Komodo 1959.00 - Houdini 6.02

Guenther · Post by **Guenther** » Mon Nov 13, 2017 5:30 pm

peter wrote:After all that said, Guenther, it still doesn't make the position before 52.d5 so easy for many engines to see the move and the winning eval with empty hash at short TC, so I still think it can be used as a test position too. Not finding 52.d5 still can give the half point away,

Nothing to say against, Peter.
I insist though that the avoid move 48...Bxb7? should be a good test position too :)

Guenther

Uri · Post by **Uri** » Mon Nov 13, 2017 5:58 pm

Many times I get the impression that humans will always be better than computers at chess.

Computers are just too stupid to understand the game in a very deep way while the best humans can.

leavenfish · Post by **leavenfish** » Mon Nov 13, 2017 6:02 pm

peter wrote:After all that said, Günther, it still doesn't make the position before 52.d5 so easy for many engines to see the move and the winning eval with empty hash at short TC, so I still think it can be used as a test position too. Not finding 52.d5 still can give the half point away,

Let me throw something out there.

During another game the same day (and mainly because I am a subscriber who would sure like to get his hands on these versions!), I asked Mark about the current developmental version they are working on - what the improvements were. "speedups?" I asked. No, he said, and one of a couple of things he mentioned was...improvements in the handling of passed pawns.

The irony. Lets hope whatever change this is, it would not lead to Komodo missing this idea. That would indeed be a real shame.

peter · Post by **peter** » Mon Nov 13, 2017 6:08 pm

Guenther wrote:
peter wrote:After all that said, Guenther, it still doesn't make the position before 52.d5 so easy for many engines to see the move and the winning eval with empty hash at short TC, so I still think it can be used as a test position too. Not finding 52.d5 still can give the half point away,
Nothing to say against, Peter.
I insist though that the avoid move 48...Bxb7? should be a good test position too

Nothing to say against that neither, Günther!

But look at this output, that's one of the best SF- branches as for solving tactical postiions:

[d]3r2k1/pr6/1p2q2p/5R2/3P3p/8/5RP1/3Q2K1 w - - 0 1

Analysis by SF-McBrain v3.0 TCEC-X 64 POPCNT:

52.Qg4+ Kh7 53.Qxh4 Rg8 54.Re5 Qc6 55.Qe4+ Qxe4 56.Rxe4
=/+ (-0.45) Depth: 7/9 00:00:00 67kN
...
52.Qg4+ Kh7 53.Qf3 Rg7 54.Rf6 Qd5 55.Qe3 Qg5 56.Qe4+ Kg8 57.Qe6+ Kh7 58.Qe4+
= (0.00) Depth: 20/14 00:00:04 5256kN
52.Rf6
= (0.07 ++) Depth: 21/14 00:00:07 12550kN
...
52.Rf6 Qe3 53.Qh5 Re7 54.Rf7 Qe6 55.Rxe7 Qxe7 56.Qg6+ Qg7 57.Qe6+ Kh8 58.Rf6 Qg5 59.Rg6 Qc1+ 60.Kh2 Qc7+ 61.Kh3 Qc8 62.Rxh6+ Kg7 63.Rg6+ Kh7 64.Rh6+ Kg7
= (0.00) Depth: 22/37 00:00:11 27980kN
...
52.Rf6 Qe3 53.Qh5 Re7 54.Rxh6 Qxd4 55.Rg6+ Rg7 56.Rxg7+ Kxg7 57.Qg5+ Kh7 58.Qh5+ Kg7
= (0.00) Depth: 42/15 00:30:42 50388MN

After half an hour at 24 threads with 32G hash still nothing to be seen of the winining move 52.d5, neither anything of the already reached decisive advantage of White as for the eval,

Robert Flesher · Post by **Robert Flesher** » Tue Nov 14, 2017 11:28 pm

peter wrote:
Guenther wrote:
peter wrote:After all that said, Guenther, it still doesn't make the position before 52.d5 so easy for many engines to see the move and the winning eval with empty hash at short TC, so I still think it can be used as a test position too. Not finding 52.d5 still can give the half point away,
Nothing to say against, Peter.
I insist though that the avoid move 48...Bxb7? should be a good test position too
Nothing to say against that neither, Günther!

But look at this output, that's one of the best SF- branches as for solving tactical postiions:

[d]3r2k1/pr6/1p2q2p/5R2/3P3p/8/5RP1/3Q2K1 w - - 0 1

Analysis by SF-McBrain v3.0 TCEC-X 64 POPCNT:

52.Qg4+ Kh7 53.Qxh4 Rg8 54.Re5 Qc6 55.Qe4+ Qxe4 56.Rxe4
=/+ (-0.45) Depth: 7/9 00:00:00 67kN
...
52.Qg4+ Kh7 53.Qf3 Rg7 54.Rf6 Qd5 55.Qe3 Qg5 56.Qe4+ Kg8 57.Qe6+ Kh7 58.Qe4+
= (0.00) Depth: 20/14 00:00:04 5256kN
52.Rf6
= (0.07 ++) Depth: 21/14 00:00:07 12550kN
...
52.Rf6 Qe3 53.Qh5 Re7 54.Rf7 Qe6 55.Rxe7 Qxe7 56.Qg6+ Qg7 57.Qe6+ Kh8 58.Rf6 Qg5 59.Rg6 Qc1+ 60.Kh2 Qc7+ 61.Kh3 Qc8 62.Rxh6+ Kg7 63.Rg6+ Kh7 64.Rh6+ Kg7
= (0.00) Depth: 22/37 00:00:11 27980kN
...
52.Rf6 Qe3 53.Qh5 Re7 54.Rxh6 Qxd4 55.Rg6+ Rg7 56.Rxg7+ Kxg7 57.Qg5+ Kh7 58.Qh5+ Kg7
= (0.00) Depth: 42/15 00:30:42 50388MN

After half an hour at 24 threads with 32G hash still nothing to be seen of the winining move 52.d5, neither anything of the already reached decisive advantage of White as for the eval,

Very interesting position. I had an older version of Komodo look at this position on very slow hardware.

New game
3r2k1/pr6/1p2q2p/5R2/3P3p/8/5RP1/3Q2K1 w - - 0 1

Analysis by Komodo 9.02 64-bit:

1.Qg4+
= (0.25 ++) Depth: 6 00:00:00 161kN
1.Qg4+
+/= (0.33 ++) Depth: 6 00:00:00 161kN
1.Qg4+
+/= (0.48 ++) Depth: 6 00:00:00 161kN
1.Qg4+
+/= (0.68 ++) Depth: 6 00:00:00 161kN
1.Qg4+ Kh7
+/= (0.48 --) Depth: 6 00:00:00 162kN
1.Qg4+ Kh7 2.Qxh4 Qe1+ 3.Kh2 Rg8 4.Re5 Qb1
+/= (0.48) Depth: 6 00:00:00 175kN
1.Qg4+
+/= (0.55 ++) Depth: 7 00:00:00 179kN
1.Qg4+ Kh7
+/= (0.48 --) Depth: 7 00:00:00 179kN
1.Qg4+ Kh7
+/= (0.38 --) Depth: 7 00:00:00 179kN
1.Qg4+ Kh7
= (0.23 --) Depth: 7 00:00:00 181kN
1.Qg4+ Kh7
= (0.01 --) Depth: 7 00:00:00 182kN
1.Rf6
+/= (0.34 ++) Depth: 7 00:00:00 183kN
1.Rf6 Qd5 2.Rxh6 Qxd4 3.Qh5 Rc8 4.Qf5
+/= (0.64) Depth: 7 00:00:00 184kN
1.Rf6 Qd5
+/= (0.57 --) Depth: 8 00:00:00 185kN
1.Rf6 Qd5
+/= (0.50 --) Depth: 8 00:00:00 189kN
1.Rf6 Qd5
+/= (0.40 --) Depth: 8 00:00:00 190kN
1.Rf6 Qd5
= (0.25 --) Depth: 8 00:00:00 191kN
1.Rf6 Qd5
= (0.03 --) Depth: 8 00:00:00 195kN
1.Rf6 Qe4
=/+ (-0.30 --) Depth: 8 00:00:00 201kN
1.Qg4+ Kh7 2.Qxh4 Rg8 3.Qf4 Rbg7 4.Rh5 Rd8 5.Qd2
= (0.08) Depth: 8 00:00:00 203kN
1.Qg4+ Kh7
= (0.01 --) Depth: 9 00:00:00 209kN
1.Qg4+ Kh7
= (-0.06 --) Depth: 9 00:00:00 226kN
1.Rf6 Qe4 2.Qh5 Qxd4 3.Qf5 Rg7 4.Qe6+ Kh8 5.Rxh6+ Rh7 6.Rxh7+ Kxh7 7.Qe7+ Kh8
= (0.00) Depth: 9 00:00:00 231kN
1.Rf6 Qd5 2.Qe1 Qxd4 3.Qe6+ Kh8 4.Rxh6+ Rh7 5.Rxh7+ Kxh7 6.Qf5+ Kh6 7.Qf7 Rd6
= (0.00) Depth: 10 00:00:00 261kN
1.Rf6
= (0.07 ++) Depth: 11 00:00:00 284kN
1.Rf6
= (0.14 ++) Depth: 11 00:00:00 296kN
1.Rf6 Qe3 2.Qg4+ Rg7 3.Rg6 Qxd4 4.Qe6+ Kh8 5.Rxh6+ Rh7 6.Qg6 Rxh6 7.Qxh6+ Kg8 8.Qg5+ Kh7 9.Qh5+ Kg8
= (0.08) Depth: 11 00:00:00 318kN
1.Rf6 Qe3
= (0.01 --) Depth: 12 00:00:00 343kN
1.Rf6 Qe3 2.Qg4+ Rg7 3.Rg6 Qxd4 4.Qe6+ Kh8 5.Rxh6+ Rh7 6.Rxh7+ Kxh7 7.Qf5+ Kh6 8.Qe6+ Kh7
= (0.00) Depth: 12 00:00:00 401kN
1.Rf6 Qe3
= (-0.07 --) Depth: 13 00:00:00 581kN
1.Qg4+ Kh7 2.Qxh4 Rg8 3.Rf6 Qe3 4.Rxh6+ Qxh6 5.Qe4+ Kh8 6.Qxb7 Qc1+ 7.Kh2 Qh6+ 8.Kg1
= (0.00) Depth: 13 00:00:00 686kN
1.Qg4+ Kh7 2.Qxh4 Rg8 3.Rf6 Qe3 4.Rxh6+ Qxh6 5.Qe4+ Kh8 6.Qxb7 Qc1+ 7.Kh2 Qh6+ 8.Kg1
= (0.00) Depth: 14 00:00:00 1001kN
1.Rf6
= (0.07 ++) Depth: 14 00:00:00 1048kN
1.Rf6
= (0.14 ++) Depth: 14 00:00:00 1054kN
1.Rf6
= (0.24 ++) Depth: 14 00:00:00 1071kN
1.Rf6
+/= (0.39 ++) Depth: 14 00:00:00 1236kN
1.Rf6 Qe3 2.Qh5 Re7 3.d5 Qc1+ 4.Rf1 Re1 5.Qf7+ Kh8 6.Rxh6+ Qxh6 7.Rxe1 Rf8 8.Re8 Qe3+ 9.Rxe3 Rxf7 10.Re4
+/= (0.43) Depth: 14 00:00:01 1727kN
1.Rf6 Qe3
+/= (0.36 --) Depth: 15 00:00:01 2188kN
1.Rf6 Qe3
+/= (0.29 --) Depth: 15 00:00:01 2509kN
1.Rf6 Qe3
= (0.19 --) Depth: 15 00:00:01 2649kN
1.Rf6 Qe3
= (0.04 --) Depth: 15 00:00:01 2925kN
1.Rf6 Qe3 2.Qh5 Re7 3.R6f4 Qc1+ 4.Kh2 Qc7 5.Qxh6 Rxd4 6.Kh1 Re1+ 7.Kh2 Re7
= (0.00) Depth: 15 00:00:01 3110kN
1.Rf6 Qe3 2.Qh5 Re7 3.R6f4 Qc1+ 4.Kh2 Qc7 5.Qxh6 Rxd4 6.Kh1 Re1+ 7.Kh2 Rxf4 8.Rxf4 Re7 9.Kg1 Qc1+ 10.Kh2 Qc7
= (0.00) Depth: 16 00:00:01 3447kN
1.Rf6 Qe3 2.Qh5 Re7 3.R6f4 Qc1+ 4.Kh2 Qc7 5.Qxh6 Rxd4 6.Qf8+ Kh7 7.Qf5+ Kg8 8.Qg6+ Rg7 9.Qe8+ Kh7 10.Qh5+ Kg8 11.Qe8+
= (0.00) Depth: 17 00:00:02 4084kN
1.Rf6 Qe3 2.Qh5 Re7 3.R6f4 Qc1+ 4.Kh2 Qc7 5.Qxh6 Rxd4 6.Qf8+ Kh7 7.Qf5+ Kg8 8.Qg6+ Rg7 9.Qe8+ Kh7 10.Qh5+ Kg8 11.Qe8+
= (0.00) Depth: 18 00:00:02 4923kN
1.Rf6
= (0.07 ++) Depth: 19 00:00:03 7477kN
1.Rf6
= (0.14 ++) Depth: 19 00:00:03 8117kN
1.Rf6 Qe3
= (0.07 --) Depth: 19 00:00:04 8608kN
1.Rf6 Qe3 2.Qh5 Re7 3.R6f4 Qc1+ 4.Kh2 Qc7 5.Qxh6 Rxd4 6.Qf8+ Kh7 7.Qf5+ Kg8 8.Qg6+ Rg7 9.Qe8+ Kh7 10.Qh5+ Kg8 11.Qe8+
= (0.00) Depth: 19 00:00:04 9118kN
1.Rf6 Qe3 2.Qh5 Re7 3.R6f4 Qc1+ 4.Kh2 Qc7 5.Qxh6 Rxd4 6.Qf8+ Kh7 7.Qf5+ Kg8 8.Qg6+ Rg7 9.Qe8+ Kh7 10.Qh5+ Kg8 11.Qe8+
= (0.00) Depth: 20 00:00:05 10702kN
1.Rf6 Qe3 2.Qh5 Re7 3.R6f4 Qc1+ 4.Kh2 Qc7 5.Qxh6 Rxd4 6.Qf8+ Kh7 7.Qf5+ Kg8 8.Qg6+ Rg7 9.Qe8+ Kh7 10.Qh5+ Kg8 11.Qe8+
= (0.00) Depth: 21 00:00:06 13881kN
1.Rf6 Qe3 2.Qh5 Re7 3.R6f4 Qc1+ 4.Kh2 Qc7 5.Qxh6 Rxd4 6.Qf8+ Kh7 7.Qf5+ Kg8 8.Qg6+ Rg7 9.Qe8+ Kh7 10.Qh5+ Kg8 11.Qe8+
= (0.00) Depth: 22 00:00:08 19787kN
1.Rf6 Qe3 2.Qh5 Re7 3.R6f4 Qc1+ 4.Kh2 Qc7 5.Qxh6 Rxd4 6.Qf8+ Kh7 7.Qf5+ Kg8 8.Qg6+ Rg7 9.Qe8+ Kh7 10.Qh5+ Kg8 11.Qe8+
= (0.00) Depth: 23 00:00:11 26944kN
1.d5
= (0.07 ++) Depth: 23 00:00:17 39492kN
1.d5
= (0.14 ++) Depth: 23 00:00:19 44503kN
1.d5 Qe3
= (0.07 --) Depth: 23 00:00:21 49108kN
1.d5
= (0.15 ++) Depth: 23 00:00:23 54235kN
1.d5
+/= (0.37 ++) Depth: 23 00:00:29 68720kN
1.d5 Qe3
= (0.20 --) Depth: 23 00:00:33 76961kN
1.d5 Qe3 2.d6 Rg7 3.R5f3 Qg5 4.Rf6 Kh8 5.Qd3 a5 6.Kh1 h3 7.Qxh3 Rg6 8.Rxg6 Qxg6 9.Qh4 Qg5 10.Qd4+ Kg8 11.Rf3 Rd7 12.Qxb6 Qg6 13.Qxa5 Qxd6 14.Qf5 Rg7 15.Rd3 Qg6 16.Rd8+ Kh7 17.Qf3 Qb1+ 18.Rd1
= (0.18) Depth: 23 00:00:37 87458kN
1.d5 Qe3
= (0.11 --) Depth: 24 00:00:39 91223kN
1.d5 Qe3
= (0.04 --) Depth: 24 00:00:45 105mN
1.d5 Qe3 2.d6 Rg7 3.R5f3 Qg5 4.Rf6 Kh8 5.Qd3 Re8 6.Rf8+ Rxf8 7.Rxf8+ Rg8 8.Qd4+ Kh7 9.Qe4+ Kh8 10.Rf7 Rg7 11.Rxg7 Kxg7 12.Qb7+ Kg6 13.Qe4+ Kg7
= (0.00) Depth: 24 00:00:47 109mN
1.d5
= (0.07 ++) Depth: 25 00:00:49 115mN
1.d5 Qe3
= (0.00 --) Depth: 25 00:00:52 122mN
1.d5
= (0.07 ++) Depth: 25 00:01:00 143mN
1.d5
= (0.22 ++) Depth: 25 00:01:02 148mN
1.d5
+/= (0.44 ++) Depth: 25 00:01:08 159mN
1.d5 Qe3
+/= (0.27 --) Depth: 25 00:01:10 165mN
1.d5 Qe3 2.d6 Rg7 3.R5f3 Qg5 4.Rf6 Kh8 5.Qd3 Rc8 6.Rf8+ Rg8 7.Rxc8 Rxc8 8.d7 Rc1+ 9.Rf1 Rxf1+ 10.Kxf1 Qd8 11.Qf5 Kg7 12.Ke2 a5 13.Kd3 b5 14.Qe6 Kf8 15.Kd4 a4 16.Kc5 a3 17.Kxb5 a2 18.Qxh6+ Ke7 19.Qxh4+ Kxd7 20.Qd4+ Kc8
= (0.14) Depth: 25 00:01:16 180mN
1.d5
= (0.21 ++) Depth: 26 00:01:20 188mN
1.d5
+/= (0.28 ++) Depth: 26 00:01:34 223mN
1.d5 Qe3
= (0.21 --) Depth: 26 00:01:44 247mN
1.d5
+/= (0.29 ++) Depth: 26 00:01:55 275mN

(, 14.11.2017)

S.Taylor · Post by **S.Taylor** » Wed Nov 15, 2017 12:31 am

Uri wrote:Many times I get the impression that humans will always be better than computers at chess.

Computers are just too stupid to understand the game in a very deep way while the best humans can.

I think this exact game shows a very big closing of any gap that is left, with this Komodo.

Vinvin · Post by **Vinvin** » Wed Nov 15, 2017 2:42 pm

Guenther wrote:
peter wrote:After all that said, Guenther, it still doesn't make the position before 52.d5 so easy for many engines to see the move and the winning eval with empty hash at short TC, so I still think it can be used as a test position too. Not finding 52.d5 still can give the half point away,
Nothing to say against, Peter.
I insist though that the avoid move 48...Bxb7? should be a good test position too

Guenther

I agree, after 48...Bxb7 49.axb7 the eval is very good for white.
3 best moves with McBrain 3.0 :

Code: Select all

FEN&#58; 3r2k1/pPr5/1p2pq1p/4RP2/3P3p/8/3R2P1/3Q2K1 b - - 0 49

McBrain_2017_v30_TCEC-x86-64-modern&#58;
...
 28/49	01&#58;26	 977 989k	11 261k	+6,04	49. ... h3 50.Qg4+ Rg7 51.Qxh3 exf5 52.Rxf5 Qc6 53.Rdf2 Rxb7 54.Rf6 Qc1+ 55.Rf1 Qc4 56.Rxh6 Qxd4+ 57.Kh1 Qe5 58.Rg6+ Rg7 59.Rxg7+ Qxg7 60.Qe6+ Kh8 61.Rf7 Rf8 62.Rxg7 Kxg7 63.Qe5+ Kg6 64.Qe4+ Kg7 65.Qg4+ Kf6 66.Qf3+ Kg7 67.Qg3+ Kf6 68.Qf4+ Kg7 69.Qg5+ Kf7 70.Kg1 Ke6 71.Qg6+ Ke7
 29/55	01&#58;26	 977 989k	11 261k	+3,81	49. ... Rxb7 50.fxe6 Rg7 51.Qb3 Rg3 52.Qc4 b5 53.Qxb5 Rg7 54.Qc4 Re7 55.Rf2 Qg7 56.Re4 Rb8 57.Rxh4 Qg6 58.d5 Rb1+ 59.Kh2 Rg7 60.g4 Qg5 61.Rh5 Qe3 62.Qf4 Qxf4+ 63.Rxf4 Rd1 64.Kg3 a5 65.Kf2 Rc7 66.Ke2 Rcc1 67.Rf7 Rd4 68.Ra7 Kf8 69.Rxh6 Kg8 70.Rxa5 Rc2+ 71.Kf3 Rc3+ 72.Ke2
 29/64	01&#58;26	 977 989k	11 261k	+3,15	49. ... exf5 50.Rf2 Rxb7 51.Rexf5 Qe6 52.d5 Qe3 53.d6 Rg7 54.R5f3 Qg5 55.Rf6 Kh8 56.Qd3 a5 57.R6f5 Qc1+ 58.Kh2 Rdg8 59.Qd4 b5 60.Rc5 Qa3 61.d7 Kh7 62.Rc8 Qg3+ 63.Kg1 Qg5 64.Qd3+ Qg6 65.d8Q Rxd8 66.Qxd8 Qb1+ 67.Kh2 Qe4 68.Qh8+ Kg6 69.Kg1 Qd4 70.Rc6+ Kg5 71.Qxh6+ Kg4

Vinvin · Post by **Vinvin** » Wed Nov 15, 2017 4:36 pm

So, 2 interesting positions :

avoid 48...Bxb7? and find 52.d5!

Code: Select all

3r2k1/pBr5/Pp2pq1p/3bRP2/3P3p/8/3R2P1/3Q2K1 b - - am Bxb7;
3r2k1/pr6/1p2q2p/5R2/3P3p/8/5RP1/3Q2K1 w - - bm d5;

Komodo 1959.00 - Houdini 6.02

Re: Komodo 1959.00 - Houdini 6.02

Re: Komodo 1959.00 - Houdini 6.02

Re: Komodo 1959.00 - Houdini 6.02

Re: Komodo 1959.00 - Houdini 6.02

Re: Komodo 1959.00 - Houdini 6.02

Re: Komodo 1959.00 - Houdini 6.02

Re: Komodo 1959.00 - Houdini 6.02

Re: Komodo 1959.00 - Houdini 6.02