c8 a2-a3 a6-a5 h2-h3 Qc7-d8 f4-f5 e6xf5 e4xf5 a5xb4 a3xb4 d6-d5 f5xg6 h7xg6
 
SFdev's second suboptimal move was at move 16, where SFdev lost about 170 centipawns by playing Qh5-g4 rather than g6-g5. The problem with Qh5-h4 is that it allows the black queen to get pushed around to b4 (because of few available squares for the black queen), and this allows white to fork the black queen and black's bishop on e7. But this dark-squared bishop is more valuable for black, precisely because of black's dark-square weaknesses on the kingside caused by 11...g7-g6.

Move 16: black to move: hash from subsequent moves: multipv=3: contempt=0: threads=15
FEN: 2b1kb1r/r2n1p1p/pp2pnp1/7q/2P2B2/1NNB2Q1/PP4PP/R4RK1 b k - 5 16
Stockfish_21062023_x64_avx2:
[...]
---------------------------------------------------------------------------
 44/64    31:17     14,817,111k    7,893k    -3.20    Qh5-g4 Qg3-f2 Qg4-h5 Ra1-d1 Nf6-g4 Qf2-e1 Bf8-h6 Qe1-g3 g6-g5 Bf4-d6 Qh5-h4 Qg3-f3 Nd7-e5 Bd6xe5 Ng4xe5 Qf3-f6 Qh4xh2+ Kg1xh2 Ne5-g4+ Kh2-g1 Ng4xf6 Rf1xf6 Bh6-g7 Rf6-f2 O-O Nc3-a4 Ra7-b7 Rf2-c2 b6-b5 Na4-c5 b5xc4 Bd3xc4 Rb7-c7 Bc4-e2 Bc8-b7 Rc2-d2 Bb7-d5 Nc5xa6 Rc7-b7 Be2-f3 Bd5xb3 Bf3xb7 Bb3xd1 Rd2xd1 Bg7xb2 a2-a4 Bb2-c3 Kg1-f1 h7-h5 Na6-c5 Bc3-a5 Rd1-b1 Kg8-g7 Nc5-e4 Ba5-d8 Rb1-b5 Kg7-h6 Kf1-e2 g5-g4
  44/75    31:17     14,817,111k    7,893k    -1.53    g6-g5 Bf4-e5 Nd7xe5 Qg3xe5 Nf6-g4 Qe5-b8 Qh5xh2+ Qb8xh2 Ng4xh2 Kg1xh2 h7-h5 Ra1-d1 h5-h4 Bd3-e2 Bf8-g7 Nc3-e4 Bg7-e5+ Kh2-g1 f7-f5 Ne4xg5 Ra7-g7 Ng5-h3 Be5xb2 c4-c5 b6xc5 Nb3xc5 Ke8-e7 Kg1-h1 Bb2-a3 Nc5-d3 Rh8-d8 Nh3-f4 Ba3-d6 Be2-f3 h4-h3 g2xh3 Rg7-g3 Bf3-g2 Rg3-e3 Rd1-e1 Re3xe1 Rf1xe1 Ke7-f6 Nf4-h5+ Kf6-g5 Bg2-f3 Bd6-g3 Re1-d1 Kg5-h4 Kh1-g2 Bg3-b8 Rd1-b1 Bb8-d6 Nh5-f4 Bd6xf4 Nd3xf4 Rd8-d2+ Bf3-e2 Rd2xa2
---------------------------------------------------------------------------

At move 16 the latest SFdev abandons Qh5-g4 and finds g6-g5 at d=18

Move 16: black to move: hash cleared: multipv=1: contempt=0: threads=15
FEN: 2b1kb1r/r2n1p1p/pp2pnp1/7q/2P2B2/1NNB2Q1/PP4PP/R4RK1 b k - 5 16

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
 [...]
 15/25    00:00     3,448k    8,751k    -1.95    Qh5-g4 Qg3-f2 Qg4-h5 Ra1-d1 g6-g5 Bd3-e2 Qh5-g6 Bf4-e3 Bf8-e7 Be3xb6 Nd7xb6 Qf2xb6 Ra7-d7 Nb3-d4 Nf6-e4 Nc3xe4 Qg6xe4 Nd4-c6 Rd7xd1
 16/25    00:00     4,519k    8,658k    -2.04    Qh5-g4 Qg3-f2 Bf8-e7 h2-h3 Qg4-h5 Bd3-e2 Qh5-f5 Qf2-e3 e6-e5 Bf4-h6 Qf5-e6 Ra1-d1 Bc8-b7 Bh6-g7 Rh8-g8 Bg7xf6 Nd7xf6 c4-c5 Ra7-a8 c5xb6 Ra8-c8
 17/24    00:00     6,298k    8,675k    -2.05    Qh5-g4 Qg3-f2 Bf8-e7 h2-h3 Qg4-h5 Bd3-e2 Qh5-f5 Qf2-e3 e6-e5 Bf4-h6 Qf5-e6 Ra1-d1 Bc8-b7 Bh6-g7 Rh8-g8 Bg7xf6 Nd7xf6 c4-c5 Ra7-a8 c5xb6 Ra8-c8 Kg1-h1 Be7-b4 Be2-f3
 18/29    00:01     10,561k    8,664k    -2.20    g6-g5 Bf4-e5 Nd7xe5 Qg3xe5 Nf6-g4 Qe5-b8 Qh5xh2+ Qb8xh2 Ng4xh2 Kg1xh2 h7-h5 Nc3-a4 Ra7-b7 Ra1-d1 Bc8-d7 Bd3-e4 Rb7-b8 Rd1xd7 Ke8xd7 Rf1xf7+ Kd7-d6 Be4-b7 Bf8-e7 Bb7xa6
 19/33    00:01     12,666k    8,640k    -2.07    g6-g5 Bf4-e5 Nd7xe5 Qg3xe5 Nf6-g4 Qe5-b8 Qh5xh2+ Qb8xh2 Ng4xh2 Kg1xh2 h7-h5 Ra1-e1 Ra7-d7 Bd3-e2 Bf8-d6+ Kh2-g1 f7-f5 Nc3-a4 Bd6-c7 c4-c5




Bryan

unread,
Jun 21, 2021, 10:02:30 AMJun 21
to FishCooking
In Game 6 SFdev as white blundered into a rook trap at move 25.

At move 25 there were at least four moves available that (likely) save the draw.

Move 25: white to move: hash cleared: multipv=4: contempt=0: threads=15
FEN: r4rk1/p1P1Rppp/2N1p3/2n5/8/P5PB/5P1P/5K2 w - - 1 25
Stockfish_21062023_x64_avx2:
[...]
---------------------------------------------------------------------------
 50/65    11:59     6,142,307k    8,540k     0.00    Bh3-g2 a7-a5 Kf1-e2 Rf8-c8 Ke2-e3 Kg8-f8 Bg2-f1 a5-a4 Bf1-c4 h7-h5 h2-h4 f7-f6 Ke3-d4 Nc5-b3+ Kd4-d3 Ra8-a5 Kd3-c3 Ra5-c5 Re7-d7 Rc5xc6 Rd7-d8+ Kf8-f7 Rd8xc8 Nb3-a5 Rc8-a8 Rc6xc4+ Kc3-d3 Rc4xc7 Ra8xa5 Rc7-b7 Ra5xa4 Rb7-b2 Ra4-a7+ Kf7-g6 Kd3-e3 Rb2-a2 a3-a4 Ra2-a3+ Ke3-e2 e6-e5 Ke2-f1 Kg6-f5 Ra7xg7 Ra3xa4 Kf1-g2 Ra4-a2 Kg2-g1 e5-e4 Rg7-g8 Ra2-d2 Rg8-g7 Rd2-c2 Kg1-f1 Rc2-a2 Rg7-f7 Ra2-a6 Rf7-h7 Kf5-g6 Rh7-e7 Kg6-f5
 50/21    11:59     6,142,307k    8,540k     0.00    Nc6-b8 Rf8-c8 Nb8-d7 Nc5-a6 Bh3-g2 Na6xc7 Bg2xa8 Nc7xa8 h2-h4 Na8-c7 Nd7-e5 f7-f6 Ne5-c6 Nc7-b5 Re7xe6 Kg8-f7 Re6-e7+ Kf7-f8 Re7-e6
 50/66    11:59     6,142,307k    8,540k     0.00    f2-f4 Ra8-c8 Nc6xa7 g7-g6 g3-g4 Kg8-g7 g4-g5 Rc8-a8 Na7-b5 Ra8-a6 Bh3-g4 h7-h6 f4-f5 e6xf5 Bg4-f3 h6xg5 Bf3-d5 Kg7-h6 Bd5xf7 Ra6-a8 Bf7-d5 Ra8-a5 Re7-f7 Rf8-h8 Nb5-d6 Nc5-a6 c7-c8Q Rh8xc8 Nd6xc8 Ra5xd5 Nc8-e7 Rd5-d1+ Kf1-g2 Rd1-d2+ Kg2-g3 Rd2-d3+ Kg3-g2 g5-g4 Rf7-f6 Rd3-d2+ Kg2-f1 Na6-c7 Rf6xg6+ Kh6-h5 Rg6-f6 Rd2xh2 Rf6-c6 f5-f4 Rc6-c5+ Kh5-h4 Ne7-f5+ Kh4-h3 Rc5xc7 g4-g3 Rc7-h7+ Kh3-g4 Rh7xh2 g3xh2 Kf1-g2 Kg4xf5
 50/44    11:59     6,142,307k    8,540k     0.00    Kf1-e2 a7-a5 Bh3-f1 Rf8-c8 Ke2-e3 Kg8-f8 Ke3-d4 Nc5-a6 Bf1xa6 Ra8xa6 Re7-d7 Ra6xc6 Rd7-d8+ Kf8-e7 Rd8xc8 Ke7-d6 Rc8-g8 Rc6xc7 Rg8xg7 e6-e5+ Kd4-e3 f7-f5 Rg7xc7 Kd6xc7 f2-f3 Kc7-d6 g3-g4 Kd6-e6 g4xf5+ Ke6xf5 a3-a4 h7-h5 h2-h4 Kf5-e6 Ke3-e4 Ke6-d6 Ke4-f5 Kd6-d5 Kf5-g5 e5-e4 f3xe4+ Kd5xe4 Kg5xh5
---------------------------------------------------------------------------

Instead, at move 25 white played Nc6xa7, which traps the white rook on e7, because after Nc6xa7 the white knight cannot land on d8 (or e5) and therefore cannot create a safe square on f7 for the white rook, and the white knight cannot (after Nb8) land on d7 to keep the black king from traversing the f8 square.

I tried three times (each time on cleared hash) to get SFdev to prefer 25.Nc6xa7 at some iteration, and it never did. So maybe this weakness is solved.

Move 25: hash cleared: multipv=1: contempt=0: threads=15
FEN: r4rk1/p1P1Rppp/2N1p3/2n5/8/P5PB/5P1P/5K2 w - - 1 25

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 17/39    00:00     5,267k    9,506k    +0.03    Bh3-g2 a7-a5 Nc6-d4 Ra8-a6 Bg2-c6 Rf8-c8 Bc6-d7 Ra6-a8 Bd7xc8 Ra8xc8 Nd4-c6 Kg8-f8 a3-a4 Nc5-a6 Re7-d7 Na6xc7 Nc6-a7 Kf8-e8 Rd7xf7 Ke8xf7 Na7xc8 Nc7-d5 Nc8-d6+ Kf7-f6 Nd6-c4 Nd5-c3 Nc4xa5 Nc3xa4 Na5-c4 Na4-c5
 18/27    00:00     5,780k    9,460k    +0.02    Bh3-g2 a7-a5 Nc6-d4 Ra8-a6 Bg2-c6 Rf8-c8 Bc6-d7 Ra6-a8 Bd7xc8 Ra8xc8 Nd4-c6 Kg8-f8 a3-a4 Nc5-a6 Re7-d7 Na6xc7 Nc6xa5 Kf8-e8 Rd7-d4 Rc8-b8 Na5-c4 Rb8-b1+ Kf1-e2 g7-g6 Rd4-h4 Rb1-a1 Rh4xh7
 19/32    00:00     7,093k    9,395k     0.00    f2-f4 g7-g6 Kf1-f2 Rf8-c8 Bh3-f1 Nc5-e4+ Kf2-g2 Ne4-c5
 20/32    00:00     9,333k    9,361k     0.00    Bh3-g2 a7-a5 Nc6-d4 Ra8-a6 Bg2-c6 Rf8-c8 Bc6-d7 Ra6-a8 Bd7xc8 Ra8xc8 a3-a4 Kg8-f8 Nd4-c6 Nc5-a6 Re7-d7 Na6xc7 Nc6xa5 Kf8-e8 Rd7-d1 Rc8-a8 Na5-c6 Ra8-c8 a4-a5 Nc7-d5 Nc6-d4 Rc8-c4
 21/35    00:01     15,105k    9,365k     0.00    f2-f4 g7-g6 Kf1-f2 Rf8-c8 Bh3-f1 Nc5-e4+ Kf2-g1 a7-a5 Re7-d7 Ne4-c5 Rd7-d8+ Kg8-g7 Bf1-b5 Nc5-b7 Rd8-d7 Nb7-c5
 22/32    00:02     20,029k    9,307k     0.00    f2-f4 Rf8-c8 Nc6-d8 Rc8xd8 c7xd8Q+ Ra8xd8 Re7xa7 h7-h5 Bh3-g2 Rd8-d2 a3-a4 Rd2-a2 a4-a5 Ra2-a1+ Kf1-f2 Ra1-a2+
 23/17    00:02     22,344k    9,325k     0.00    f2-f4 g7-g6 Kf1-f2 Rf8-c8 Bh3-f1 Nc5-e4+ Kf2-g1 a7-a6 Re7-d7 Ne4-c5 Rd7-d8+ Kg8-g7 Bf1-g2 Nc5-b7 Rd8-d7 Nb7-c5
 24/20    00:02     26,947k    9,353k     0.00    Kf1-e2 a7-a5 Ke2-e3 Rf8-c8 Bh3-f1 Kg8-f8 Ke3-d4 Nc5-a6 Bf1xa6 Ra8xa6 Re7-d7 Ra6xc6 Rd7-d8+ Kf8-e7 Rd8xc8 h7-h5 Kd4-d3 Rc6-c5 Kd3-d4
 25/37    00:02     27,777k    9,365k     0.00    Kf1-e2 a7-a5 Ke2-e3 Rf8-c8 Bh3-f1 Kg8-f8 Ke3-d4 Nc5-a6 Bf1xa6 Ra8xa6 Re7-d7 Ra6xc6 Rd7-d8+ Kf8-e7 Rd8xc8 h7-h5 Kd4-d3 Rc6-c5 Kd3-d4
 26/40    00:03     34,153k    9,414k     0.00    Kf1-e2 a7-a5 Ke2-e3 Rf8-c8 Bh3-f1 Kg8-f8 Ke3-d4 Nc5-a6 Bf1xa6 Ra8xa6 Re7-d7 Ra6xc6 Rd7-d8+ Kf8-e7 Rd8xc8 h7-h5 Kd4-d3 Rc6-c5 h2-h4 g7-g6 a3-a4 Ke7-d6 Rc8-a8 Rc5-d5+ Kd3-e4 Rd5-e5+ Ke4-d4 Re5-d5+
 27/42    00:03     36,021k    9,420k     0.00    Kf1-e2 a7-a5 Ke2-e3 Rf8-c8 Bh3-f1 Kg8-f8 Ke3-d4 Nc5-a6 Bf1xa6 Ra8xa6 Re7-d7 Ra6xc6 Rd7-d8+ Kf8-e7 Rd8xc8 a5-a4 Kd4-e3 g7-g6 Ke3-d2 e6-e5 Rc8-a8 Rc6xc7 Ra8xa4 f7-f5 Ra4-a8 Rc7-b7 Kd2-c2 Rb7-c7+
 28/42    00:04     41,809k    9,353k     0.00    Kf1-e2 a7-a5 Ke2-e3 Rf8-c8 Bh3-f1 Kg8-f8 Ke3-d4 Nc5-a6 Bf1xa6 Ra8xa6 Re7-d7 Ra6xc6 Rd7-d8+ Kf8-e7 Rd8xc8 g7-g5 a3-a4 g5-g4 Kd4-d3 Rc6-c5 Rc8-g8 Rc5xc7 Rg8xg4 Rc7-c5 h2-h4 Rc5-d5+ Rg4-d4 Rd5-c5
 29/46    00:05     52,345k    9,322k     0.00    Kf1-e2 a7-a5 Ke2-e3 Rf8-c8 Bh3-f1 Kg8-f8 Ke3-d4 Nc5-a6 Bf1xa6 Ra8xa6 Re7-d7 Ra6xc6 Rd7-d8+ Kf8-e7 Rd8xc8 h7-h5 Kd4-d3 Rc6-c5 h2-h4 g7-g6 a3-a4 Ke7-d6 Rc8-a8 Rc5-d5+ Kd3-e4 Rd5-e5+ Ke4-d4 Re5-d5+ Kd4-c3
 30/43    00:07     72,152k    9,284k     0.00    Kf1-e2 a7-a5 Ke2-e3 Rf8-c8 Bh3-f1 Kg8-f8 Ke3-d4 Nc5-a6 Bf1xa6 Ra8xa6 Re7-d7 Ra6xc6 Rd7-d8+ Kf8-e7 Rd8xc8 g7-g5 a3-a4 h7-h6 h2-h4 g5xh4 g3xh4 Ke7-d6 Rc8-a8 Rc6xc7 Ra8xa5 Rc7-c2 f2-f4 Rc2-h2 Ra5-a7 Rh2xh4 Ra7xf7 e6-e5+ Kd4-e4 Kd6-e6 Rf7-f5 e5xf4 Rf5xf4 Rh4-h2 Rf4-f3 h6-h5 Ke4-f4 Ke6-e7 Kf4-g3
 31/42    00:08     79,441k    9,282k     0.00    Kf1-e2 a7-a5 Ke2-e3 Rf8-c8 Bh3-f1 Kg8-f8 Ke3-d4 Nc5-a6 Bf1xa6 Ra8xa6 Re7-d7 Ra6xc6 Rd7-d8+ Kf8-e7 Rd8xc8 g7-g5 a3-a4 h7-h6 h2-h4 g5xh4 g3xh4 Ke7-d6 Rc8-a8 Rc6xc7 Ra8xa5 Rc7-c2 f2-f4 Rc2-f2 Kd4-e3 Rf2-h2 h4-h5 Rh2-h3+ Ke3-d4 Rh3-h4 Kd4-e3 f7-f5 Ra5-a6+ Kd6-d5 Ra6-a5+ Kd5-d6




Bryan

unread,
Jun 21, 2021, 11:26:48 AMJun 21
to FishCooking
In Game 7, SFdev blundered at move 53 by playing 53...Ra8-d8. After that, it is +M23.

Move 54: white to move: hash cleared: multipv=1: contempt=0: threads=15
FEN: 3r3r/P6P/3k4/7R/3ppp2/5p2/5K1P/6R1 w - - 6 54
Stockfish_21062023_x64_avx2:
Found 145 tablebases
[...]
 52/46    06:04     6,279,678k    17,224k    +M23    Rg1-g6+ Kd6-c7 Rh5-h6 e4-e3+ Kf2xf3 e3-e2 Rg6-g7+ Rd8-d7 Rg7xd7+ Kc7xd7 Kf3xe2 f4-f3+ Ke2xf3 d4-d3 a7-a8R Rh8xa8 h7-h8Q Ra8xh8 Rh6xh8 Kd7-e7 Rh8-g8 Ke7-f7 Rg8-g1 Kf7-f6 h2-h4 d3-d2 Kf3-e2 d2-d1Q+ Ke2xd1 Kf6-f7 h4-h5 Kf7-f8 h5-h6 Kf8-f7 h6-h7 Kf7-e7 h7-h8Q Ke7-e6 Rg1-g5 Ke6-e7 Qh8-h6 Ke7-d7 Rg5-g7+ Kd7-e8 Qh6-h8+

At move 53, black to move, there were at least five moves that (likely) save the draw.

Move 53: black to move: hash cleared: multipv=5: contempt=0: threads=15
FEN: r6r/P6P/3k4/7R/3ppp2/5p2/5K1P/6R1 b - - 5 53
Stockfish_21062023_x64_avx2:
[...]
---------------------------------------------------------------------------
 51/29    17:28     10,083,807k    9,617k     0.00    Ra8-e8 Rg1-g6+ Kd6-d7 Rh5-d5+ Kd7-c7 Rg6-g7+ Kc7-c6 Rd5xd4 e4-e3+ Kf2xf3 e3-e2 a7-a8Q+ Re8xa8 Kf3xe2 Ra8-a2+ Ke2-f3 Ra2xh2 Rd4-d7 Rh2-h4 Rd7-a7 Kc6-d5 Rg7-f7 Kd5-e5 Ra7-e7+ Ke5-d5 Re7-b7 Kd5-e5
 51/12    17:28     10,083,807k    9,617k     0.00    d4-d3 Rg1-g6+ Kd6-d7 Rh5-h6 Ra8-e8 Rg6-b6 Kd7-c7 Rh6-c6+ Kc7-d7 Rc6-d6+ Kd7-c7
 51/37    17:28     10,083,807k    9,617k     0.00    Kd6-e7 Rg1-g7+ Ke7-f6 Rg7-c7 Kf6-e6 Rc7-b7 Rh8-c8 Rb7-b6+ Ke6-f7 h7-h8Q Rc8xh8 Rh5xh8 Ra8xh8 Rb6-b8 Rh8xh2+ Kf2-g1 Rh2-g2+ Kg1-f1 e4-e3 Rb8-f8+ Kf7-e7 Rf8-e8+ Ke7xe8 a7-a8Q+ Ke8-d7 Qa8-a7+ Kd7-d6 Qa7xd4+ Kd6-c6 Qd4-c4+ Kc6-d7 Qc4-d5+ Kd7-c8 Qd5xf3 Rg2-f2+ Qf3xf2
 51/23    17:28     10,083,807k    9,617k     0.00    Rh8-c8 Rg1-a1 d4-d3 h7-h8R Rc8xh8 Ra1-a6+ Kd6-e7 Rh5xh8 Ra8xh8 a7-a8Q Rh8xa8 Ra6xa8 d3-d2 Ra8-a1 Ke7-f6 Ra1-d1 Kf6-g5 Rd1xd2 e4-e3+ Kf2-e1 e3xd2+ Ke1xd2
 51/19    17:28     10,083,807k    9,617k     0.00    Kd6-e6 Rg1-g6+ Ke6-f7 Rg6-a6 Kf7-g7 Rh5-e5 Rh8-c8 h7-h8Q+ Kg7xh8 Re5xe4 Rc8-c7 Re4xf4 Rc7xa7 Ra6xa7 Ra8xa7 Rf4xd4 Ra7-a2+ Kf2xf3
---------------------------------------------------------------------------

The reason 53...Ra8-d8 is a blunder is because it either loses a tempo when white queens (a7-a8Q+) with a safe check or it allows white to exchange off one of black's rooks. If black plays 53...Ra8-e8, then if white plays Rg6+ black can play Kd6-d7 and if Rhh6 black can play d4-d3 and if white then plays Rg6-g7+, and black blocks check with Re8-e7, then if white goes on to play Rg7-g8 then black can queen the d-pawn, and when white queens on a8, it does not come with check, since the black king is on d7, and the game is drawn.

But after 53...Ra8-d8, then after Rg6+, black has only three moves available: Kd7, Kc7, and Ke7. All three are mates. If black plays Kd6-d7, white can skewer the d8 rook with Rh5-d5, and black cannot stop one of white's a-file or h-file pawns from queening, and cannot queen its d-pawn. And if black plays Kc7 and then black plays d4-d3, Rg6-g7+ Rd8-d7 (to block check) Rg7-g8 d3-d2, white has Rh6-c6+, and if black replies with Kxc6, white has a7-a8Q+ with a safe check. So after 53...Ra8-d8 Rg6+ Kc7 Rhh6 black's king is (essentially) forced to a8, which allows white to exchange off one of black's rooks, and queen the h7 pawn. And if after 53...Ra8-d8 Rg6+ Ke7, white can play Rb5, and exchange off one of black's rooks with Rb5-b7+ Rd8-d7 Rb6xd7, and then white queens on a8.

Bryan

unread,
Jun 21, 2021, 12:48:58 PMJun 21
to FishCooking
In Game 8 SFdev as black blundered at move 61. There black has a move that allows black to retain some advantage: e5-e4.

Move 61: black to move: hash cleared: multipv=1: contempt=0: threads=15
FEN: 7r/k5q1/1pQP4/p1n1p3/P2p2r1/R2P4/5PB1/4RK2 b - - 1 61

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 53/67    05:44     3,136,128k    9,091k    +0.60    e5-e4 Qc6-c7+ Qg7xc7 d6xc7 Ka7-b7 Bg2xe4+ Kb7xc7 Be4-f3 Rg4-f4 Bf3-g2 Kc7-d7 Re1-e5 Rh8-f8 Re5-e2 Kd7-d6 Re2-b2 Kd6-c7 Rb2-e2 Kc7-d7 Re2-c2 Rf8-d8 Rc2-b2 Kd7-c7 Rb2-e2 Rd8-d7 Re2-c2 Rd7-d6 Rc2-e2 Rd6-g6 Re2-e7+ Kc7-d6 Re7-e8 Rg6-f6 Re8-e2 Rf4-g4 Re2-e8 Rf6-f7 Bg2-e4 Rg4-f4 Ra3-a2 Rf7-f8 Re8xf8 Rf4xf8 Kf1-e2 Rf8-f4 Be4-g2 Rf4-h4 Bg2-e4 Rh4-h6 Be4-a8 Kd6-e5 Ba8-g2 Rh6-g6 Bg2-a8 Rg6-g1 Ba8-c6 Rg1-b1 Bc6-b5 Ke5-f5 Ke2-f3 Rb1-b4 Ra2-a1 Rb4-b3

But in the game SFdev played 61...Rh8-g8, and this loses immediately because it allows white to capture the loose pawn on e5 with 62.Re1xe5, with white's next move being a pawn-protected rook fork of black's king and queen at 63.Re5-e7+ (or a protected pin of the black queen if 62...Qg7-b7 followed by 63.Re5-e7). Black cannot capture the white rook with 62...Qg7xe5, because then white has +M12 with 63.Ra3-b3, double-attacking b6, and if white replies with Nc5xb3, white has 64.Qc6-b7#. So the knight on c5 is doing double duty (overloaded) guarding both the b7 and b3 squares.

Move 62: black to move: hash cleared: multipv=3: contempt=0: threads=3
FEN: 6r1/k5q1/1pQP4/p1n1R3/P2p2r1/R2P4/5PB1/5K2 b - - 0 62
Stockfish_21062023_x64_avx2:
[...]
---------------------------------------------------------------------------
 36/51    18:40     20,819,843k    18,573k    -152.45    Qg7-b7 Re5-e7 Rg4xg2 Re7xb7+ Nc5xb7 [after Nc5xb7 this position is +M21] Ra3-b3 Rg2-g1+ Kf1-e2 Rg1-g7 Qc6xb6+ Ka7-a8 Qb6-a6+ Ka8-b8 Qa6xa5 Rg7-d7 Qa5-c5 Kb8-a8 Qc5-c6 Rg8-g7 a4-a5 Rd7xd6 Qc6-c8+ Ka8-a7 Rb3xb7+ Rg7xb7 Qc8-c5+ Ka7-a8 Qc5xd6 Rb7-h7 Qd6xd4 Ka8-b8 Qd4-b6+ Kb8-c8 Qb6-c6+ Kc8-b8 a5-a6 Rh7-e7+ Qc6-e4 Re7-d7 Qe4-e8+ Kb8-c7 Qe8xd7+ Kc7xd7
 36/52    18:40     20,819,843k    18,573k    -49.29    Rg4-g6 Re5-e7+ Qg7xe7 d6xe7 Rg6xc6 Bg2xc6 Ka7-b8 e7-e8Q+ Rg8xe8 Bc6xe8 Kb8-c7 Be8-b5 Nc5-e6 Ra3-a1 Ne6-f4 Ra1-c1+ Kc7-d6 Rc1-c6+ Kd6-d5 Rc6xb6 Kd5-c5 Rb6-f6 Nf4-d5 Rf6-c6+ Kc5-b4 Rc6-c4+ Kb4-a3 Rc4xd4 Nd5-c3 f2-f4 Ka3-b3 f4-f5 Nc3xb5 a4xb5 a5-a4 Rd4-c4 Kb3-a3 f5-f6 Ka3-b3 f6-f7
 36/60    18:40     20,819,843k    18,573k    -16.83    Rg4xg2 Re5-e7+ Qg7xe7 d6xe7 Rg2-g1+ Kf1-e2 Rg1-g6 Qc6-c7+ Ka7-a6 Ke2-d2 Rg6-g2 Qc7-f4 Rg8-e8 Qf4-f7 Rg2-g8 Ra3-a1 Re8-c8 Ra1-c1 Rg8-e8 Rc1-c4 Nc5-b3+ Kd2-d1 Nb3-c5 Kd1-e2 Rc8-b8 f2-f4 Re8-c8 Rc4xd4 Rc8-h8 Ke2-d2 Rh8-h2+ Kd2-e1 Rh2-h1+ Ke1-f2 Rh1-h8 Kf2-e2 Rh8-h2+ Ke2-f3 Rh2-h8 f4-f5 Rh8-h3+ Kf3-g4 Rh3-h8 f5-f6 Nc5xa4 Qf7-c4+ b6-b5 Rd4-d6+ Na4-b6 Qc4-c7 Rb8-b7
---------------------------------------------------------------------------

SFdef abandons Rh8-g8 and finds e5-e4 at d=16. But in the d=14 and d=15 iterations where SFdev prefers 61...Rh8-g8, it expects white to reply with Bg2-d5, attacking that rook on g8. At those lower depths it does not see that white has the much more crushing reply of Re1xe5.

Move 61: black to move: hash cleared: multipv=1: contempt=0: threads=15
FEN: 7r/k5q1/1pQP4/p1n1p3/P2p2r1/R2P4/5PB1/4RK2 b - - 1 61


Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 6/6    00:00     25k    8,263k    +1.32    e5-e4 Qc6-c7+ Qg7xc7 d6xc7 Rg4-g7
 7/7    00:00     31k    7,786k    +1.34    e5-e4 Bg2xe4 Rg4-g1+ Kf1-e2 Qg7-g4+ Be4-f3 Rg1xe1+ Ke2xe1
 8/13    00:00     69k    8,658k    +1.32    e5-e4 Bg2xe4 Rg4-g1+ Kf1-e2 Rg1xe1+ Ke2xe1 Qg7-g1+ Ke1-e2
 9/10    00:00     90k    8,976k    +1.32    e5-e4 Bg2xe4 Rg4-g1+ Kf1-e2 Rg1xe1+ Ke2xe1 Qg7-g1+ Ke1-e2 Rh8-f8 Be4-f3
 10/13    00:00     140k    9,351k    +1.35    Rh8-g8 Re1-b1 Qg7-b7 Qc6xb7+ Nc5xb7 Bg2-d5 Rg4-g1+ Kf1-e2 Rg1xb1 Bd5xg8 Nb7xd6 Bg8-d5
 11/18    00:00     229k    8,811k    +1.62    Rh8-g8 Re1-b1 Qg7-b7 Qc6xb7+ Nc5xb7 Bg2-d5 Rg4-g1+ Kf1-e2 Rg1xb1 Bd5xg8
 12/25    00:00     884k    9,116k    +0.99    e5-e4 Qc6-c7+ Qg7xc7 d6xc7 Ka7-b7 Bg2xe4+ Kb7xc7 Ra3-a2 Rh8-g8 Kf1-e2 Nc5xe4 d3xe4 Rg4xe4+ Ke2-d2 Re4xe1
 13/24    00:00     1,109k    9,090k    +1.09    e5-e4 Qc6-c7+ Qg7xc7 d6xc7 Ka7-b7 Bg2xe4+ Kb7xc7 Ra3-a2 Rh8-g8 Kf1-e2 Nc5xe4 d3xe4 Rg4xe4+ Ke2-d2 Re4xe1 Kd2xe1 Kc7-d6
 14/27    00:00     2,132k    9,311k    +1.11    Rh8-g8 Bg2-d5 Rg4-g1+ Kf1-e2 Qg7-g4+ Bd5-f3 Rg1xe1+ Ke2xe1 Qg4-e6 Ke1-e2 Rg8-c8 Qc6-d5 Rc8-d8 Qd5-c6 Qe6xd6 Ra3-a1 e5-e4 Bf3xe4 Nc5xe4 Qc6xe4
 15/27    00:00     2,916k    9,316k    +1.39    Rh8-g8 Bg2-d5 Rg4-g1+ Kf1-e2 Qg7-g4+ Bd5-f3 Rg1xe1+ Ke2xe1 Qg4-g1+ Ke1-e2 e5-e4 Bf3xe4 Qg1-b1 d6-d7 Qb1-c2+ Ke2-f3 Rg8-f8+ Kf3-g4 Qc2-d1+ Kg4-h4 Rf8-h8+ Kh4-g5
 16/25    00:00     3,962k    9,279k    +0.97    e5-e4 Qc6-c7+ Qg7xc7 d6xc7 Ka7-b7 Bg2xe4+ Kb7xc7 Ra3-a2 Kc7-b8 Kf1-e2 Rh8-h2 Ke2-f1 Rg4-f4 Ra2-e2 Nc5xa4 Be4-g2 Na4-c5 Re2-e8+ Kb8-c7 Re1-e7+ Kc7-d6
 17/26    00:00     5,461k    9,178k    +1.06    e5-e4 Qc6-c7+ Qg7xc7 d6xc7 Ka7-b7 Ra3-a2 Kb7xc7 Bg2xe4 Kc7-b8 Ra2-a3 Rg4-f4 Kf1-g2 Kb8-c7 Ra3-a2 Kc7-d6 Ra2-c2 Rh8-g8+ Kg2-f1 Nc5xa4 Rc2-c6+ Kd6-d7 Re1-c1
 18/39    00:00     8,650k    9,173k    +0.98    e5-e4 Qc6-c7+ Qg7xc7 d6xc7 Ka7-b7 Bg2xe4+ Kb7xc7 Ra3-a2 Kc7-b8 Ra2-a3 Rg4-f4 Ra3-a2 Rf4-h4 Re1-e2 Rh4-g4 Be4-f3 Rg4-g7 Ra2-a3 Kb8-c7 Re2-c2 Rh8-g8 Rc2-c4 Rg7-g1+ Kf1-e2 Rg8-e8+ Ke2-d2


Bryan

unread,
Jun 21, 2021, 1:52:36 PMJun 21
to FishCooking
In Game 9, SFdev as white blundered at move 43 by playing Rc8-c1 instead of Kc5 or Kb5. The reason Rc8-c1 is a mistake is that it allows black's king to beat the white king to the c3 square to defend the black pawn on c2, because Rc8-c1 puts the rook on a square where black can gain a tempo with d3-d2 (at move 45 in the game), hitting the white rook on c1. After 43.Rc8-c1 Kd5 44.Kb5  Kd4 45.Kxa4 d3-d2, if white plays Rc1-a1 then the black king comes to b2, white's rook goes to d1 and captures on d2, but black's f-pawn queens. And after 43.Rc8-c1 Kd5 44.Kb5  Kd4 45.Kxa4 d3-d2 if white plays or Rc1-b1 or Rc1-d1, then black's d-pawn queens.

White is already lost at move 43, black to move.

Move 43: black to move: hash cleared: multipv=1: contempt=0: threads=15
FEN: 8/6p1/1K2kp2/4n2p/p4p1P/3p4/P7/2R2B2 b - - 3 43
Stockfish_21062023_x64_avx2:
Found 145 tablebases

NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 34/52    01:49     1,627,530k    14,875k    +6.62    Ke6-d5 Kb6-a5 Kd5-d4 Rc1-c8 g7-g5 Ka5xa4 d3-d2 Rc8-d8+ Kd4-c3 h4xg5 f6xg5 Ka4-a3 g5-g4 Rd8-c8+ Kc3-d4 Ka3-b2 Kd4-e3 Kb2-c2 Ke3-f2 Bf1-a6 g4-g3 Rc8-e8 g3-g2 Re8xe5 d2-d1Q+ Kc2xd1 g2-g1Q+ Kd1-c2 h5-h4 Re5-e2+ Kf2-f3 Ba6-d3 h4-h3 Re2-e6 Kf3-f2 Kc2-b3 h3-h2 Bd3-e4 f4-f3 Re6-h6 Qg1-d1+ Kb3-c4 Kf2-e3

At move 43, white had two moves that (likely) would have saved the draw: Kc5 and Kb5.

Move 43: white to move: multipv=2: contempt=0: threads=15
FEN: 2R5/6p1/1K2kp2/4n2p/p4p1P/3p4/P7/5B2 w - - 2 43
Stockfish_21062023_x64_avx2:
[...]
---------------------------------------------------------------------------
 50/49    07:20     4,886,213k    11,087k     0.00    Kb6-b5 f4-f3 Rc8-c1 g7-g5 h4xg5 f6xg5 Kb5xa4 g5-g4 Ka4-b4 h5-h4 a2-a4 h4-h3 Bf1xh3 g4xh3 a4-a5 d3-d2 Rc1-f1 Ne5-g4 a5-a6 Ng4-e3 a6-a7 Ne3xf1 a7-a8Q d2-d1Q Qa8-e4+ Ke6-d6 Qe4-f4+ Kd6-c6 Qf4-c4+ Kc6-d7 Qc4-g4+ Kd7-e7 Qg4-h4+ Ke7-f7 Qh4-f4+ Kf7-e8 Qf4-e4+ Ke8-d8 Qe4-h4+ Kd8-c7 Qh4-c4+ Kc7-b8 Qc4-f4+ Kb8-b7 Qf4-f7+ Kb7-c8 Qf7-c4+ Kc8-d8
 50/36    07:20     4,886,213k    11,087k     0.00    Kb6-c5 Ke6-f5 Rc8-a8 a4-a3 Ra8xa3 d3-d2 Bf1-d3+ Kf5-g4 Bd3-e2+ Kg4xh4 Ra3-a4 g7-g5 Ra4-d4 g5-g4 Rd4xd2 g4-g3 Rd2-d1 f4-f3 Be2xf3 Ne5xf3 a2-a4 g3-g2 a4-a5 Nf3-d2 Rd1-g1 Nd2-b3+ Kc5-d5 Nb3xa5 Rg1xg2 Na5-b3 Kd5-e4 Nb3-c5+ Ke4-f5 Kh4-h3 Kf5xf6

SFdev abandons Rc8-c1 and finds Kb6-c5 at d=16.

Move 43: white to move: hash cleared: multipv=1: contempt=0: threads=15
FEN: 2R5/6p1/1K2kp2/4n2p/p4p1P/3p4/P7/5B2 w - - 2 43

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 14/23    00:00     3,226k    11,011k     0.00    Rc8-c1 Ke6-d5 Kb6-b5 a4-a3 Kb5-a4 Kd5-d4 Ka4xa3 Ne5-c4+ Ka3-b3 Nc4-d2+ Kb3-b2 g7-g5 a2-a4 Nd2-c4+ Rc1xc4+ Kd4xc4
 15/23    00:00     4,338k    10,954k    -0.02    Rc8-c1 Ke6-f5 Kb6-b5 a4-a3 Kb5-c5 f4-f3 Kc5-d4 Kf5-f4 Bf1xd3 g7-g5 h4xg5 f6xg5 Bd3-b5 f3-f2 Rc1-f1 Ne5-f3+ Kd4-d3 Kf4-g3
 16/19    00:00     5,332k    11,038k    -0.75    Kb6-c5 Ke6-f5 Rc8-a8 f4-f3 Ra8xa4 g7-g5 h4xg5 f6xg5 Kc5-d5 g5-g4 Ra4-e4 Ne5-g6 Re4-e1 Ng6-f4+ Kd5-d4 g4-g3 Bf1xd3+ Nf4xd3 Kd4xd3
 17/24    00:00     7,242k    11,176k    -0.84    Kb6-c5 Ke6-f5 Rc8-a8 f4-f3 Ra8xa4 g7-g5 h4xg5 f6xg5 Kc5-d6 g5-g4 Ra4-a5 g4-g3 Ra5xe5+ Kf5-f4 Bf1xd3 g3-g2 Re5-e4+ Kf4-g3 Re4-e8 h5-h4 Re8-g8+ Kg3-f4
 18/33    00:00     9,935k    11,226k    -0.91    Kb6-c5 Ke6-f5 Rc8-a8 f4-f3 Ra8xa4 g7-g5 Ra4-b4 g5-g4 a2-a4 g4-g3 Rb4-b2 Kf5-e4 a4-a5 Ke4-e3 a5-a6 d3-d2 Rb2-b1 g3-g2 a6-a7 g2-g1Q
 19/35    00:00     10,797k    11,165k    -0.19    Kb6-c5 Ke6-f5 Rc8-a8 f4-f3 Ra8xa4 g7-g5 Ra4-b4 g5-g4 a2-a4 g4-g3 Rb4-b2 Kf5-f4 a4-a5 Kf4-e3 a5-a6 d3-d2 Rb2-b1 g3-g2 Bf1xg2 f3xg2 a6-a7
 20/39    00:01     14,612k    11,053k     0.00    Kb6-c5 g7-g5 h4xg5 f6xg5 Rc8-e8+ Ke6-f6 Re8-f8+ Kf6-e6
 21/35    00:01     17,391k    10,952k     0.00    Kb6-c5 f4-f3 Bf1-h3+ Ke6-e7 Rc8-c7+ Ke7-e8 Rc7-c8+ Ke8-e7
 22/8    00:01     17,543k    10,944k     0.00    Kb6-c5 f4-f3 Bf1-h3+ Ke6-e7 Rc8-c7+ Ke7-e8 Rc7-c8+
 23/27    00:01     18,797k    10,922k     0.00    Kb6-c5 f4-f3 Bf1-h3+ Ke6-e7 Rc8-c7+ Ke7-e8 Rc7-c8+
 24/8    00:01     21,261k    10,898k     0.00    Kb6-c5 f4-f3 Bf1-h3+ Ke6-e7 Rc8-c7+ Ke7-e8 Rc7-c8+
 25/12    00:02     23,552k    10,884k     0.00    Kb6-c5 f4-f3 Bf1-h3+ Ke6-e7 Rc8-c7+ Ke7-e8 Rc7-c8+
 26/8    00:02     25,379k    10,846k     0.00    Kb6-c5 f4-f3 Bf1-h3+ Ke6-e7 Rc8-c7+ Ke7-e8 Rc7-c8+
 27/33    00:02     29,699k    10,831k     0.00    Kb6-c5 f4-f3 Bf1-h3+ Ne5-g4 Kc5-c4 d3-d2 Rc8-d8 Ke6-e5 Rd8xd2 Ke5-f4 Rd2-d4+ Kf4-g3 Bh3xg4 h5xg4 Kc4-d3 Kg3xh4 Rd4-f4 g7-g5 Rf4xf6 Kh4-h3 Kd3-e3 Kh3-g2 a2-a3 g4-g3 Rf6xf3 Kg2-h2 Rf3-f5 g3-g2 Rf5xg5 g2-g1Q+ Rg5xg1

At iteration 14, the mistake is thinking that black will play 44...a4-a3, when black has the much better move 44...Kd5-d4. If white responds to 44...Kd5-d4 with 45.Kb4, then black plays d3-d2, and white cannot stop either the d-pawn or the g-pawn from queening.

Bryan

unread,
Jun 21, 2021, 4:08:46 PMJun 21
to FishCooking
In Game 10, SFdev as black blundered at move 32 by playing Ne5xf3. The better move (and only game-saving move) there is Bf8-h6.

At move 33 white is already winning:

Move 33: white to move: hash cleared: multipv=1: contempt=0: threads=15
FEN: 2kr1brq/1p6/p3bp2/2P5/1P1N1Q2/R4nP1/P1P5/2B1R1K1 w - - 0 33

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 50/87    08:10     5,707,066k    11,642k    +4.75    Qf4xf3 Be6-d5 Qf3-f5+ Kc8-b8 Bc1-f4+ Bf8-d6 c5xd6 Qh8-h1+ Kg1-f2 Qh1-h2+ Kf2-e3 Rd8-e8+ Ke3-d3 Rg8xg3+ Re1-e3 Rg3xe3+ Bf4xe3 Bd5-e4+ Qf5xe4 Re8xe4 Kd3xe4 Qh2-e5+ Ke4-d3 Qe5xd6 Kd3-c3 f6-f5 Ra3-b3 f5-f4 Be3-f2 Qd6-d5 Kc3-b2 Qd5-g2 Rb3-f3 Qg2-g5 a2-a3 Qg5-e5 Rf3-c3 Qe5-e8 Rc3-c5 Qe8-e4 Rc5-f5 Kb8-c8 Rf5-f8+ Kc8-d7 Rf8-f7+ Kd7-e8 Rf7-f6 Ke8-d7 Nd4-b3 Kd7-e7 Bf2-d4 f4-f3 Bd4-c3 Qe4-e3 Rf6-f5 Ke7-d7 Nb3-c5+ Kd7-c7 Nc5-d3 Qe3-e4 Rf5-f4 Qe4-d5 a3-a4 b7-b6 Rf4-f6 Kc7-c8 Rf6xb6 Qd5-c4

As best as I can tell, the reason white is winning here is because by move 45 white has a material advantage (a rook and two pieces for a queen), only one pawn island compared to black's two pawn islands, no isolated pawns (compared to one isolated pawn for black), a more developed king protected within the white pawns.

At move 32, black to move, black could have kept the position roughly equal with Bf8-h6.

Move 32: black to move: hash from subsequent moves: multipv=1: contempt=0: threads=15
FEN: 2kr1brq/1p6/p3bp2/2P1n3/1P1N1Q2/R4BP1/P1P5/2B1R1K1 b - - 0 32

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 48/69    11:38     6,172,028k    8,834k    -0.30    Bf8-h6 Qf4xh6 Rg8xg3+ Bf3-g2 Qh8xh6 Bc1xh6 Rg3xg2+ Kg1xg2 Rd8xd4 Ra3-g3 Rd4xb4 Re1-f1 Ne5-d7 Bh6-f4 Rb4-c4 Kg2-g1 Rc4xc5 Rg3-e3 Nd7-e5 Bf4-g3 Rc5xc2 Rf1xf6 Ne5-g4 Rf6-f8+ Kc8-d7 Re3-d3+ Kd7-e7 Rf8-f1 Be6-c4 Rf1-e1+ Ke7-f6 Rd3-d4 Ng4-h6 a2-a3 Nh6-f5 Bg3-e5+ Kf6-g5 Rd4-f4 Bc4-d5 Rf4-f2 Rc2-c4 Rf2-d2 Bd5-c6 Kg1-f2 Kg5-g6 Re1-g1+ Kg6-f7 Rd2-e2 Kf7-e6 Rg1-g6+ Ke6-d5 Rg6-g5 Kd5-e6 Be5-b2+ Rc4-e4 Rg5-g4 Re4xe2+ Kf2xe2 Bc6-b5+ Ke2-d2 Ke6-d6 Rg4-g6+ Kd6-c5 Rg6-g5 Bb5-d7 Kd2-c1 Bd7-e6 Rg5-g6

In that line, by move 38 (after 32...Bh6 33.Qxh6 Rxg3 34.Bg2 Qxh6 35.Bxh6 Rxg2 36.Kxg2 Rxd4 37.Rg3 Rxb4) materially white is up a rook for a knight, the bishops are opposite colored, white has an isolated pawn on the a-file and doubled pawns on the c-file, and white is not on the same flank as the white pawns.

SFdev abandons Ne5xf3+ and finds Bf8-h6 at d=27:

Move 32: black to move: hash cleared: multipv=1: contempt=0: threads=15
FEN: 2kr1brq/1p6/p3bp2/2P1n3/1P1N1Q2/R4BP1/P1P5/2B1R1K1 b - - 0 32

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 24/42    00:03     30,994k    9,333k    -0.12    Ne5xf3+ Qf4xf3 Be6-d5 Qf3-f5+ Kc8-b8 Bc1-f4+ Bf8-d6 Nd4-f3 Bd6xf4 Qf5xf4+ Kb8-a8 Re1-d1 Rg8xg3+ Qf4xg3 Rd8-g8 Qg3xg8+ Qh8xg8+ Kg1-f2 Bd5xf3 Ra3xf3 Qg8-h7 Rd1-d8+ Ka8-a7 c2-c3 Qh7-c2+ Kf2-f1 Qc2-b1+ Kf1-e2 Qb1xa2+ Rd8-d2 Qa2-c4+ Rd2-d3 a6-a5 b4xa5 Qc4-e6+ Ke2-f1 Qe6-a6 Kf1-g1 Qa6-b5 Rd3-d4 Qb5xc5 Kg1-f2
 25/45    00:04     38,901k    9,302k    -0.06    Ne5xf3+ Qf4xf3 Be6-d5 Qf3-f5+ Kc8-b8 Bc1-f4+ Bf8-d6 Nd4-f3 Bd6xf4 Qf5xf4+ Kb8-a8 Re1-d1 Rg8xg3+ Qf4xg3 Rd8-g8 Qg3xg8+ Qh8xg8+ Kg1-f2 Bd5-c6 Rd1-d6 f6-f5 Nf3-d4 Bc6-d5 Ra3-d3 Bd5-c4 Nd4-c6 Bc4xd3 Rd6-d8+ Qg8xd8 Nc6xd8 Bd3xc2 Kf2-e3 Bc2-b1 a2-a4 Bb1-a2 a4-a5 Ka8-b8 c5-c6 b7xc6 Nd8xc6+ Kb8-a8 Ke3-d2 Ba2-b1
 26/47-    00:06     60,630k    9,255k    -0.14    Ne5xf3+ Qf4xf3
 26/47+    00:06     62,492k    9,232k    -0.06    Ne5xf3+
 26/47    00:06     63,315k    9,224k    -0.12    Ne5xf3+ Qf4xf3 Be6-d5 Qf3-f5+ Kc8-b8 Bc1-f4+ Bf8-d6 Nd4-f3 Bd6xf4 Qf5xf4+ Kb8-a8 Kg1-f2 f6-f5 Re1-d1 Bd5-e4 Rd1-d6 Qh8-b2 Ra3-b3 Qb2xa2 Qf4-d2 Rd8-c8 Qd2-c3 f5-f4 g3xf4 Rc8-f8 Qc3-b2 Qa2-a4 Rb3-a3 Qa4xc2+ Qb2xc2 Be4xc2 Rd6-d4 Rg8-g4 b4-b5 Rg4xf4 Rd4xf4 Rf8xf4
 27/52-    00:10     99,263k    9,140k    -0.20    Ne5xf3+ Qf4xf3
 27/52-    00:10     100,107k    9,136k    -0.28    Ne5xf3+ Qf4xf3
 27/52-    00:12     115,591k    9,161k    -0.59    Ne5xf3+ Qf4xf3
 27/52    00:17     161,039k    9,027k    -0.49    Bf8-h6 Qf4xh6 Rg8xg3+ Bf3-g2 Qh8xh6 Bc1xh6 Rg3xg2+ Kg1xg2 Rd8xd4 Bh6-g7 Be6-d5+ Kg2-f2 Kc8-d7 Bg7-f8 Kd7-c6 Bf8-d6 Rd4-f4+ Kf2-e2 Rf4xb4 Bd6xe5 f6xe5 Ke2-d2 e5-e4 Re1-f1 Rb4-c4 Rf1-f6+ Kc6xc5 Rf6-f5 Kc5-d6 Rf5-h5 Rc4-c5 Rh5-h6+ Kd6-e5 Ra3-h3 Bd5xa2 c2-c3 Rc5-b5 Rh3-h5+ Ke5-f4
 28/41-    00:18     168,855k    9,013k    -0.57    Bf8-h6 Qf4xh6
 28/41-    00:19     176,237k    9,014k    -0.65    Bf8-h6 Qf4xh6
 28/41+    00:22     200,720k    9,005k    -0.57    Bf8-h6
 28/44    00:22     204,869k    9,001k    -0.67    Bf8-h6 Qf4xh6 Rg8xg3+ Bf3-g2 Rg3xg2+ Kg1xg2 Qh8xh6 Bc1xh6 Rd8xd4 Ra3-g3 Rd4xb4 Re1-f1 Rb4-c4 Rf1xf6 Rc4xc2+ Kg2-g1 Be6xa2 Rg3-g7 Ne5-f7 Bh6-e3 a6-a5 Rf6xf7 Ba2xf7 Rg7xf7 a5-a4 Rf7-f8+ Kc8-d7 Rf8-a8 Rc2-e2 Be3-f2 Re2-e4 Kg1-g2 Kd7-c6 Kg2-f3 Re4-b4 Ra8-a7 Kc6-d5 Ra7-a5 Rb4-c4 Bf2-e3 Rc4-b4
 29/46+    00:23     208,836k    8,998k    -0.59    Bf8-h6
 29/46    00:24     220,544k    8,994k    -0.52    Bf8-h6 Qf4xh6 Rg8xg3+ Bf3-g2 Qh8xh6 Bc1xh6 Rg3xg2+ Kg1xg2 Rd8xd4 Ra3-g3 Rd4xb4 Re1-f1 Rb4-c4 Rf1xf6 Rc4xc2+ Kg2-g1 Be6xa2 Rg3-g7 Ne5-f7 Bh6-e3 a6-a5 Rg7xf7 Ba2xf7 Rf6xf7 a5-a4 Rf7-f8+ Kc8-d7 Rf8-a8 Rc2-e2 Be3-f4 Kd7-c6 Kg1-f1 Re2-e4 Bf4-d6 Kc6-b5 Kf1-f2 Kb5-c6 Ra8-a7 Re4-b4 Kf2-f3 b7-b6

The weakness can be seen at iteration=26 where SFdev thinks white will play 36.Nd4-f3. We can narrow in on that weakness by looking at move 36, white to move:

Move 36: white to move: hash cleared: multipv=1: contempt=0: threads=15
FEN: 1k1r2rq/1p6/p2b1p2/2Pb1Q2/1P1N1B2/R5P1/P1P5/4R1K1 w - - 5 36

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
 [...]
 10/14    00:00     179k    8,960k    +0.01    Nd4-f3 Bd6xf4 Qf5xf4+ Kb8-a8 b4-b5 Qh8-h3 Kg1-f2 Rg8-g4
 11/22    00:00     363k    8,845k    +0.11    Nd4-f3 Bd6xf4 Qf5xf4+ Kb8-a7 b4-b5 Qh8-h3 Kg1-f2 Bd5xf3 b5-b6+ Ka7-a8
 12/19    00:00     486k    9,004k    +0.11    Nd4-f3 Bd6xf4 Qf5xf4+ Kb8-a7 b4-b5 Qh8-h3 Kg1-f2 Bd5xf3 b5-b6+ Ka7-a8 Qf4xf3 Rd8-d2+ Re1-e2 Qh3-h2+ Kf2-e1 Rd2xe2+ Qf3xe2 Qh2-g1+ Ke1-d2 Qg1xc5
 13/25    00:00     683k    8,874k    +0.16    Nd4-f3 Rg8xg3+ Bf4xg3 Bd6xg3 Re1-d1 Qh8-g8 Rd1-d4 Bg3-e5+ Rd4-g4 Bd5-e6 Qf5xe5+ f6xe5 Rg4xg8 Rd8xg8+ Kg1-f2 Rg8-f8 Ra3-e3 Rf8-f4 Re3xe5 Be6xa2 c2-c3 Rf4-c4 Re5-e8+ Kb8-c7
 14/32    00:00     1,709k    9,044k    +0.44    c5xd6 Qh8-h1+ Kg1-f2 Qh1-g2+ Kf2-e3 Rd8-e8+ Nd4-e6
 15/31    00:00     2,470k    9,216k    +0.17    Nd4-f3 Bd6xf4 Qf5xf4+ Kb8-a8 Kg1-f2 Bd5-c6 Ra3-d3 Rd8xd3 c2xd3 Qh8-h3 Re1-e7 Rg8-g4 Qf4-d6 Rg4xb4 Qd6xf6 Rb4-a4 Qf6-d6 Ra4xa2+ Kf2-e3
 16/34    00:00     4,852k    9,403k    +0.90    c5xd6 Qh8-h1+ Kg1-f2 Qh1-g2+ Kf2-e3 Rd8-e8+ Ke3-d3 Rg8xg3+ Nd4-f3 Rg3xf3+ Kd3-d4 Re8xe1 d6-d7+ Kb8-a7 Bf4-b8+ Ka7xb8 d7-d8Q+ Kb8-a7 Ra3xf3 Qg2-d2+ Rf3-d3
 17/26    00:00     6,459k    9,714k    +2.53    c5xd6 Qh8-h1+ Kg1-f2 Qh1-g2+ Kf2-e3 Rg8-e8+ Ke3-d3 Re8xe1 Kd3-c3 Qg2-e4 Qf5xf6 Rd8-d7 Qf6-h8+ Qe4-e8 Qh8xe8+ Re1xe8 Ra3-a5 Bd5-e4

Tracking that down, we can go to move 36, black to move, and search it with one thread:

Move 36: black to move: hash cleared: multipv=1: contempt=0: threads=1
FEN: 1k1r2rq/1p6/p2P1p2/3b1Q2/1P1N1B2/R5P1/P1P5/4R1K1 b - - 0 36

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 12/18    00:00     61k    989k     0.00    Qh8-h1+ Kg1-f2 Qh1-g2+ Kf2-e3 Rd8-e8+ Nd4-e6 Re8xe6+ Ke3-d4 Re6xe1 d6-d7+ Re1-e5 Bf4xe5+ f6xe5+ Qf5xe5+ Kb8-a7
 13/22    00:00     72k    985k     0.00    Qh8-h1+ Kg1-f2 Qh1-g2+ Kf2-e3 Rd8-e8+ Nd4-e6 Re8xe6+ Ke3-d4 Re6xe1 d6-d7+ Re1-e5 Bf4xe5+ f6xe5+ Qf5xe5+ Kb8-a7 Qe5xd5 Rg8-g4+ Kd4-c5 Qg2-f2+ Kc5-d6 Qf2-f8+ Kd6-e6
 14/28    00:00     121k    995k     0.00    Qh8-h1+ Kg1-f2 Qh1-g2+ Kf2-e3 Rd8-e8+ Nd4-e6 Re8xe6+ Ke3-d4 Re6xe1 d6-d7+ Kb8-a7 Qf5xd5 Qg2xc2 Qd5xg8 Re1-d1+ Kd4-e3 Rd1-e1+
 15/18    00:00     138k    997k    +0.30    Qh8-h1+ Kg1-f2 Qh1-g2+ Kf2-e3 Rd8-e8+ Nd4-e6 Re8xe6+ Ke3-d4 Re6xe1 d6-d7+ Kb8-a7 Qf5xd5 Qg2xc2 Qd5-c5+ Qc2xc5+ Kd4xc5
 16/32    00:00     767k    1,065k    -1.37    Qh8-h1+ Kg1-f2 Qh1-g2+ Kf2-e3 Rd8-e8+ Ke3-d3 Rg8xg3+ Nd4-f3 Rg3xf3+ Kd3-d4 Re8xe1 d6-d7+ Rf3xf4+ Qf5xf4+ Kb8-a7 Kd4-c3 Re1-e4 Qf4-c7 Re4-c4+ Qc7xc4 Bd5xc4 d7-d8Q Qg2-g3+ Kc3-b2 Qg3-e5+ Ra3-c3 Bc4xa2 Qd8-a5 Qe5xa5 b4xa5 Ba2-d5 Rc3-c8
 17/44    00:02     2,458k    1,122k    -2.76    Qh8-h1+ Kg1-f2 Qh1-h2+ Kf2-e3 Rd8-e8+ Ke3-d3 Rg8xg3+ Re1-e3 Rg3xe3+ Bf4xe3 Bd5-e4+ Qf5xe4 Re8xe4 Kd3xe4 Qh2-e5+ Ke4-d3 Qe5xd6 c2-c3 b7-b6 Ra3xa6 Kb8-b7 Ra6-a3 Qd6-d5 Kd3-d2
 18/35    00:03     3,403k    1,122k    -3.23    Qh8-h1+ Kg1-f2

There the mistake in iterations 12, 13, and 14 is Nd4-e6 (marked in bold font), which gives up that white knight, and also loses the rook on e1. White has a much better move here: Ke3-d3. This leads to white going up two pieces and two pawns, for a rook (after 39.Kd3 Rxe1 40.d7 Ka8 41.Kc3 Re7 42.Nb5 Qe4 43.Qxd5 Qxd5 44.Nc7+ Ka7 45.Nxd5 Rxd7 46.Nxf6 Rc8+ 47.Kb2).

Bryan

unread,
Jun 21, 2021, 5:45:27 PMJun 21
to FishCooking
In Game 11, SFdev as white blundered at move 41, playing Qc4-c2. The only game-saving move there is Qc4-a2, which threatens Qa2-a8#.

At move 41, black to move, white is already lost.

Move 41: black to move: hash cleared: multipv=1: contempt=0: threads=15
FEN: 1kn5/3q1p2/3p1b2/1p5r/1R2B2r/3N2pP/2Q3P1/4R2K b - - 9 41

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 38/70    07:14     4,209,645k    9,683k    +4.80    Bf6-d8 Kh1-g1 Bd8-b6+ Kg1-f1 Bb6-a5 Be4-f3 Nc8-b6 Bf3xh5 Rh4xh5 Rb4-e4 Nb6-d5 Re4-e8+ Kb8-b7 Re1-e3 Nd5xe3+ Re8xe3 Rh5-f5+ Kf1-e2 Ba5-b6 Re3-e4 Bb6-c5 Re4-f4 Kb7-b6 Qc2-b3 Qd7-e7+ Ke2-d1 Qe7-f6 Rf4xf5 Qf6xf5 Nd3-e1 Qf5-f2 Qb3-c3 b5-b4 Qc3-c4 Qf2-b2 h3-h4 f7-f5 Qc4-g8 Qb2-d4+ Kd1-c1 Qd4-c3+ Ne1-c2 b4-b3 Qg8-b8+ Kb6-a5 Qb8-a8+ Ka5-b5 Qa8-b7+ Bc5-b6 Qb7-d5+ Kb5-a6 Qd5xf5 Bb6-e3+ Kc1-b1 b3xc2+ Qf5xc2 Qc3-e1+ Kb1-a2 Qe1-a5+ Ka2-b3 Qa5-b5+ Kb3-a2 Qb5-d5+ Ka2-b1 Ka6-b6 Qc2-c3 Qd5-d1+ Kb1-a2 Qd1-e2+ Qc3-b2+ Qe2xb2+ Ka2xb2

At move 41, white to move, the move Qc4-a2 holds the draw for white:

Move 41: white to move: hash cleared: multipv=1: contempt=0: threads=15
FEN: 1kn5/3q1p2/3p1b2/1p5r/1RQ1B2r/3N2pP/6P1/4R2K w - - 8 41

Stockfish_21062023_x64_avx2:
NNUE evaluation using nn-75980ca503c6.nnue enabled
[...]
 57/46    20:02     10,090,262k    8,391k     0.00    Qc4-a2 Nc8-b6 Re1-f1 Rh4xh3+ g2xh3 Qd7xh3+ Kh1-g1 g3-g2 Be4xg2 Qh3-h2+ Kg1-f2 Rh5-f5+ Kf2-e2 Qh2xg2+ Rf1-f2 Rf5xf2+ Nd3xf2 Qg2-c6 Qa2-b3 Kb8-c7 Rb4xb5 Nb6-c4 Rb5-f5 Qc6-e8+ Ke2-d3 Nc4-e5+ Kd3-d2 Bf6-g7 Rf5-h5 Qe8-c6 Rh5xe5 Bg7xe5 Qb3xf7+ Kc7-d8 Qf7-g8+ Kd8-d7 Qg8-h7+ Kd7-c8 Qh7-g6 Be5-f4+ Kd2-e2 Qc6-c4+ Qg6-d3 Qc4xd3+ Nf2xd3

The move 41...Qc4-c2 allows the black bishop to reach a5 and skewer the white rooks. However, if black plays 41...Qc4-a2, then the mate threat of Qa2-a8# requires black to use tempi to prevent it, and also disallows black's bishop from maneuvering from f6 to d8 to a5. The black queen guards the c6 square, but not the a6 square. So Qc4-a2 is more aggressive (than Qc4-c2), and gives white more tempi by allowing the white queen to move up the a-file and harass the rather unprotected black king.





Fauzi

unread,
Jun 21, 2021, 6:03:40 PMJun 21
to FishCooking
What is the time control for these games?

Bryan

unread,
Jun 21, 2021, 6:09:13 PMJun 21
to FishCooking
@Fauzi, all around tc=24.5.
Message has been deleted

Bryan

unread,
Jun 21, 2021, 7:08:48 PMJun 21
to FishCooking
In Game 12, SF_20210619 played two suboptimal moves: at move 14 and at move 16.

At move 14, black to move, SF_20210619 played 14...b7-b5. This lost about 120 centipawns. The better move here is e6-e5, which forces Nf5 Bxf5 gxf5, and this removes white's g-pawn threat to black's already advanced h-pawn. (Allowing white to play 15.g4-g5 significantly weakens black's king defense.)

Move 14: black to move: hash cleared: multipv=1: (contempt removed): threads=15
FEN: r1b2rk1/1pq1bpp1/p2ppn1p/8/3NPPP1/1PN1BQ2/1PP4P/3R1RK1 b - g3 0 14
Stockfish_21062123_x64_avx2:
NNUE evaluation using nn-190f102a22c3.nnue enabled
[...]
 44/66    12:42     6,525,196k    8,561k    -0.11    e6-e5 Nd4-f5 Bc8xf5 g4xf5 b7-b5 Rf1-f2 b5-b4 f4xe5 b4xc3 e5xf6 Be7xf6 b2xc3 Qc7xc3 Be3-f4 Rf8-e8 Qf3xc3 Bf6xc3 Rd1-d3 Bc3-b4 Rf2-e2 a6-a5 Bf4xd6 Bb4xd6 Rd3xd6 a5-a4 Rd6-d4 a4-a3 Rd4-d1 Re8-c8 c2-c4 Rc8-b8 Rd1-b1 Ra8-a4 Re2-c2 a3-a2 Rb1-a1 Rb8xb3 Ra1xa2 Ra4xa2 Rc2xa2 Rb3-c3 Ra2-a8+ Kg8-h7 Ra8-c8 Rc3-e3 Kg1-f2 Re3xe4 Kf2-f3 Re4-h4 Kf3-g3 Rh4-d4 Rc8-c7 f7-f6 c4-c5 h6-h5 c5-c6 Rd4-c4 Kg3-f3 Rc4-c2 Kf3-e3 h5-h4

SFdev (SF_21062123) abandons b7-b5 at d=15 and finds e6-e5 at d=21.

Move 14: black to move: hash cleared: multipv=1: threads=15
FEN: r1b2rk1/1pq1bpp1/p2ppn1p/8/3NPPP1/1PN1BQ2/1PP4P/3R1RK1 b - g3 0 14
Stockfish_21062123_x64_avx2:
NNUE evaluation using nn-190f102a22c3.nnue enabled
[...]
 12/20    00:00     736k    8,759k    -0.12    b7-b5 b3-b4 Bc8-b7 h2-h4 h6-h5 g4xh5 a6-a5 h5-h6 g7-g6 Qf3-g3 Nf6-h5 Qg3-g2 a5xb4
 13/21    00:00     1,407k    8,687k    -0.19    b7-b5 b3-b4 e6-e5 Nd4-f5 Bc8xf5 e4xf5 e5xf4 Qf3xf4 Rf8-e8 h2-h4 Qc7-c6 g4-g5
 14/24    00:00     2,172k    8,756k    -0.27    b7-b5 g4-g5 h6xg5 f4xg5 Nf6-d7 h2-h4 b5-b4 Nc3-e2 a6-a5 Ne2-f4 Nd7-e5 Qf3-g2 Ne5-g6 Nf4xg6 f7xg6
 15/29    00:00     3,310k    8,803k    -0.52    Nf6-h7 Qf3-g2 b7-b5 b3-b4 a6-a5 Nc3xb5 Qc7-d7 b4xa5 Ra8xa5 c2-c4 Bc8-b7 g4-g5 h6xg5 f4xg5
 16/24    00:00     3,855k    8,782k    -0.27    Nf6-h7 Qf3-g2 b7-b5 g4-g5 h6xg5 f4xg5 b5-b4 Nc3-a4 g7-g6 h2-h4 Ra8-b8 c2-c4 e6-e5 Nd4-f3 Bc8-b7 Nf3-h2 Qc7-c6
 17/26    00:00     6,458k    8,751k    -0.11    Nf6-h7 Qf3-g2 b7-b5 b3-b4 a6-a5 Nc3xb5 Qc7-b7 c2-c4 a5xb4 g4-g5 h6xg5 f4xg5 g7-g6 Nd4-c2 b4-b3 Nc2-d4 Bc8-d7 h2-h4
 18/29    00:01     9,844k    8,720k    -0.36    Nf6-h7 Qf3-g2 b7-b5 h2-h4 b5-b4 Nc3-a4 Bc8-b7 g4-g5 h6xg5 f4xg5 Ra8-c8 c2-c4 b4xc3/ep Na4xc3 g7-g6 Rf1-f2 e6-e5 Nd4-f3
 19/28    00:01     12,693k    8,682k    -0.29    Nf6-h7 h2-h4 b7-b5 g4-g5 b5-b4 Nc3-e2 g7-g6 g5xh6 Be7xh4 f4-f5 e6-e5 f5xg6 f7xg6 Qf3-g2 Bh4-f6 Nd4-f5 Kg8-h8 Nf5xd6 Bc8-e6 c2-c4 b4xc3/ep Ne2xc3
 20/30    00:02     22,060k    8,594k    -0.49    Nf6-h7 h2-h4 b7-b5 b3-b4 Be7xh4 Qf3-h1 Bh4-f6 g4-g5 Bf6xd4 Rd1xd4 f7-f5 Rd4-d2 Bc8-b7 Qh1-h3 Kg8-h8 Rd2-g2 Ra8-e8 g5xh6 g7xh6 Qh3xh6 Rf8-g8 Be3-d4+ e6-e5
 21/29    00:03     27,678k    8,556k    -0.45    e6-e5 Nd4-f5 Bc8xf5 e4xf5 e5xf4 Qf3xf4 Ra8-e8 Be3-d4 Nf6-h7 Qf4-f2 Be7-f6 Nc3-d5 Qc7-c6 Qf2-g2 Bf6xd4+ Rd1xd4 Re8-e5 c2-c3 Rf8-e8 f5-f6 Nh7xf6 Nd5xf6+ g7xf6 Qg2xc6 b7xc6 Rd4xd6 Re5-g5 Rf1-f4 Re8-e2 Rd6xf6

At move 16, black to move, SF_20210619 played Nf6-d7. This lost about 430 centipawns. The only move here is Nf6-h7. Again, the problem here is black's king protection, and Nf6-h7 allows black to target white's pawn on g5, and add king-protection to the h-file.

Move 16: black to move: hash cleared: multipv=1: (contempt removed): threads=15
FEN: r1b2rk1/2q1bpp1/p2ppn2/1p4P1/3NP3/1PN1BQ2/1PP4P/3R1RK1 b - - 0 16
Stockfish_21062123_x64_avx2:
NNUE evaluation using nn-190f102a22c3.nnue enabled
[...]
 43/78    03:28     1,877,506k    8,986k    -1.30    Nf6-h7 Qf3-h5 b5-b4 Rf1xf7 Nh7xg5 Rf7xf8+ Be7xf8 Be3xg5 b4xc3 Rd1-f1 Qc7-d7 b2xc3 Bc8-b7 Rf1-f4 Qd7-e8 Qh5-h3 Qe8-g6 Qh3xe6+ Qg6xe6 Nd4xe6 Ra8-e8 Ne6xf8 Re8xf8 Rf4-g4 Rf8-e8 Bg5-f6 Bb7xe4 Rg4xg7+ Kg8-f8 Rg7-c7 Be4xc2 b3-b4 Re8-e2 Rc7-d7 Re2-e6 Bf6-g5 Bc2-a4 Rd7-d8+ Kf8-f7 Kg1-f2 Ba4-b5 h2-h4 Re6-e2+ Kf2-g3 Re2-c2 Rd8-c8 Bb5-e2 Bg5-e3 Kf7-e7 Be3-d4 Rc2-c1 Kg3-f2 Be2-b5 Kf2-g2 Bb5-f1+ Kg2-f3 Bf1-b5 h4-h5 Rc1-h1 Rc8-h8 Rh1-f1+ Kf3-g3

Move 17: black to move: hash cleared: multipv=1: (contempt removed): threads=15
FEN: r1b2rk1/2qnbpp1/p2pp3/1p4PQ/3NP3/1PN1B3/1PP4P/3R1RK1 b - - 2 17
Stockfish_21062123_x64_avx2:
Found 145 tablebases
NNUE evaluation using nn-190f102a22c3.nnue enabled
[...]
 33/51    01:50     1,065,221k    9,650k    -5.69    Nd7-e5 Rf1-f4 Be7xg5 Qh5xg5 b5-b4 Nc3-e2 f7-f6 Qg5-g2 a6-a5 Rf4-h4 Qc7-f7 Rd1-f1 a5-a4 b3xa4 Ra8xa4 Ne2-f4 g7-g5 Rh4-h5 Ne5-g6 Nf4xg6 Qf7xg6 Be3xg5 Bc8-a6 Rh5-h6 Qg6-g7 Nd4xe6 Ba6xf1 Qg2-g3 Ra4-a5 h2-h4 f6xg5 Ne6xg7 Kg8xg7 Rh6xd6 Bf1-e2 Rd6-d2 Rf8-f7 Rd2xe2 Kg7-g6 Qg3-d3 Ra5-a1+ Kg1-g2 g5xh4 Qd3-d6+ Kg6-g5 Qd6xb4

SFdev (SF_21062123) abandons Nf6-d7 and finds Nf6-h7 at d=15.

Move 16: black to move: hash cleared: multipv=1: threads=15
FEN: r1b2rk1/2q1bpp1/p2ppn2/1p4P1/3NP3/1PN1BQ2/1PP4P/3R1RK1 b - - 0 16
Stockfish_21062123_x64_avx2:
NNUE evaluation using nn-190f102a22c3.nnue enabled
[...]
 13/15    00:00     1,156k    8,757k    -0.34    Nf6-d7 h2-h4 Nd7-e5 Qf3-g2 Bc8-b7 h4-h5 Ra8-e8 g5-g6 f7xg6 Rf1xf8+ Be7xf8
 14/17    00:00     1,261k    8,760k    -0.08    Nf6-d7 h2-h4 Nd7-e5 Qf3-g2 Bc8-b7 h4-h5 Ra8-e8 Nc3-e2 d6-d5 g5-g6 Be7-c5 Qg2-g3 Bc5-b6 Ne2-f4 d5xe4
 15/29    00:00     4,802k    8,977k    -0.89    Nf6-h7 Qf3-h5 b5-b4 Nc3-d5 e6xd5 g5-g6 Nh7-g5 Rf1xf7 Rf8xf7 Be3xg5 Be7xg5 g6xf7+ Qc7xf7 Qh5xg5 Bc8-h3 Qg5xd5 Ra8-f8 Qd5xf7+ Rf8xf7 Nd4-c6 Rf7-f4 Nc6xb4
 16/26    00:00     7,185k    9,084k    -1.37    Nf6-h7 Qf3-h5 b5-b4 Nc3-d5 e6xd5 g5-g6 f7xg6 Qh5xd5+ Kg8-h8 Rf1xf8+ Nh7xf8 Qd5xa8 Qc7-d7 Rd1-f1 Kh8-h7 Qa8-c6 Qd7-g4+ Kg1-h1 Be7-f6 Rf1-g1 Qg4-h4 Qc6-e8 Bf6xd4
 17/37    00:00     8,330k    9,134k    -1.26    Nf6-h7 Qf3-h5 b5-b4 Rf1xf7 Rf8xf7 g5-g6 Nh7-f6 g6xf7+ Kg8-f8 Qh5-h8+ Kf8xf7 Nd4-f3 Bc8-d7 Qh8xa8 b4xc3 Nf3-g5+ Kf7-g6 e4-e5 Nf6-d5 e5xd6 Be7xd6
 18/30    00:01     9,961k    9,122k    -0.92    Nf6-h7 Qf3-h5 b5-b4 Nc3-d5 e6xd5 g5-g6 Nh7-g5 Rf1xf7 Rf8xf7 Be3xg5 Be7xg5 Qh5xg5 Rf7-f6 Qg5-h5 Rf6xg6+ Qh5xg6 Qc7-e7 Nd4-c6 Qe7-h4 Qg6-g3
 19/41    00:01     14,660k    9,197k    -1.50    Nf6-h7 Qf3-h5 b5-b4 Rf1xf7 Rf8xf7 g5-g6 Nh7-f6 g6xf7+ Kg8-f8 Qh5-h8+ Kf8xf7 Nd4-f3 Be7-f8 e4-e5 d6xe5 Nf3-g5+ Kf7-g6 Nc3-e4 Bc8-b7 Ne4xf6 g7xf6 Rd1-f1 Qc7-g7 Qh8-h3 Bb7-c8 Qh3-g4 f6-f5 Qg4-h4 Kg6-f6 Qh4-h5 Kf6-e7

Bryan

unread,
Jun 21, 2021, 9:38:03 PMJun 21
to FishCooking
In Game 13 SF_20210619 as black blundered in two places: move 47 and move 53.

At move 47, black to play, Kg5-f6 was winning.

Move 47: black to move: hash cleared: multipv=1: threads=15
FEN: 1r6/1p4p1/p2N4/2P3kp/1P2p3/P3Pb2/1K3P1R/8 b - - 3 47

Stockfish_21062123_x64_avx2:
NNUE evaluation using nn-190f102a22c3.nnue enabled
[...]
 43/88    11:00     6,574,518k    9,955k    +2.67    Kg5-f6 Kb2-c3 Kf6-e6 a3-a4 b7-b6 Kc3-c4 Rb8-d8 Rh2-h3 b6xc5 Nd6-b7 Rd8-c8 b4xc5 Ke6-e5 Nb7-d6 Bf3-e2+ Kc4-b4 a6-a5+ Kb4xa5 Rc8xc5+ Nd6-b5 g7-g5 Rh3-h2 h5-h4 Ka5-b6 Rc5-c1 Nb5-d4 Be2-d3 a4-a5 Ke5-f6 a5-a6 Rc1-b1+ Kb6-c5 Bd3xa6 f2-f4 Ba6-f1 Rh2-h1 h4-h3 f4xg5+ Kf6xg5 Nd4-e6+ Kg5-h4 Ne6-f4 Kh4-g3 Rh1-g1+ Kg3-h2 Rg1-g4 Bf1-g2 Kc5-d4 Rb1-b8 Kd4-c3 Rb8-h8 Kc3-d2 Kh2-g1

But at move 47 SF_20210619 played Kg5-g4, and this is a draw:

Move 48: white to move: hash from prior move: multipv=1: threads=15
FEN: 1r6/1p4p1/p2N4/2P4p/1P2p1k1/P3Pb2/1K3P1R/8 w - - 4 48

Stockfish_21062123_x64_avx2:
NNUE evaluation using nn-190f102a22c3.nnue enabled
[...]
 57/42    09:19     5,120,851k    9,158k     0.00    a3-a4 h5-h4 b4-b5 a6xb5 a4xb5 Bf3-e2 Kb2-c3 h4-h3 Kc3-b4 g7-g5 Kb4-a5 Be2-f1 Ka5-b6 Kg4-f3 Kb6-a7 Rb8-f8 Ka7xb7 Bf1-g2 c5-c6 Kf3xf2 c6-c7 g5-g4 Kb7-a7 g4-g3 Nd6xe4+ Kf2-f3 Rh2xh3 Bg2xh3 Ne4-g5+ Kf3-g4 Ng5xh3 Kg4xh3 b5-b6 Rf8-c8 Ka7-a6 g3-g2 b6-b7 Rc8xc7 b7-b8Q g2-g1Q Qb8xc7

The latest SFdev abandons Kg5-g4 and finds Kg5-f6 at d=20:

Move 47: black to move: hash cleared: multipv=1: threads=15
FEN: 1r6/1p4p1/p2N4/2P3kp/1P2p3/P3Pb2/1K3P1R/8 b - - 3 47

Stockfish_21062123_x64_avx2:
NNUE evaluation using nn-190f102a22c3.nnue enabled
[...]
 17/30    00:00     801k    10,966k    +1.59    Kg5-g4 a3-a4 h5-h4 Kb2-c3 h4-h3 Kc3-d4 Bf3-g2 b4-b5 Kg4-f3 c5-c6 b7xc6 b5xc6 Kf3xf2 c6-c7 Rb8-f8 c7-c8R Rf8xc8 Nd6xc8 Kf2-g3 Rh2xh3+ Kg3xh3 Nc8-d6 g7-g5
 18/28    00:00     1,187k    11,411k    +2.38    Kg5-g4 a3-a4 h5-h4 b4-b5 h4-h3 Kb2-a3 a6xb5 a4xb5 Bf3-g2 c5-c6 Rb8-a8+ Ka3-b4 b7xc6 b5xc6 Kg4-f3 c6-c7 Kf3xf2
 19/30    00:00     3,742k    11,443k    +0.84    Kg5-g4 a3-a4 h5-h4 Kb2-c3 h4-h3 b4-b5 Bf3-g2 Kc3-d4 Rb8-d8 Kd4-e5 Rd8-f8 Nd6xb7 Rf8xf2 b5-b6 Rf2-c2 Ke5-d5 Rc2-d2+ Kd5-e5 Kg4-g3 Rh2xh3+ Bg2xh3
 20/32    00:00     6,368k    10,388k    +1.08    Kg5-f6 Kb2-c3 Kf6-e6 Kc3-c4 g7-g5 a3-a4 b7-b6 b4-b5 b6xc5 Kc4xc5 Rb8-d8 Nd6-c4 a6xb5 a4xb5 Bf3-e2 Kc5-b4 Rd8-b8 Rh2-h1 h5-h4 Rh1-g1 Be2xc4 Kb4xc4 Ke6-f5
 21/31    00:00     9,066k    9,865k    +1.05    Kg5-f6 Kb2-c3 Kf6-e6 Kc3-c4 g7-g5 a3-a4 b7-b6 b4-b5 b6xc5 Kc4xc5 Rb8-d8 Nd6-b7 Rd8-c8+ Kc5-b4 a6xb5 a4xb5 Rc8-b8 Nb7-c5+ Ke6-d6 Nc5xe4+ Bf3xe4 Rh2xh5 g5-g4 Kb4-c4 Be4-f3 Kc4-d4 Rb8-c8 e3-e4
 22/41    00:01     12,030k    9,639k    +1.27    Kg5-f6 a3-a4 Kf6-e6 Kb2-c3 Ke6-d5 a4-a5 g7-g5 Rh2-h3 h5-h4 Rh3-h2 Rb8-h8 Nd6xb7 Rh8-h7 Nb7-d6 h4-h3 Nd6-e8 Kd5-c6 Ne8-d6 g5-g4 Nd6-f5 Kc6-d5 Nf5-g3 Bf3-g2

The mistake at iteration 17 is c5-c6 (marked in bold font). There white has Nd6xb7.
The mistake at iteration 18 is c5-c6 (marked in bold font). There white has Ka3-b4.
There is no obvious mistake in iteration 19, but the position at the end of it is a draw, not +0.84 for black.

At move 53 SF_20210619 could have saved the draw with Kf3xf2, but instead lost the game by playing Rd8-b8. The problem with 53...Rd8-b8 is that it allows white to queen the c-pawn quickly after sacking the knight on b7 to create two connected passers, because c6-c7 comes with tempo by attacking the black rook on b8. But 53...Kf3xf2 followed by Rd8-c8 saves the draw by giving black enough tempi to queen the black pawn on the g-file or create a sufficient threat to do so to force white to trade off the pieces.

Move 53: black to move: hash cleared: multipv=3: threads=15
FEN: 3r4/1p4p1/pP1N4/2P1K3/P3p3/4Pk1p/5PbR/8 b - - 2 53
Stockfish_21062123_x64_avx2:
[...]
---------------------------------------------------------------------------
 37/61    55:41     47,838,009k    14,315k    -8.05    Rd8-h8 Nd6xb7 Kf3xf2 c5-c6 g7-g5 Nb7-d6 g5-g4 c6-c7 g4-g3 Nd6xe4+ Kf2-g1 Rh2xh3 Rh8-e8+ Ke5-f4 Re8xe4+ Kf4xg3 Re4-e8 Rh3-h6 Bg2-b7 e3-e4 a6-a5 Rh6-d6 Re8-g8+ Kg3-f4 Rg8-f8+ Kf4-e5 Rf8-e8+ Ke5-f5 Re8-f8+ Rd6-f6 Rf8-e8 e4-e5 Kg1-g2 Kf5-f4 Re8-g8 e5-e6 Rg8-h8 Kf4-f5 Rh8-h3 e6-e7 Rh3-e3 Rf6-e6 Re3-f3+ Kf5-g6 Bb7-c8 Re6-d6 Rf3-g3+ Kg6-f6
 37/61    55:41     47,838,009k    14,315k    -0.22    g7-g5 Nd6xb7 Rd8-c8 Ke5-d4 g5-g4 Nb7-d6 Rc8-d8 b6-b7 Kf3xf2 Kd4-e5 g4-g3 Nd6xe4+ Kf2xe3 Ne4xg3 Ke3-d3 Ng3-f5 Kd3-c4 c5-c6 Kc4-c5 Rh2xg2 Rd8-e8+ Ke5-f6 h3xg2 c6-c7 g2-g1Q c7-c8Q+ Re8xc8 b7xc8Q+ Kc5-b4 Qc8-b8+ Kb4xa4 Qb8-f4+ Ka4-a3 Nf5-d4 Qg1-h1 Qf4-d6+ Ka3-a4 Qd6-c5 Qh1-h6+ Kf6-f7 Qh6-h7+ Kf7-e8 Qh7-h8+ Ke8-d7 Qh8-h7+ Kd7-d6 Qh7-h6+ Nd4-e6 Qh6-d2+ Kd6-c6 Qd2-b4 Qc5-c2+ Ka4-a5 Ne6-c5 Qb4-b5+ Kc6-d5 Ka5-b6
 37/42    55:41     47,838,009k    14,315k     0.00    Kf3xf2 Nd6xb7 Rd8-c8 Ke5-d6 g7-g5 c5-c6 g5-g4 Nb7-c5 g4-g3 Rh2xh3 Bg2xh3 Nc5xe4+ Kf2xe3 Ne4xg3 Rc8-d8+ Kd6-c5 Rd8-d2 b6-b7 Rd2-b2 Kc5-d6 Bh3-g2 Ng3-f5+ Ke3-d3 Kd6-c7 Kd3-c4 Nf5-e3+ Kc4-c5 Ne3xg2 Rb2xg2 b7-b8N Kc5-b4 Kc7-b6 Kb4xa4 c6-c7 Rg2-b2+ Kb6-c6 Rb2-c2+ Kc6-d7 Ka4-b5 c7-c8Q Rc2xc8
---------------------------------------------------------------------------

The latest SFdev abandons Rd8-b8 and finds Kf3xf2 at d=13.

Move 53: black to move: hash cleared: multipv=1: threads=15
FEN: 3r4/1p4p1/pP1N4/2P1K3/P3p3/4Pk1p/5PbR/8 b - - 2 53

Stockfish_21062123_x64_avx2:
NNUE evaluation using nn-190f102a22c3.nnue enabled
[...]
 8/10    00:00     44k    6,264k    +2.36    Rd8-b8 Ke5-d5 g7-g6 a4-a5 Kf3xf2 c5-c6 b7xc6+ Kd5xc6 Rb8-h8
 9/10    00:00     64k    7,958k    +2.43    Rd8-b8 Ke5-e6 Kf3xf2 Ke6-d7 Rb8-f8 Nd6xb7 Kf2-g1 Nb7-d8 Kg1xh2
 10/13    00:00     159k    9,926k    +1.30    Rd8-b8 Ke5-d5 Kf3xf2 Nd6xb7 Kf2-g3 Rh2xh3+ Bg2xh3 c5-c6 Kg3-f3 a4-a5 Kf3xe3 Nb7-d6 Ke3-f4 Nd6xe4
 11/19    00:00     450k    10,970k    +0.67    Rd8-b8 Ke5-d5 Rb8-d8 Kd5-e5 Kf3xf2 Nd6xb7 Rd8-e8+ Ke5-d4 g7-g5 Nb7-d6 Re8-d8
 12/25    00:00     1,040k    10,832k    +1.85    Rd8-b8 Ke5-d5 Kf3xf2 Nd6xb7 Rb8xb7 c5-c6 Rb7xb6 c6-c7 Kf2-g3 c7-c8Q
 13/16    00:00     2,116k    10,634k    -0.11    Kf3xf2 Nd6xb7 Rd8-e8+ Ke5-d4 g7-g5 c5-c6 g5-g4 Nb7-d6 g4-g3 Rh2xh3 Bg2xh3 Nd6xe8 g3-g2 b6-b7 g2-g1Q b7-b8Q
 14/23    00:00     2,541k    10,500k    -0.19    Kf3xf2 Nd6xb7 Rd8-e8+ Ke5-d4 g7-g5 c5-c6 g5-g4 Nb7-d6 Re8-d8 Kd4-c5 Rd8xd6 b6-b7 g4-g3 Rh2xh3 Bg2xh3 Kc5xd6 g3-g2 b7-b8Q
 15/24    00:00     4,038k    10,656k    -1.02    Kf3xf2 Nd6xb7 Rd8-c8 Ke5-d6 g7-g5 c5-c6 g5-g4 Nb7-c5 g4-g3 Rh2xh3 Rc8-d8+ Kd6-c7 Bg2xh3 Nc5xe4+ Kf2xe3 Ne4xg3 Rd8-c8+ Kc7-d6 Rc8-d8+ Kd6-c5 Rd8-f8 b6-b7 Bh3-g2 Kc5-b6
 16/29    00:00     4,189k    10,631k    -0.84    Kf3xf2 Nd6xb7 Rd8-c8 Ke5-d6 g7-g5 c5-c6 g5-g4 Nb7-c5 g4-g3 Rh2xh3 Rc8-d8+ Kd6-c7 Bg2xh3 Nc5xe4+ Kf2xe3 Ne4xg3
 17/26    00:00     5,788k    10,543k    -0.89    Kf3xf2 Nd6xb7 Rd8-c8 Ke5-d6 g7-g5 c5-c6 g5-g4 Nb7-c5 g4-g3 Rh2xh3 Bg2xh3 Nc5xe4+ Kf2xe3 Ne4xg3 Rc8-d8+ Kd6-c5 Rd8-d1 b6-b7 Rd1-b1 Kc5-d6 Ke3-d4 Ng3-h5 Kd4-c4 Nh5-f4 Bh3-g4 Nf4-d5

The mistake in iterations 8 through 12 is the same: expecting white to reply with Ke5-d5. But Ke5-d5 is -6.06 (at d=32) and Nd6xb7 is +5.81 (at d=32). That's because Ke5-d5 loses a tempo compared to Nd6xb7.

Move 54: white to move: hash cleared: multipv=3: threads=15
FEN: 1r6/1p4p1/pP1N4/2P1K3/P3p3/4Pk1p/5PbR/8 w - - 3 54
Stockfish_21062123_x64_avx2:
[...]
---------------------------------------------------------------------------
 32/67    05:09     3,955,682k    12,779k    -6.06    Ke5-d5 Kf3xf2 Nd6xb7 Rb8xb7 a4-a5 Rb7-f7 Kd5-e6 Rf7-f6+ Ke6-e5 Rf6-f8 c5-c6 Kf2-g1 c6-c7 Kg1xh2 b6-b7 Kh2-g1 c7-c8R Rf8xc8 b7xc8Q h3-h2 Qc8xa6 h2-h1Q Qa6-b6 Qh1-h5+ Ke5-d4 Qh5-d1+ Kd4-e5 Bg2-f1 Ke5xe4 Bf1-g2+ Ke4-f4 Qd1-h5 e3-e4+ Kg1-h1 Qb6-b1+ Kh1-h2 Qb1-b8 g7-g5+ Kf4-e3+ Kh2-h3 Ke3-d4 Qh5-d1+ Kd4-e5 Qd1-f3 Qb8-h8+ Kh3-g3 Ke5-d6 Qf3-a3+ Kd6-e6 Qa3-b3+ Ke6-e7 Qb3-b4+ Ke7-f7 Qb4-b7+ Kf7-f6 Qb7xe4 Qh8-d8 Qe4-f4+ Kf6-g7 g5-g4 a5-a6
 32/44    05:09     3,955,682k    12,779k    +5.81    Nd6xb7 Rb8xb7 a4-a5 Rb7-b8 c5-c6 Kf3xf2 c6-c7 Rb8-c8 Ke5-d6 Kf2-g1 Rh2xh3 Bg2xh3 b6-b7 Kg1-f2 b7-b8Q Bh3-f5 Kd6-e5 g7-g6 Ke5-f4 Kf2-e2 Qb8-b6 Ke2-f2 Qb6-b2+ Kf2-e1 Qb2-c1+ Ke1-e2 Qc1-c2+ Ke2-e1 Qc2-c6 Bf5-h3 Qc6-c1+ Ke1-e2 Kf4xe4 g6-g5 Qc1-c2+ Ke2-e1 Ke4-d4 g5-g4 Qc2-h2
---------------------------------------------------------------------------

Bryan

unread,
Jun 21, 2021, 10:17:57 PMJun 21
to FishCooking
In Game 14 SF_20210619 blundered at move 40 by playing Nb4-d3 instead of Ra5-a7.

Move 40: black to move: hash cleared: multipv=1: threads=15
FEN: 8/6k1/4p3/r1pbN2B/1n1r2p1/1P4P1/5R1P/5RK1 b - - 1 40

Stockfish_21062123_x64_avx2:
NNUE evaluation using nn-190f102a22c3.nnue enabled
[...]
 57/66    19:32     10,943,245k    9,333k    -0.03    Ra5-a7 Ne5xg4 Bd5xb3 Rf2-f7+ Ra7xf7 Rf1xf7+ Kg7-h8 Ng4-f6 Bb3-c2 Rf7-c7 Nb4-d5 Nf6xd5 e6xd5 Rc7xc5 Bc2-e4 Rc5-c3 Rd4-c4 Rc3xc4 d5xc4 Kg1-f2 c4-c3 Kf2-e1 Kh8-h7 Ke1-d1 c3-c2+ Kd1-c1 Kh7-h6 Bh5-g4 Kh6-g6 Bg4-e6 Kg6-h5 Be6-d7 Be4-g6 Kc1-d2 Kh5-g5 Bd7-c8 Bg6-d3 Kd2-c1 Bd3-e4 Bc8-h3 Be4-g6 Kc1-b2 Bg6-e4 Bh3-d7 Be4-g6 h2-h3 Bg6-d3 Bd7-g4 Bd3-h7 Kb2-c1 Bh7-g6 Kc1-d2 Kg5-f6 Bg4-d7 Bg6-e4 Bd7-a4 Kf6-g5 Ba4-b5 Be4-g6 Bb5-d7 Bg6-d3 Bd7-c8 Bd3-g6 h3-h4+ Kg5-h5 Kd2-c1 Bg6-h7

Move 41: white to move: hash cleared: multipv=1: threads=15
FEN: 8/6k1/4p3/r1pbN2B/3r2p1/1P1n2P1/5R1P/5RK1 w - - 2 41

Stockfish_21062123_x64_avx2:
NNUE evaluation using nn-190f102a22c3.nnue enabled
 1/1    00:00     899    899k    +M23    Rf2-f7+ Kg7-h6
[...]
 76/46    01:49     2,005,012k    18,240k    +M23    Rf2-f7+ Kg7-h6 Rf7-f6+ Kh6-g7 Rf6-g6+ Kg7-h7 Ne5-d7 e6-e5 Nd7-f6+ Kh7-h8 Rg6-h6+ Kh8-g7 Rh6-h7+ Kg7-f8 Nf6xd5+ Nd3-f4 Nd5-f6 Nf4xh5 Nf6xh5+ Kf8-e8 Rf1-f7 Rd4-d1+ Kg1-f2 Ra5-a2+ Kf2-e3 Rd1-e1+ Ke3-d3 Ra2-a6 Rf7-g7 Ra6-d6+ Kd3-c4 Re1-f1 Rg7-a7 Rd6-d4+ Kc4-b5 Rd4-b4+ Kb5xc5 Rb4-b5+ Kc5xb5 Rf1-f7 Ra7xf7 e5-e4 Nh5-f6+ Ke8-d8 Rf7-f8+

The problem with 40...Nb4-d3 is that it allows white to play Rf2-f7+, and all white's pieces are on the black king's flank, but all of black's pieces are on the other flank. By contrast, 40...Ra5-a7 controls the seventh rank, and doesn't allow white to sustain doubled rooks on the f-file (because after Rf2-f7+ black plays Ra7xf7).

The latest SFdev abandons Nb4-d3 at d=10 and finds Ra5-a7 at d=12.

Move 40: black to move: hash cleared: multipv=1: threads=10
FEN: 8/6k1/4p3/r1pbN2B/1n1r2p1/1P4P1/5R1P/5RK1 b - - 1 40

Stockfish_21062123_x64_avx2:
NNUE evaluation using nn-190f102a22c3.nnue enabled
 1/1    00:00     2k    941k    -0.72    Nb4-d3 Rf2-f7+ Kg7-h6
 2/2    00:00     3k    1,337k    -0.72    Nb4-d3 Rf2-f7+ Kg7-h6
 3/3    00:00     4k    1,842k    -0.72    Nb4-d3 Rf2-f7+ Kg7-h6
 4/4    00:00     5k    2,368k    -0.51    Nb4-d3 Rf2-f7+ Kg7-h6
 5/5    00:00     7k    2,303k    -0.51    Nb4-d3 Rf2-f7+ Kg7-h6 Ne5xd3 Rd4xd3
 6/6    00:00     10k    3,338k    -0.51    Nb4-d3 Ne5xd3 Rd4xd3 Rf2-f7+ Kg7-h6 Bh5xg4
 7/7    00:00     23k    5,784k    -0.84    Nb4-c6 Ne5xg4 Ra5-a7 Ng4-f6 Bd5xb3 Nf6-e8+ Kg7-h6
 8/9    00:00     31k    6,229k    -0.72    Nb4-d3 Ne5xd3 Rd4xd3 Rf2-f7+ Kg7-h6 Bh5xg4 Kh6-g5
 9/11    00:00     46k    7,624k    -0.36    Nb4-d3 Rf2-f7+ Kg7-h6 Ne5xd3 Kh6xh5 Nd3-f4+ Kh5-h6 Rf7-f6+ Kh6-h7 Nf4xe6
 10/17    00:00     239k    8,546k    -0.46    Ra5-a2 Rf2xa2 Nb4xa2 Ne5xg4 Kg7-g8 Ng4-f6+ Kg8-g7 g3-g4 Bd5xb3 g4-g5 Na2-c3 Rf1-f3 Nc3-e2+ Kg1-f1
 11/17    00:00     449k    8,982k    -0.60    Ra5-a2 Rf2xa2 Nb4xa2 Ne5xg4 Kg7-g8 Ng4-f6+ Kg8-g7 Nf6-e8+ Kg7-h6 Bh5-f7 Bd5xb3 h2-h4 Rd4-d1 g3-g4 Rd1xf1+ Kg1xf1
 12/19    00:00     604k    9,015k    -0.36    Ra5-a7 Ne5xg4 Bd5xb3 Rf2-f6 Bb3-c2 Rf6xe6 Bc2-d3 Rf1-e1 c5-c4 Re6-e7+ Ra7xe7 Re1xe7+ Kg7-h8
 13/19    00:00     1,121k    9,044k    -0.48    Ra5-a7 Ne5xg4 Bd5xb3 Rf2-f7+ Ra7xf7 Rf1xf7+ Kg7-h8 Ng4-e5 Nb4-d5 Bh5-g6 Rd4-d1+ Kg1-g2 Rd1-d2+ Kg2-h3 Bb3-c2 Rf7-h7+ Kh8-g8
 14/23    00:00     1,354k    9,027k    -0.37    Ra5-a7 Ne5xg4 Bd5xb3 Rf2-f6 Bb3-c2 Rf6-f7+ Ra7xf7 Rf1xf7+ Kg7-h8 Rf7-c7 Bc2-f5 Ng4-f6 Nb4-d5 Nf6xd5 e6xd5 Rc7xc5 Bf5-e4 Rc5-c3

Bryan

unread,
Jun 22, 2021, 12:08:01 AMJun 22
to FishCooking
In Game 15, SF_20210619 as white blundered at move 24, by playing Nf3-h2.

At move 24, black to move, black is winning.

Move 24: black to move: hash cleared: multipv=1: threads=15
FEN: 4r1rk/pb3q1p/1p2pp2/3p4/3Pn2Q/3BP2P/PP3P1N/2R3RK b - - 4 24
Stockfish_21062123_x64_avx2:
Found 145 tablebases

NNUE evaluation using nn-190f102a22c3.nnue enabled
[...]
 35/58    02:06     1,691,437k    13,352k    +7.76    Qf7-h5 Qh4-f4 e6-e5 d4xe5 f6xe5 Qf4-f3 Qh5xf3+ Nh2xf3 Ne4xf2+ Kh1-h2 Nf2xd3 Rc1-c7 Rg8xg1 Nf3xg1 Bb7-c8 a2-a4 Nd3xb2 Ng1-f3 Bc8-f5 Rc7-f7 Bf5-g6 Rf7xa7 Nb2-c4 Nf3-g5 Kh8-g8 Ra7-a6 h7-h6 Ng5-f3 Kg8-g7 Kh2-g3 Bg6-f5 Nf3-h2 Re8-e6 Ra6-a7+ Kg7-g6 Ra7-a8 Kg6-f7 Ra8-a7+ Kf7-e8 Kg3-h4 Nc4xe3 a4-a5 b6xa5 Ra7xa5

But at move 24, white to move, white has at least three moves that give white a winning advantage: Kh2, a3, and Bd3xe4:

Move 24: white to move: hash cleared: multipv=8: threads=15
FEN: 4r1rk/pb3q1p/1p2pp2/3p4/3Pn2Q/3BPN1P/PP3P2/2R3RK w - - 3 24
Stockfish_21062123_x64_avx2:
[...]
---------------------------------------------------------------------------
 44/66    1:14:13     38,102,562k    8,556k    +1.51    Kh1-h2 Rg8xg1 Nf3xg1 Qf7-e7 Ng1-e2 e6-e5 Rc1-g1 Re8-g8 Rg1xg8+ Kh8xg8 b2-b3 a7-a5 f2-f3 Ne4-g5 Qh4-g4 Kg8-h8 Kh2-g2 Bb7-c6 d4xe5 f6xe5 Qg4-c8+ Qe7-e8 Qc8-c7 Bc6-d7 Qc7xb6 Bd7xh3+ Kg2-g3 Bh3-e6 f3-f4 Ng5-f7 f4-f5 Be6-c8 Qb6xa5 Qe8-d7 Kg3-f2 e5-e4 Qa5-c5 Kh8-g7 Bd3-c2 Qd7xf5+ Ne2-f4 Nf7-e5 Qc5xd5 Ne5-d3+ Bc2xd3 Qf5xd5 Nf4xd5 e4xd3 Nd5-f4 Bc8-a6 Kf2-e1 Kg7-f6 a2-a4 Kf6-e5 Ke1-d2 Ke5-e4 a4-a5 Ba6-b5 Nf4-h5 Ke4-f5
 44/74    1:14:13     38,102,562k    8,556k    +1.62    a2-a3 Re8-f8 b2-b4 Bb7-c8 Bd3xe4 d5xe4 Nf3-d2 Bc8-b7 Nd2-c4 Bb7-d5 Nc4-d6 Rg8xg1+ Rc1xg1 Qf7-e7 Qh4-f4 Rf8-d8 Nd6-b5 Rd8-f8 Rg1-c1 e6-e5 Qf4-h6 Qe7-g7 Qh6xg7+ Kh8xg7 d4xe5 f6xe5 Rc1-c7+ Kg7-h8 Nb5-c3 Bd5-e6 Rc7-e7 Rf8-f6 b4-b5 Kh8-g8 a3-a4 Be6xh3 Nc3xe4 Rf6-g6 Kh1-h2 Bh3-e6 f2-f3 Be6-c4 Re7xe5 Bc4-b3 f3-f4 Rg6-e6 Re5xe6 Bb3xe6 Kh2-g3 Kg8-f8
 44/63    1:14:13     38,102,562k    8,556k    +1.75    Bd3xe4 d5xe4 Nf3-h2 f6-f5 Qh4-f4 Rg8xg1+ Rc1xg1 e6-e5 d4xe5 Qf7-e6 Rg1-g5 Bb7-c8 Qf4-h4 Bc8-a6 Rg5-h5 Qe6-f7 b2-b3 Ba6-e2 Rh5-g5 Re8-f8 Qh4-h6 Qf7-e7 Rg5-g3 Rf8-g8 e5-e6 Be2-a6 Rg3xg8+ Kh8xg8 Qh6-f4 Qe7xe6 Qf4-b8+ Ba6-c8 Kh1-g1 Qe6-d7 Nh2-f1 Kg8-f7 Nf1-g3 Kf7-e6 h3-h4 Ke6-e7 Kg1-g2 Ke7-d8 Ng3-h5 Qd7-e7 Qb8-g3 Bc8-e6 Nh5-f4 Kd8-c8 Nf4xe6 Qe7xe6 Qg3-g7 Qe6-g6+ Qg7xg6 h7xg6 Kg2-g3 Kc8-d8

One reason why 24. Nf3-h2 is a mistake is that after Nf3-h2 the white king on h1 is left with no safe squares, and for that reason if the f2 square is not protected, black can play Ne4xf2#. So at move 24 when black plays Qf7-h5, white cannot capture the unprotected queen with Qh4xh5, because the white queen is the only piece defending the f2 square. As soon as black exchanges off queens at move 26, black can fork the king and bishop with Ne4xf2. So this black knight on e4 is a real threat in this position. That's why the much better move at move 24 is Bd3xe4, eliminating this knight.

SFdev abandons 24.Nf3-h2 at d=18.

Move 24: white to move: hash cleared: multipv=1: threads=15
FEN: 4r1rk/pb3q1p/1p2pp2/3p4/3Pn2Q/3BPN1P/PP3P2/2R3RK w - - 3 24

Stockfish_21062123_x64_avx2:
NNUE evaluation using nn-190f102a22c3.nnue enabled
 [...]
 15/26    00:00     2,232k    9,457k    +1.74    Nf3-h2 Qf7-e7 Nh2-g4 Rg8-g6 f2-f3 f6-f5 Qh4xe7 Re8xe7 Ng4-e5 Rg6xg1+ Kh1xg1 Ne4-g5 Kg1-h2 Ng5-f7 Kh2-g3 Nf7xe5 d4xe5 Kh8-g7 Kg3-f4 Kg7-h6 h3-h4 a7-a6
 16/30    00:00     2,541k    9,447k    +1.75    Nf3-h2 Qf7-e7 Nh2-g4 Rg8-g6 f2-f3 f6-f5 Qh4xe7 Re8xe7 Ng4-e5 Rg6xg1+ Kh1xg1 Ne4-g5 Kg1-h2 Ng5-f7 Kh2-g3 Nf7xe5 d4xe5 Kh8-g7 Kg3-f4 Kg7-h6 Bd3-b5 a7-a6
 17/34    00:00     4,552k    9,347k    +1.76    Nf3-h2 Qf7-e7 Nh2-g4 Rg8-g6 f2-f3 f6-f5 Qh4xe7 Re8xe7 Ng4-e5 Rg6xg1+ Kh1xg1 Ne4-g5 Kg1-g2 Ng5-f7 Kg2-g3 Nf7xe5 d4xe5 Kh8-g7 Kg3-f4 Kg7-h6 b2-b4 Re7-g7 a2-a4 d5-d4 e3xd4 Bb7-d5 Bd3-b5
 18/26    00:00     6,294k    9,228k    +1.46    Bd3xe4 d5xe4 Nf3-h2 f6-f5 Qh4-f4 Rg8xg1+ Rc1xg1 e6-e5 d4xe5 Qf7-e6 Rg1-g5 Qe6xe5 Qf4-h4 Qe5xb2 Rg5xf5 Qb2-a1+ Kh1-g2 Re8-g8+ Nh2-g4 Qa1-g7 Kg2-h2
 19/33    00:00     7,523k    9,163k    +1.36    Bd3xe4 d5xe4 Nf3-h2 f6-f5 Qh4-f4 Rg8xg1+ Rc1xg1 e6-e5 d4xe5 Qf7-e6 Rg1-g5 Qe6xe5 Qf4-h4 Qe5xb2 Rg5xf5 Qb2-a1+ Kh1-g2 Re8-g8+ Nh2-g4 Qa1-g7 Kg2-h2 Qg7-g6 Rf5-f4 Qg6-g5 Qh4xg5 Rg8xg5 Ng4-f6 Kh8-g7 Nf6-e8+ Kg7-g6 Rf4-f6+ Kg6-h5 Ne8-d6
 20/35    00:01     10,050k    9,145k    +1.30    Bd3xe4 d5xe4 Nf3-h2 f6-f5 Qh4-f4 Rg8xg1+ Rc1xg1 e6-e5 d4xe5 Qf7-e6 Rg1-g5 Bb7-c8 Qf4-h4 Bc8-a6 a2-a3 Ba6-e2 Qh4-g3 a7-a6 Qg3-f4 Qe6xe5 Rg5xf5 Qe5xf4 Rf5xf4 h7-h5 Rf4-f6 a6-a5 Rf6xb6 Re8-d8 Kh1-g2

Iterations 15, 16, and 17 show the same mistake. At those iterations SFdev expects black to reply with Qf7-e7, when black's much better reply is Qf7-h5.
Reply all
Reply to author
Forward
0 new messages