Interesting Test Position

[Paloma]

....But as Louis shows, Ne5-g4 is much better. Does Stockfish's line transpose with the Houdini Qd2? What about Komodo here.

Wrong! Qd2 is rhe best move!

info depth 45 seldepth 87 multipv 1 score cp 317 nodes 301807158343 nps 28815068 hashfull 999 tbhits 0 time 10473935 pv e2d2 b7a6 e5g4 f8c8 d2g5 g8f8 f6f2 c8c7 g5g6 a6b5 d1e1 a8e8 g4e5 f7f6 g6h7 f8e7 h7g7 e7d8 e5f7 c7f7 g7f7 f6f5 f7a7 d6c7 a7c7 d8c7 h3h4 c7d6 f2f3 e8h8 g2g3 b5e8 g1f2 h8g8 b2b4 c5b4 c3b4 g8g4 f3f4 g4g7 g3g4 f5g4 f4f6 g4g3 f2g2 e8f7 e1e5 d6e7 f6f4 g7h7 g2g3 f7g6 a2a4 g6e4 h4h5 h7h8 a4a5 b6a5 b4a5 e7d6

info depth 45 seldepth 87 multipv 2 score cp 252 nodes 301807158343 nps 28815068 hashfull 999 tbhits 0 time 10473935 pv e5g4 f8c8 e2d2 g8f8 f6f7 f8f7 d1f1 f7e8 d2g5 b7a6 g5h5 e8d7 g4e5 d7c7 f1f7 c7b8 f7d7 d6d7 e5d7 b8b7 h5f7 c8c7 f7e6 a6b5 d7e5 c5d4 e6d5 b5c6 d5d4 a8h8 e5c6 b7c6 d4e4 c6b5 a2a4 b5a6 b2b4 c7c3 e4e7 b6b5 e7d6 a6b7

http://www.talkchess.com/forum/viewtopi ... 87&t=59033

[Eelco de Groot]

Hi Herbert,

I was referring there Ne5-g4 to the currently best move Rxd2 found by Rainbow Serpent (and Marc's Stockfish 7, not Louis's deeper search with the most recent Stockfish). Even though the score is higher for Qxd2 as first move, I think right away playing the knight is at least also a winning move, several moves in in the PV are the same and I see not clearly that the two moves would not transpose. Not completely perhaps because apparently in the Qxd2 line Stockfish does not find a necessary refutation until very late that does not seem to matter in the Ne5-g4 line.

But just the different scores are not always to be trusted, even transpositions can get a different score. I am not sure how that works but it can happen.

Eelco

Rxd2 is also not so bad it seems, but Qxd2 still not found here as winning:


r4rk1/pb3pp1/1p1qpR2/2ppN3/3P4/2P4P/PP1pQ1P1/3R2K1 w - -

Engine: Rainbow Serpent Sf20151218_008 MOD MP (512 MB)
by T. Romstad, M. Costalba, J. Kiiski, G. Linscott

32 41:25 +0.60 1.Rxd2 Qd8 2.Rf4 Qg5 3.Rg4 Qh6
4.dxc5 bxc5 5.Qf2 d4 6.cxd4 Rad8
7.Rh4 Qg5 8.Qxf7+ Rxf7 9.Rh8+ Kxh8
10.Nxf7+ Kg8 11.Nxg5 Rxd4 12.Rxd4 cxd4
13.Nxe6 d3 14.Kf2 (7.703.354.259) 3099

32 41:25 +0.41 1.Qxd2 Ba6 2.Ng4 Qe7 3.Qf2 Rac8
4.Nh6+ Kh7 5.Rxf7 Rxf7 6.Nxf7 Qf6
7.Ne5 Qxf2+ 8.Kxf2 cxd4 9.Rxd4 Rf8+
10.Kg3 Bb5 11.Nf3 Kg8 12.h4 Rf6
13.Kf2 Be8 14.Ke3 (7.703.354.259) 3099

32 41:25 +0.25 1.Rf4 cxd4 2.cxd4 f6 3.Ng6 Rfc8
4.Qxd2 Bc6 5.Rh4 Be8 6.Qd3 Bxg6
7.Qxg6 Rc7 8.Rf1 Rf8 9.Qh7+ Kf7
10.Rg4 Ke8 11.Rxg7 Rxg7 12.Qxg7 Qe7
13.Qg6+ Kd7 14.Rc1 (7.703.354.259) 3099

32 41:25 0.00 1.Qh5 gxf6 2.Qg4+ Kh8 3.Qh4+ Kg7
4.Qg3+ Kh7 5.Qh4+ Kg7 (7.703.354.259) 3099

32 41:25 0.00 1.Rdf1 Ba6 2.Qh5 Bxf1 3.Nxf7 Rxf7
4.Qxf7+ Kh8 5.Qh5+ Kg8 6.Qf7+ (7.703.354.259) 3099

32 41:25 0.00 1.Qg4 cxd4 2.Qh5 gxf6 3.Qg4+ Kh7
4.Qh4+ Kg7 5.Qg3+ Kh7 6.Qh4+ (7.703.354.259) 3099

______________________________________________________________________________

33 108:57 +1.54 1.Ng4 Rfe8 2.Qxd2 Kf8 3.Rf3 cxd4
4.cxd4 f5 5.Qg5 Ba6 6.Ne5 Kg8 7.Qh5 Qf8
8.Re1 Qf6 9.Rg3 Bb5 10.Ng6 f4 11.Rg4 Bd3
12.Qh8+ Kf7 13.Ne5+ Qxe5 14.Qxe8+ (20.435.778.122) 3125

33 108:57 +0.53 1.Rxd2 Qd8 2.Rf4 Qg5 3.Rg4 Qh6
4.dxc5 bxc5 5.Qf2 d4 6.cxd4 Rad8
7.Rh4 Qg5 8.Qxf7+ Rxf7 9.Rh8+ Kxh8
10.Nxf7+ Kg8 11.Nxg5 Rxd4 12.Rxd4 cxd4
13.Nxe6 d3 14.Kf2 (20.435.778.122) 3125

33 108:57 +0.39 1.Qxd2 Ba6 2.Ng4 Qe7 3.Qf2 Rac8
4.Nh6+ Kh7 5.Rxf7 Rxf7 6.Nxf7 Qf6
7.Ne5 Qxf2+ 8.Kxf2 cxd4 9.Rxd4 Rf8+
10.Kg3 Bb5 11.Rf4 Kg8 12.h4 Rf6
13.b3 Kf8 14.a4 (20.435.778.122) 3125

33 108:57 +0.30 1.Rf4 cxd4 2.cxd4 f6 3.Ng6 Rfc8
4.Qxd2 Bc6 5.Rh4 Be8 6.Qd3 Bxg6
7.Qxg6 Rc7 8.Rf1 Rac8 9.Qh7+ Kf7
10.Rh6 Qe7 11.Rg6 Ke8 12.Qg8+ Kd7
13.Qxg7 Qxg7 14.Rxg7+ (20.435.778.122) 3125

33 108:57 0.00 1.Qh5 gxf6 2.Qg4+ Kh8 3.Qh4+ Kg7
4.Qg3+ Kh7 5.Qh4+ Kg7 (20.435.778.122) 3125

33 108:57 0.00-- 1.Rdf1 Ba6 2.Qh5 Bxf1 3.Nxf7 Rxf7
4.Qxf7+ Kh8 5.Qh5+ Kg8 6.Qf7+ (20.435.778.122) 3125

________________________________________________________________________________

34 256:04 +2.08 1.Ng4 Rfc8 2.Qxd2 Kf8 3.Rxf7+ Kxf7
4.Rf1+ Ke8 5.Qg5 Ba6 6.Qh5+ Kd7
7.Ne5+ Kc7 8.Rf7+ Kb8 9.Nd7+ Kc7
10.Nxc5+ Kc6 11.Nxa6 b5 12.Qe2 Kb6
13.Nc5 Rc7 14.Rxc7 (48.899.761.173) 3182

34 256:04 +1.10 1.Rxd2 Qd8 2.Rf4 g5 3.Rf2 f6 4.Ng6 Rf7
5.Qxe6 Kg7 6.Ne5 Rf8 7.dxc5 Qe8
8.Qxe8 Raxe8 9.Nd3 Re4 10.cxb6 axb6
11.Nb4 Rd8 12.Rf1 Re5 13.a3 Kg6
14.Rfd1 (48.899.761.173) 3182

34 256:04 +0.47 1.Qxd2 Ba6 2.Ng4 Qe7 3.Qf2 Bc4
4.dxc5 bxc5 5.Qh4 Rfe8 6.Qg3 Kf8
7.Rf2 f5 8.Ne5 Kg8 9.Nxc4 dxc4
10.Re1 Rad8 11.Rxf5 exf5 12.Rxe7 Rxe7
13.Qf4 Rd5 14.Qxc4 (48.899.761.173) 3182

34 256:04 +0.27 1.Rf4 cxd4 2.cxd4 f6 3.Ng6 Rfc8
4.Qxd2 Bc6 5.Rh4 Be8 6.Qd3 Bxg6
7.Qxg6 Rc7 8.Rf1 Rf8 9.Qh7+ Kf7
10.Rg4 Ke8 11.Rxg7 Rxg7 12.Qxg7 Qe7
13.Qg6+ Kd7 14.Rc1 (48.899.761.173) 3182

34 256:04 0.00 1.Qh5 gxf6 2.Qg4+ Kh8 3.Qh4+ Kg7
4.Qg3+ Kh7 5.Qh4+ Kg7 (48.899.761.173) 3182

34 256:04 0.00 1.Rdf1 Ba6 2.Qh5 Bxf1 3.Nxf7 Rxf7
4.Qxf7+ Kh8 5.Qh5+ Kg8 6.Qf7+ (48.899.761.173) 3182

_______________________________________________________________________________

35 324:36 +2.22 1.Ng4 Rfc8 2.Qxd2 Kf8 3.Rxf7+ Kxf7
4.Rf1+ Ke8 5.Qg5 Ba6 6.Rf2 Kd7
7.Ne5+ Kc7 8.Rf7+ Kb8 9.Rd7 Qxd7
10.Nxd7+ Kb7 11.Qe7 Rc7 12.Qxe6 Bb5
13.Ne5 cxd4 14.Qxd5+ (62.358.673.988) 3201

35 324:36 +1.35 1.Rxd2 Qd8 2.Rf4 Qg5 3.Rg4 Qh6 4.Qf2 f6
5.Ng6 cxd4 6.Rd3 dxc3 7.Rxc3 Rfc8
8.Ne7+ Kf7 9.Nxc8 Rxc8 10.Rgg3 Rxc3
11.Rxc3 Ba6 12.Rc7+ Kg8 13.Rxa7 Bd3
14.Rb7 (62.358.673.988) 3201

[peter]

I was referring there Ne5-g4 to the currently best move Rxd2 found by Rainbow Serpent (and Marc's Stockfish 7, not Louis's deeper search with the most recent Stockfish). Even though the score is higher for Qxd2 as first move, I think right away playing the knight is at least also a winning move, several moves in in the PV are the same and I see not clearly that the two moves would not transpose. Not completely perhaps because apparently in the Qxd2 line Stockfish does not find a necessary refutation until very late that does not seem to matter in the Ne5-g4 line.

Based on the variants of Marco's Houdini and Louis' SF I made this little .pgn, evals are from K9.3 at the variants' ends:

[Event "Tata Steel 2016, CCC"]
[Site "?"]
[Date "2016.01.25"]
[Round "?"]
[White "Carlsen"]
[Black "Tomashevsky"]
[Result "1-0"]
[Annotator "Martan,Peter"]
[FEN "r4rk1/pb3pp1/1p1qpR2/2ppN3/3P4/2P4P/PP1pQ1P1/3R2K1 w - - 0 1"]
[SetUp "1"]

1.Ng4 ( 1.Rxd2 $6 Qd8 2.Rf4 Qg5 3.Rg4 Qh6 4.dxc5 bxc5 5.Qf2 d4 6.cxd4 ) ( 1.Qxd2 $5 Ba6 ( 1...Qe7 2.Rdf1 ) ( 1...cxd4 $2 2.Qg5 { [%eval 284,28] } ) 2.Ng4 Qe7 ( 2...Rfc8 3.Qg5 Kf8 ( 3...Qe7 4.Rd2 Rc7 5.Rdf2 cxd4 6.Nh6+ ) 4.Rf2 Rc7 5.Qg6 cxd4 ( 5...Bb5 6.Re1 ) 6.cxd4 ) 3.Qf2 Bb5 ( 3...Bc4 $2 4.dxc5 bxc5 5.b3 Ba6 6.Qh4 { [%eval 174,27] } ) ( 3...Rfd8 4.Qg3 ) ( 3...Rac8 $6 4.Qh4 Be2 5.Nh6+ { [%eval 132,30] } ) ( 3...Rfc8 $2 4.Qg3 { [%eval 161,27] } ) 4.dxc5 bxc5 ( 4...Rfc8 $2 5.Rd4 { [%eval 281,27] } ) ( 4...Rfd8 $2 5.Qh4 { [%eval 186,28] } ) 5.Qh4 Be2 6.Nh6+ { [%eval 127,31] } ) ( 1.Rf4 $6 f6 2.Ng6 Rfe8 3.Qh5 e5 4.dxe5 Rxe5 5.Nxe5 Qxe5 6.Qxe5 fxe5 { [%eval 29,27] } ) 1...Rfc8 ( 1...cxd4 2.Rdf1 Qc5 3.Nh6+ { [%eval 228,28] } ) ( 1...Rfe8 2.Qxd2 ) 2.Qxd2 Kf8 3.Rxf7+ Kxf7 4.Rf1+ Ke8 5.Qg5 Ba6 6.Rf2 { [%eval 210,30] } 1-0

After Resolving all of that with komodo, I get this output:

[d]r4rk1/pb3pp1/1p1qpR2/2ppN3/3P4/2P4P/PP1pQ1P1/3R2K1 w - - 0 1

Analysis by Komodo 9.3 64-bit:

1.Ng4 Rfc8 2.Qxd2 Kf8 3.Rxf7+ Kxf7 4.Rf1+ Ke8 5.Ne5 Ba6 6.Rf2 Qe7 7.Qd1 Kd8 8.Qh5 Rab8 9.Rf7 Rb7 10.Rxe7 Rxe7 11.Qh4 Rcc7 12.Ng6 cxd4 13.cxd4 Bb5 14.b3 Ba6 15.a4 Bd3 16.Nxe7 Rxe7 17.Qg3 Ba6 18.Qb8+ Bc8 19.Qd6+ Ke8 20.Kf2 Rf7+ 21.Ke3 Rd7 22.Qb8 Kd8 23.Qf4 Ke8 24.b4 Rb7 25.Qd6 Rf7 26.Qe5 Kd8 27.Qg5+ Ke8 28.Qg3 Kd7 29.b5 Ke8
+- (2.92) Depth: 40 00:19:12 30134MN

I would have changed to 3 MV after the Backward, but that deletes komodo's hash at once, as almost every GUI- command given during analysis does, eg. "Save Hash" especially.


3MV with SF after Backward keeps 1.Ng4 on top as well

[zullil]

I was referring there Ne5-g4 to the currently best move Rxd2 found by Rainbow Serpent (and Marc's Stockfish 7, not Louis's deeper search with the most recent Stockfish). Even though the score is higher for Qxd2 as first move, I think right away playing the knight is at least also a winning move, several moves in in the PV are the same and I see not clearly that the two moves would not transpose. Not completely perhaps because apparently in the Qxd2 line Stockfish does not find a necessary refutation until very late that does not seem to matter in the Ne5-g4 line.

Based on the variants of Marco's Houdini and Louis' SF I made this little .pgn, evals are from K9.3 at the variants' ends:

[Event "Tata Steel 2016, CCC"]
[Site "?"]
[Date "2016.01.25"]
[Round "?"]
[White "Carlsen"]
[Black "Tomashevsky"]
[Result "1-0"]
[Annotator "Martan,Peter"]
[FEN "r4rk1/pb3pp1/1p1qpR2/2ppN3/3P4/2P4P/PP1pQ1P1/3R2K1 w - - 0 1"]
[SetUp "1"]

1.Ng4 ( 1.Rxd2 $6 Qd8 2.Rf4 Qg5 3.Rg4 Qh6 4.dxc5 bxc5 5.Qf2 d4 6.cxd4 ) ( 1.Qxd2 $5 Ba6 ( 1...Qe7 2.Rdf1 ) ( 1...cxd4 $2 2.Qg5 { [%eval 284,28] } ) 2.Ng4 Qe7 ( 2...Rfc8 3.Qg5 Kf8 ( 3...Qe7 4.Rd2 Rc7 5.Rdf2 cxd4 6.Nh6+ ) 4.Rf2 Rc7 5.Qg6 cxd4 ( 5...Bb5 6.Re1 ) 6.cxd4 ) 3.Qf2 Bb5 ( 3...Bc4 $2 4.dxc5 bxc5 5.b3 Ba6 6.Qh4 { [%eval 174,27] } ) ( 3...Rfd8 4.Qg3 ) ( 3...Rac8 $6 4.Qh4 Be2 5.Nh6+ { [%eval 132,30] } ) ( 3...Rfc8 $2 4.Qg3 { [%eval 161,27] } ) 4.dxc5 bxc5 ( 4...Rfc8 $2 5.Rd4 { [%eval 281,27] } ) ( 4...Rfd8 $2 5.Qh4 { [%eval 186,28] } ) 5.Qh4 Be2 6.Nh6+ { [%eval 127,31] } ) ( 1.Rf4 $6 f6 2.Ng6 Rfe8 3.Qh5 e5 4.dxe5 Rxe5 5.Nxe5 Qxe5 6.Qxe5 fxe5 { [%eval 29,27] } ) 1...Rfc8 ( 1...cxd4 2.Rdf1 Qc5 3.Nh6+ { [%eval 228,28] } ) ( 1...Rfe8 2.Qxd2 ) 2.Qxd2 Kf8 3.Rxf7+ Kxf7 4.Rf1+ Ke8 5.Qg5 Ba6 6.Rf2 { [%eval 210,30] } 1-0

After Resolving all of that with komodo, I get this output:

[d]r4rk1/pb3pp1/1p1qpR2/2ppN3/3P4/2P4P/PP1pQ1P1/3R2K1 w - - 0 1

Analysis by Komodo 9.3 64-bit:

1.Ng4 Rfc8 2.Qxd2 Kf8 3.Rxf7+ Kxf7 4.Rf1+ Ke8 5.Ne5 Ba6 6.Rf2 Qe7 7.Qd1 Kd8 8.Qh5 Rab8 9.Rf7 Rb7 10.Rxe7 Rxe7 11.Qh4 Rcc7 12.Ng6 cxd4 13.cxd4 Bb5 14.b3 Ba6 15.a4 Bd3 16.Nxe7 Rxe7 17.Qg3 Ba6 18.Qb8+ Bc8 19.Qd6+ Ke8 20.Kf2 Rf7+ 21.Ke3 Rd7 22.Qb8 Kd8 23.Qf4 Ke8 24.b4 Rb7 25.Qd6 Rf7 26.Qe5 Kd8 27.Qg5+ Ke8 28.Qg3 Kd7 29.b5 Ke8
+- (2.92) Depth: 40 00:19:12 30134MN

I would have changed to 3 MV after the Backward, but that deletes komodo's hash at once, as almost every GUI- command given during analysis does, eg. "Save Hash" especially.


3MV with SF after Backward keeps 1.Ng4 on top as well

Here's where I stopped:

Code: Select all

info depth 46 seldepth 92 multipv 1 score cp 372 nodes 558241513116 nps 28636946 hashfull 999 tbhits 0 time 19493751 pv e2d2 b7a6 e5g4 f8c8 d2g5 g8f8 f6f2 c8c7 g5g6 a6b5 d1e1 a8e8 g4e5 f7f6 g6h7 f8e7 h7g7 e7d8 e5f7 c7f7 g7f7 f6f5 f7a7 b5d7 d4c5 b6c5 f2e2 e8e7 a7a5 d8c8 e2e5 c8b7 c3c4 d7c6 h3h4 b7c8 a5a6 c8d7 g1f2 d5c4 a6a7 d7e8 a7c5 d6d7 c5c4 c6e4 e5e4 f5e4 e1e4 d7d6 f2g1 e8f8 c4b4 d6b4 e4b4 e7c7 g2g4 f8f7 g1f2 f7f6 a2a4 c7h7 g4g5 f6f5 f2e3 h7a7 b4f4 f5e5

info depth 46 seldepth 92 multipv 2 score cp 275 nodes 558241513116 nps 28636946 hashfull 999 tbhits 0 time 19493751 pv e5g4 f8c8 e2d2 g8f8 f6f7 f8f7 d1f1 f7e8 d2g5 b7a6 g5h5 e8d7 g4e5 d7c7 f1f7 c7b8 f7d7 d6d7 e5d7 b8b7 h5f7 c8c7 f7e6 a6b5 d7e5 a8c8 e5f7 b7a6 f7d6 c8d8 d4c5 c7c5 d6b5 a6b5 e6e7 c5c8 e7g7 d5d4 c3d4 b5a6 g7e5 c8c1 g1h2 c1c2 e5e3 c2c7 h3h4 c7d7 h4h5 d7d4 b2b3 d4h4 h2g3 h4h1 e3e2 a6b7 e2e4 b7c7 e4f4 c7c8 g3f2 d8d1 f4c4 c8d7 g2g4 h1h3 c4f7 d7c6 f7e6 c6c7 e6e7 c7c6 e7e8 c6c7 e8e5 c7d7 f2g2 h3d3 g4g5 d1d2 g2f1 d3d5

[peter]

Hi Louis!

And here some SF- MateFinder- Output, started with empty hash again:

r4rk1/pb3pp1/1p1qpR2/2ppN3/3P4/2P4P/PP1pQ1P1/3R2K1 w - - 0 1

Analysis by SF MateFinder 64 POPCNT:

1.Tf2 cxd4 2.cxd4 De7 3.Txd2 Tac8 4.Dh5
-/+ (-0.71) Tiefe: 7/7 00:00:00 109kN
1.Txd2 La6 2.Dg4 cxd4 3.cxd4 Tac8 4.Tdf2 Tc7
=/+ (-0.41) Tiefe: 8/10 00:00:00 247kN
…
1.Txd2 Dd8 2.Tf4 Dg5 3.Tg4 Dh6 4.dxc5 bxc5 5.Df2 d4 6.cxd4 Tad8 7.Th4 Dg5 8.Dxf7+ Txf7 9.Th8+ Kxh8 10.Sxf7+ Kg8 11.Sxg5 Txd4 12.Txd4 cxd4 13.Sxe6 d3 14.Kf2 Lxg2 15.Sf4 d2 16.Ke2 Lc6 17.Kxd2 Kf7 18.h4 Ld7 19.Sh5 Kg6 20.Sg3 Kf6 21.b3 Ke5 22.Ke3 Lg4 23.Se4 Lf5 24.Sd2 Lg6 25.Sc4+ Kf6 26.Kd4 Lb1 27.a4
+/= (0.59) Tiefe: 31/60 00:02:29 3005MN
1.Sg4
+/= (0.66 ++) Tiefe: 32/60 00:03:43 4651MN
…
1.Sg4 Tfc8 2.Dxd2 Kf8 3.Txf7+ Kxf7 4.Tf1+ Ke8 5.Dg5 La6 6.Dh5+ Kd7 7.Se5+ Kc7 8.Tf7+ Kb8 9.Td7 Df8 10.Sf7 Tc6 11.Dg5 De8 12.Td8+ Kb7 13.Txe8 Txe8 14.dxc5 Ka8 15.Sd6 Tg8 16.De7 bxc5 17.Dd7 Tb6 18.Se8 c4 19.Sc7+ Kb8 20.Sxe6 d4 21.cxd4 Tc8 22.Sc5 Lb7 23.Sxb7 Txb7 24.Dd6+ Ka8 25.Dd5 Kb8
+- (1.98) Tiefe: 32/61 00:07:57 10602MN
…
1.Sg4 Tfc8 2.Dxd2 Kf8 3.Txf7+ Kxf7 4.Tf1+ Ke8 5.Dg5 La6 6.Dh5+ Kd7 7.Se5+ Kc7 8.Tf7+ Kb8 9.Td7 Df8 10.Sf7 Tc6 11.Dg5 De8 12.Td8+ Kb7 13.Txe8 Txe8 14.dxc5 Ka8 15.Sd6 Tg8 16.De7 Txc5 17.Dxe6 Tb8 18.De7 g5 19.Kh2 Tc6 20.De5 Td8 21.Sf7 Tdc8 22.Dxg5 Lb7 23.Sd8 Td6 24.Sxb7 Kxb7 25.De7+ Kc6 26.Dxa7 d4 27.cxd4 Txd4 28.De7 Tcd8
+- (2.37) Tiefe: 34/68 00:13:56 19087MN
…
1.Sg4 Tfc8 2.Dxd2 Kf8 3.Txf7+ Kxf7 4.Tf1+ Ke8 5.Dg5 La6 6.Dh5+ Kd7 7.Se5+ Kc7 8.Tf7+ Kb8 9.Td7 Df8 10.Sf7 Tc6 11.Dg5 De8 12.Td8+ Kb7 13.Txe8 Txe8 14.dxc5 Ka8 15.Sd6 Tg8 16.De7 Txc5 17.Dxe6 Tb8 18.De7 g6 19.h4 Tc6 20.a3 Th8 21.Dd7 Tc5 22.Dg7 Td8 23.Sf7 Tc7 24.Dxg6 Tf8 25.Sd6 Tc6 26.Dg3 d4 27.cxd4 Td8 28.Sf5 Tc2 29.Db3 Td2 30.Dc3 Td1+ 31.Kh2
+- (2.44) Tiefe: 35/68 00:17:31 24163MN

As for your latest SF- output- variant with 1.Qxd2, according to my .pgn in the posting you answered to, 2...Rfc8(?!) isn't quite as strong as 2...Qe7(!) I guess.
Have a look at komodo's end- evals in the .pgn after the one and after the other one continuation

[peter]

As for your latest SF- output- variant with 1.Qxd2, according to my .pgn in the posting you answered to, 2...Rfc8(?!) isn't quite as strong as 2...Qe7(!) I guess.

And here is SF after the Resolving of the variants in the .pgn in my first posting in 2 MV mode.

1.Ng4 is up from the latest ply of th Backward at once (given 1.Qxd2 as second and last one Backward- variant even).

Here is depth 35 (reached with already full hash in 10 minutes):

[d]r4rk1/pb3pp1/1p1qpR2/2ppN3/3P4/2P4P/PP1pQ1P1/3R2K1 w - - 0 1

Analysis by Stockfish 240116 64 POPCNT:

1. +- (2.33): 1.Ng4 Rfc8 2.Qxd2 Kf8 3.Rxf7+ Kxf7 4.Rf1+ Ke8 5.Qg5 Ba6 6.Rf2 Kd7 7.Ne5+ Kc7 8.Rf7+ Kb8 9.Rd7 Qxd7 10.Nxd7+ Kb7 11.Qe7 Rc7 12.Qxe6 Bb5 13.Ne5 Rac8 14.Nf7 Ka6 15.Nd6 Rd8 16.dxc5 Rxc5 17.Nxb5 Kxb5 18.Qe7 Rdc8 19.Qxg7 R8c7 20.Qe5 Rd7 21.h4 d4 22.Qe8 Rcc7 23.cxd4 Kc4 24.Qe4 Re7

2. +/- (1.37): 1.Qxd2 Ba6 2.Ng4 Qe7 3.Qf2 Bc4 4.dxc5 bxc5 5.b3 Ba6 6.Qh4 Be2 7.Nh6+ gxh6 8.Rd2 Bh5 9.Rdf2 Kg7 10.Rg6+ Bxg6 11.Qxe7 c4 12.bxc4 dxc4 13.g4 Rfe8 14.Qf6+ Kh7 15.h4 Rg8 16.g5 hxg5 17.h5 Bf5 18.Qxf7+ Rg7 19.Qf6 Rb7 20.Rxf5 exf5 21.Qxf5+ Kg7 22.Qxg5+ Kh7 23.Qf5+ Kh8


I'll let it run for some time longer and come back, if there is some relevant change, but I don't think so, because SF now (as the important difference to the 2MV-output of your latest trial) "knows" already 2...Qe7! instead of 2...Rfc8?! following 1.Qxd2 (!?)

As for this better chance as Black's second move, that Black has not quite as well after 1.Ng4!, this best move has the edge over the probably yet also the full point winning 1.Qxd2!?, which therefore is a winner to me too, but only second best.

[MMoisan]

...


As for your latest SF- output- variant with 1.Qxd2, according to my .pgn in the posting you answered to, 2...Rfc8(?!) isn't quite as strong as 2...Qe7(!) I guess.
Have a look at komodo's end- evals in the .pgn after the one and after the other one continuation

It seems that 21. Re1! is much stronger than 21.Qf2 after 20...Qe7 :

Code:
1: Carlsen,M - Tomashevsky,E, 78th Tata Steel GpA 2016
r4rk1/p3qpp1/bp2pR2/2pp4/3P2N1/2P4P/PP1Q2P1/3R2K1 w - - 0 1

Analysis by Houdini 3 x64 (after 20...Qe7):

...

21.Te1 Tfc8 22.Te5 Rf8 23.Th5 Dxf6 24.Cxf6 gxf6 25.dxc5 Re7 26.cxb6 axb6 27.Th6 Tc6 28.Dd4 e5 29.Dxd5 Fb7 30.Th7 Tf8 31.h4 Fa8 32.Db3 Td6 33.h5 Fd5 34.c4 Fe6 35.Db4 Tfd8 36.Rh2 Ff5 37.Tg7 Th8 38.g4 Fe6 39.Rg3 f5 40.Dc3 fxg4 41.Dxe5 Td3+ 42.Rg2
+- (3.40) Profondeur: 23/70 00:09:36 2952mN

[peter]

That seems to be a point for you and Houdini, Marc.

Even if in 2MV SF still was at

r4rk1/pb3pp1/1p1qpR2/2ppN3/3P4/2P4P/PP1pQ1P1/3R2K1 w - - 0 1

Analysis by Stockfish 240116 64 POPCNT:

1. +- (2.72): 1.Ng4 Rfc8 2.Qxd2 Kf8 3.Rxf7+ Kxf7 4.Rf1+ Ke8 5.Qg5 Ba6 6.Rf2 Kd7 7.Ne5+ Kc7 8.Rf7+ Kb8 9.Rd7 Qxd7 10.Nxd7+ Kb7 11.Qe7 Rc7 12.Qxe6 Bb5 13.Ne5 cxd4 14.Qxd5+ Bc6 15.Qxd4 Rh8 16.b4 Rh6 17.c4 a6 18.c5 b5 19.Nxc6 Rhxc6 20.Qe4 Kc8 21.g4 Kd7 22.Qh7 Kc8 23.Qf5+ Kb7 24.Qe5 Rf7 25.Qe4 Rff6 26.Qe7+ Rc7 27.Qe5 Rfc6 28.Kg2 Rd7 29.Qe4 Kc7 30.a3 g6 31.Kg3 Rd8 32.h4 Kd7 33.h5 gxh5 34.gxh5 Rf6 35.Kg4 Rg8+ 36.Kh4

2. +/- (1.36): 1.Qxd2 Ba6 2.Ng4 Qe7 3.Qf2 Rac8 4.Qh4 Be2 5.Nh6+ gxh6 6.Rd2 Bh5 7.Rdf2 Kg7 8.dxc5 bxc5 9.Rg6+ Bxg6 10.Qxe7 a5 11.g4 Rg8 12.Kh2 Rh8 13.h4 h5 14.Qf6+ Kh7 15.gxh5 Bxh5 16.Qe5 Bg6 17.h5 Bf5 18.Rxf5 exf5 19.Qxd5 Rhg8 20.Qxf5+ Kh6 21.Kh3 Rce8 22.Kh4 Re2 23.Qf6+ Kh7 24.Qxf7+ Rg7 25.Qf5+ Kh8 26.Kh3 Rxb2 27.h6 Rg1

in depth 42, just after jumping to the second move after 1.Qxd2 2...Qe7 SF comes up with 3.Re1 on ist own:

r4rk1/p3qpp1/bp2pR2/2pp4/3P2N1/2P4P/PP1Q2P1/3R2K1 w - - 0 1

Analysis by Stockfish 240116 64 POPCNT:

3.Re1 Rfd8 4.Re5 Qxf6 5.Nxf6+ gxf6 6.Rh5 Kf8 7.Qf4 Ke7 8.Rh6 Rd7 9.Qxf6+ Kd6 10.Qf4+ Ke7 11.Qg5+ Kd6 12.Rf6 Rf8 13.Qg7 Rc8 14.Rxf7 Rxf7 15.Qxf7 Rc7 16.Qf4+ Kc6 17.Qg4 Kd7 18.dxc5 bxc5 19.Qa4+ Rc6 20.h4 e5 21.h5 d4 22.h6 Bd3 23.cxd4 exd4 24.Qb3 c4 25.Qb7+ Rc7 26.Qd5+ Ke8 27.Qxd4 Re7 28.Kf2
+- (4.20) Depth: 34/64 00:01:30 2139MN

So we have at the position of interest (very fine one, thanks Marc!) after new Backward, this time given in 1.Ng4 after 1.Qxd2:


[d]r4rk1/pb3pp1/1p1qpR2/2ppN3/3P4/2P4P/PP1pQ1P1/3R2K1 w - - 0 1

Analysis by Stockfish 240116 64 POPCNT:

1. +- (3.24): 1.Qxd2 cxd4 2.Qg5 dxc3 3.bxc3 Qe7 4.Rdf1 d4 5.R1f4 Be4 6.Rg4 Bg6 7.Rxg6 Qxg5 8.R6xg5 g6 9.Rxd4 Rfd8 10.Rgg4 Rd5 11.Rxd5 exd5 12.Ra4 a5 13.Nd7 b5 14.Rd4 Rc8 15.Rxd5 Rxc3 16.Rxb5 Ra3 17.Rb2 Kg7 18.Nb6 a4 19.Rc2 g5 20.Nc4 Rd3 21.Ne5 Ra3 22.Rf2 f5 23.Nf3 g4 24.hxg4 fxg4 25.Ne5 g3 26.Rf7+ Kh6

2. +- (2.10): 1.Ng4 Rfc8 2.Qxd2 Kf8 3.Rxf7+ Kxf7 4.Rf1+ Ke8 5.Qg5 Ba6 6.Rf3 Kd7 7.Ne5+ Kc7 8.Rf7+ Kb8 9.Rd7 Qxd7 10.Nxd7+ Kb7 11.Qe7 Rc7 12.Qxe6 Rd8 13.Qxd5+ Kc8 14.Nxb6+ axb6 15.Qa8+ Kd7 16.Qxa6 cxd4 17.cxd4 Kc6 18.Qd3 Rcd7 19.Qg6+ Kb7 20.Kh2 Rxd4 21.Qxg7+ R8d7 22.Qf6 R7d6 23.Qf3+ R6d5 24.Qf7+ Rd7 25.Qe6 R7d6 26.Qe3 R6d5 27.Qf2 Ka6 28.Qf3 Ka7 29.Qf6 Rd6 30.Qe7+ Rd7 31.Qe5

Depth 36/67

Best move 1.Qxd2!, second best 1.Ng4!?

Here is the corrected .pgn:

[Event "Tata Steel 2016, CCC"]
[Site "?"]
[Date "2016.01.25"]
[Round "?"]
[White "Carlsen"]
[Black "Tomashevsky"]
[Result "1-0"]
[Annotator "Martan,Peter"]
[SetUp "1"]
[FEN "r4rk1/pb3pp1/1p1qpR2/2ppN3/3P4/2P4P/PP1pQ1P1/3R2K1 w - - 0 1"]
[PlyCount "11"]

1. Qxd2 $1 (1. Rxd2 $6 Qd8 2. Rf4 Qg5 3. Rg4 Qh6 4. dxc5 bxc5 5. Qf2 d4 6. cxd4 ) (1. Ng4 $5 Rfc8 (1... cxd4 2. Rdf1 Qc5 3. Nh6+ {[%eval 228,28]}) (1... Rfe8 2. Qxd2 {[%eval 278,29]}) 2. Qxd2 Kf8 (2... Ba6 3. Qg5 {[%eval 265,27]}) 3. Rxf7+ Kxf7 4. Rf1+ Ke8 5. Qg5 Ba6 6. Rf2 {[%eval 210,30]}) (1. Rf4 $6 f6 2. Ng6 Rfe8 3. Qh5 e5 4. dxe5 Rxe5 5. Nxe5 Qxe5 6. Qxe5 fxe5 {[%eval 29,27]}) 1... Ba6 (1... Qe7 2. Rdf1) (1... cxd4 $2 2. Qg5 {[%eval 284,28]}) 2. Ng4 Rfc8 (2... Rfb8 $6 3. Qg5 Kf8 4. Rf2 Rb7 5. Qg6 cxd4 6. cxd4 Rc8 7. Qh7 {[%eval 339,31]}) (2... Qe7 $6 3. Re1 (3. Qf2 $6 Bb5 (3... Bc4 $2 4. dxc5 bxc5 5. b3 Ba6 6. Qh4 { [%eval 174,27]}) (3... Rfd8 4. Qg3) (3... Rac8 $6 4. Qh4 Be2 5. Nh6+ {[%eval 132,30]}) (3... Rfc8 $2 4. Qg3 {[%eval 161,27]}) 4. dxc5 bxc5 (4... Rfc8 $2 5. Rd4 {[%eval 281,27]}) (4... Rfd8 $2 5. Qh4 {[%eval 186,28]}) 5. Qh4 Be2 6. Nh6+ {[%eval 127,31]}) 3... gxf6 (3... Rfd8 4. Re5 Qxf6 5. Nxf6+ gxf6 6. Rh5 Kf8 7. Qf4 {[%eval 458,27]}) 4. Qh6 Bd3 5. Nxf6+ Qxf6 6. Qxf6 Kh7 7. dxc5 {[%eval 427, 27]}) 3. Qg5 Kf8 (3... Qe7 4. Rd2 Rc7 5. Rdf2 cxd4 6. Nh6+) 4. Rf2 Rc7 5. Qg6 cxd4 (5... Bb5 6. Re1 {[%eval 365,33]}) 6. cxd4 {[%eval 350,30]} 1-0

BTW SF dev. has become noticable better in Resolving such since SF7 on many cores as for my pov

[peter]

Sting prefers Qxd2 rather quickly, but I fear not quite for the right reasons.

r4rk1/pb3pp1/1p1qpR2/2ppN3/3P4/2P4P/PP1pQ1P1/3R2K1 w - - 0 1

Analysis by Sting SF 6 64bit:
...
1.Txd2
= (0.08 ++) Tiefe: 29 00:01:09 896MN
...
1.Txd2 Dd8 2.Tf4 Dg5 3.Tg4 Dh6 4.dxc5 bxc5 5.Df2 f6 6.Sd7 Tf7 7.Sxc5 Te8 8.Th4 Dg6 9.Te2 e5 10.Tg4 Dh6 11.Th4 Dg6
= (0.00) Tiefe: 29 00:02:04 1751MN
1.Dxd2
= (0.08 ++) Tiefe: 30 00:03:18 2853MN
...
1.Dxd2 La6 2.Sg4 De7 3.Df2 Lc4 4.b3 La6 5.Sh6+ Kh7 6.Txf7 Txf7 7.Sxf7 Df6 8.Se5 cxd4 9.cxd4 Tc8 10.Dd2 Df5 11.Tc1 Txc1+ 12.Dxc1 Ld3 13.Dd2 Le4 14.Kh2 Df1 15.Kg3 Db1 16.b4 Lc2 17.Sf7 Kg6 18.Sg5
+/= (0.44) Tiefe: 30 00:04:17 3757MN
...
1.Dxd2 La6 2.Sg4 De7 3.Df2 Tac8 4.Sh6+ Kh7 5.Txf7 Txf7 6.Sxf7 Df6 7.Se5 cxd4 8.Dxd4 Tf8 9.Te1 Kg8 10.Sf3 Lc8 11.De3 Df4 12.Dxf4 Txf4 13.Sg5 Tf8 14.h4 Ld7 15.a3 Te8 16.Kf2 e5 17.Sf3 e4 18.Sd4 Kf7 19.Ke3 Ke7 20.Tf1 Th8 21.g3 g6 22.Tf2
+/= (0.44) Tiefe: 34 00:10:36 9775MN
1.Dxd2 La6 2.Sg4 De7 3.Df2 Tac8 4.Sh6+ Kh7 5.Txf7 Txf7 6.Sxf7 Df6 7.Se5 cxd4 8.Dxd4 Tf8 9.De3 Df4 10.Dxf4 Txf4 11.Te1 Lc8 12.g3 Tf6 13.a4 Kg8 14.Tf1 Kf8 15.Tf4 Ke7 16.Sc6+ Kf7 17.b3 Txf4 18.gxf4 Kf6
+/= (0.44) Tiefe: 35 00:14:19 13460MN

After 2...Qe7 3.Qf2 stays in output- line, so I suspect 1.Ng4 to get better evals if "found" too. The eval isn't a winning one for the favored move.

And so indeed, being "shown" 1.Ng4..., Sting changes its mind:

r4rk1/pb3pp1/1p1qpR2/2ppN3/3P4/2P4P/PP1pQ1P1/3R2K1 w - - 0 1

Analysis by Sting SF 6 64bit:
...
1.Sg4
= (0.08 ++) Tiefe: 19 00:00:03 6642kN
...
1.Sg4 Tfc8 2.Dxd2 Tc7 3.Dg5 Kf8 4.Tf2 cxd4 5.cxd4 De7 6.Dg6 f5 7.Te1 Df7 8.Txe6 Dxg6 9.Txg6 Lc8 10.Se3 Tf7 11.Td6 Ld7 12.Sxd5 Te8 13.Sf4 Ke7 14.Tg6 Kf8 15.d5 Te1+ 16.Tf1 Txf1+ 17.Kxf1 Te7 18.Kf2 Kf7 19.Td6 Lb5 20.Te6 Lc4
+/- (1.49) Tiefe: 30 00:01:48 1390MN


The position is a damned good one to test engines' Forward- Backward- solving abilities.

[peter]

This SF- derivative from "Bullet" does even better, I'd say.

With empty hash again:

r4rk1/pb3pp1/1p1qpR2/2ppN3/3P4/2P4P/PP1pQ1P1/3R2K1 w - - 0 1

Analysis by Stockfish 251215bt 64 SSE42:
...
1.Tf2 De7 2.Dh5 g6 3.Dh6 f6 4.Sxg6 Dh7 5.Dxh7+ Kxh7 6.Sxf8+ Txf8 7.Tfxd2 Kg6 8.Kf2 Kf7 9.Te1 cxd4 10.Txd4 e5 11.Tdd1 Tg8
+- (1.85) Tiefe: 13/24 00:00:18 166MN
1.Dxd2 La6 2.Tf4 f6 3.Sg6 f5 4.Df2 Tf6 5.Se5 De7 6.Th4 Th6 7.Txh6 gxh6
+- (1.85) Tiefe: 14/24 00:00:21 247MN
...
1.Dxd2 La6 2.Sg4 De7 3.Df2 Tfc8 4.Kh1 gxf6 5.Sxf6+ Kf8 6.Dh4 Dd6 7.dxc5 bxc5 8.Dh6+ Ke7 9.Dg5 Kf8 10.b3 Tab8
+- (1.85) Tiefe: 15/24 00:00:31 547MN
...
1.Dxd2 La6 2.Sg4 De7 3.Df2 Tac8 4.Dh4 Le2 5.Sh6+ gxh6 6.Td2 Lh5 7.Tdf2 Kg7 8.g4 Tg8 9.dxc5 Txc5 10.Tf1 Tc7
+- (1.90) Tiefe: 16/30 00:00:49 1079MN
...
1.Dxd2 La6 2.Sg4 De7 3.Df2 Tac8 4.Dh4 Le2 5.Sh6+ gxh6 6.Td2 Lh5 7.Tdf2 Kg7 8.dxc5 bxc5 9.Tg6+ Lxg6 10.Dxe7 a5 11.h4 Kh7 12.g4 Tg8 13.g5 a4 14.Df6 Ta8 15.h5
+- (1.76) Tiefe: 19/35 00:02:28 4092MN
...
1.Dxd2 cxd4 2.Dg5 De7 3.Tdf1 dxc3 4.bxc3 d4 5.T1f4 Le4 6.Tg4 Lg6 7.Txg6 Dxg5 8.T6xg5 g6 9.Txd4 Tac8 10.Td7 Txc3 11.Txa7 Tc5 12.h4 Ta5 13.Txa5 bxa5 14.Sc6 Kg7 15.Txa5 Th8 16.g3 Kf6 17.Kf2 Tc8 18.Se5
+- (2.94) Tiefe: 22/44 00:08:28 14914MN
...
1.Dxd2 cxd4 2.Dg5 De7 3.Tdf1 dxc3 4.bxc3 d4 5.T1f4 Le4 6.Tg4 Lg6 7.Txg6 Dxg5 8.T6xg5 g6 9.Txd4 Tac8 10.Td7 a6 11.Tg3 Tcd8 12.Tgd3 Txd7 13.Sxd7 Tc8 14.Sxb6 Tc6 15.Sd7 Kg7 16.Kf2 a5 17.Ke3 Tc7 18.Kd2 f6 19.Ke3 Kf7 20.Sb6 Tc6 21.Sd7
+- (3.07) Tiefe: 25/46 00:30:39 52928MN