Under Promotion Game

[Howard E]

Game 1 of the Silver Opening Suite.

I was watching this game when Doch's eval jumped up on move 17, only to lose because it missed Spark's Knight underpromotion. 40/10 repeating,i7-920,4 threads(virtual) for Spark , 512mb hash,

[Event "DochSpark"]
[Site "HOWARD-PC"]
[Date "2010.01.10"]
[Round "1"]
[White "Doch64-134-ja"]
[Black "Spark-0.3a-vc64-mp"]
[Result "0-1"]
[BlackElo "3000"]
[Time "21:09:17"]
[WhiteElo "2900"]
[TimeControl "40/600:40/600:40/600"]
[SetUp "1"]
[FEN "1rbq1rk1/1pp2pbp/p1np1np1/4p3/2PPP3/2N1BP2/PP1Q2PP/R1N1KB1R w KQ e6 0 1"]
[Termination "adjudication"]
[PlyCount "61"]
[WhiteType "program"]
[BlackType "program"]
[Comment "; id \"Silver Suite - King's Indian, Saemisch : ECO E84\";"]

1. d5 {+0.43/17 17} Nd4 {-0.56/20 19} 2. Nb3 {+0.57/17 18} Nxb3 {-0.62/19
26} 3. axb3 {+0.64/14 1} Nh5 {-0.46/20 9} 4. O-O-O {+0.67/17 17} f5
{-0.04/19 15} 5. exf5 {+0.62/16 9} gxf5 {0.00/19 11} 6. Kb1 {+0.51/18 53}
b5 {+0.09/19 10} 7. Bd3 {+0.49/17 16} Bd7 {+0.05/17 16} 8. Rhe1 {+0.48/16
33} bxc4 {+0.03/16 18} 9. Bxc4 {+0.37/16 32} f4 {+0.46/17 17} 10. Bf2
{+0.34/17 9} Bf5+ {+0.41/17 9} 11. Ne4 {+0.27/16 8} Qg5 {+0.49/18 8} 12. g4
{+0.31/17 9} Bxe4+ {-0.05/21 20} 13. Rxe4 {+0.05/18 10} Nf6 {-0.05/21 18}
14. Ree1 {+0.01/17 10} Qg6+ {-0.03/19 17} 15. Qc2 {+0.41/17 12} Qxc2+
{-0.17/19 8} 16. Kxc2 {+0.73/14 0} e4 {-0.32/22 23} 17. g5 {+1.80/18 8} e3
{-0.10/24 10} 18. gxf6 {+1.59/18 8} exf2 {-0.07/23 8} 19. fxg7 {+4.12/12 0}
fxe1=N+ {+2.10/23 9} 20. Rxe1 {-1.44/16 1} Rfe8 {+2.35/24 10} 21. Rg1
{-1.67/19 10} Re7 {+2.50/24 16} 22. Rg2 {-1.92/21 27} a5 {+2.98/23 10} 23.
Bd3 {-2.17/22 12} Rxg7 {+3.06/26 12} 24. Re2 {-2.16/22 8} h5 {+3.32/25 12}
25. Kc3 {-2.68/20 14} h4 {+3.30/24 8} 26. Re4 {-2.75/20 10} Rg2 {+3.76/24
12} 27. Rxf4 {-2.75/20 9} Rb4 {+4.08/25 19} 28. Bc4 {-2.92/20 9} Rxh2
{+4.41/24 11} 29. Rg4+ {-3.02/21 10} Kf7 {+4.68/24 16} 30. Kd4 {-3.20/19 9}
h3 {+5.07/21 7} 31. Rf4+ {-4.20/20 11} ... {0-1 Arena Adjudication} 0-1

[Vinvin]

I suppose that means that Doch 1.3.4 doesn't support underpromotion in the search

[rvida]

I suppose that means that Doch 1.3.4 doesn't support underpromotion in the search

It is possible that underpromotions are sorted late in the movelist and are reduced much more than queen promotions and/or quiet moves. Maybe Doch requires more depth to see them.

Richard

[Howard E]

Can others confirm this?

[D] 1r3rk1/2p3Pp/p2p4/3P4/2B2p2/1P3P2/1PK2p1P/3RR3 b - -

Doch64-134-ja:
16 00:01 2.847.667 1.584.920 -4.18 Kxg7 Re2 Rfe8 Rxf2 Kf6 Ra1 Rb6 Rxa6 Rxa6 Bxa6 Re3 Bd3 Ke5 Be4 Kd4 Rd2+ Kc5 h3 h6 Rd3 Re2+ Kc3 Re3 b4+ Kb5 Bf5
17 00:02 4.170.504 1.570.202 -4.21 Kxg7 Re2 Rfe8 Rxf2 Kf6 Ra1 Rb6 Rxa6 Rxa6 Bxa6 Re3 Bd3 Ke5 Be4 h5 h4 Kd4 Rg2 Kc5 Rg7 Kb4 Rxc7 Re2+ Kd3 Rxb2 Rb7+ Kc5 Kc3
17 00:03 4.764.310 1.571.862 -4.21 Kxg7 Re2 Rfe8 Rxf2 Kf6 Ra1 Rb6 Rxa6 Rxa6 Bxa6 Re3 Bd3 Ke5 Be4 h5 h4 Kd4 Rg2 Kc5 Rg7 Kb4 Rxc7 Re2+ Kd3 Rxb2 Rb7+ Kc5 Kc3
18 00:03 5.862.052 1.563.343 -4.19 Kxg7 Re2 Rfe8 Rxf2 Kf6 Ra1 Rb6 Rxa6 Rxa6 Bxa6 Re3 Bd3 Ke5 Be4 h6 Rg2 Kd4 h4 Kc5 Rg7 Re2+ Kc1 Kb4 Rxc7 Kxb3 Bd3 Re1+ Kd2
18 00:04 6.876.565 1.566.323 -4.19 Kxg7 Re2 Rfe8 Rxf2 Kf6 Ra1 Rb6 Rxa6 Rxa6 Bxa6 Re3 Bd3 Ke5 Be4 h6 Rg2 Kd4 h4 Kc5 Rg7 Re2+ Kc1 Kb4 Rxc7 Kxb3 Bd3 Re1+ Kd2
19 00:07 11.990.477 1.528.794 -4.28 Kxg7 Re2 Rfe8 Rxf2 Kf6 Ra1 Rb6 Rxa6 Rxa6 Bxa6 Re3 Bd3 Ke5 Be4 h6 Rg2 Kd4 h4 Re1 Rg4 Re2+ Kb1 Re3 Ka2 Rxe4 fxe4 Kxe4 Ka3 h5
19 00:08 13.788.704 1.537.544 -4.28 Kxg7 Re2 Rfe8 Rxf2 Kf6 Ra1 Rb6 Rxa6 Rxa6 Bxa6 Re3 Bd3 Ke5 Be4 h6 Rg2 Kd4 h4 Re1 Rg4 Re2+ Kb1 Re3 Ka2 Rxe4 fxe4 Kxe4 Ka3 h5
20 00:11 18.061.987 1.517.143 -4.23 Kxg7 Re2 Rfe8 Rxf2 Kf6 Ra1 Rb6 Rxa6 Rxa6 Bxa6 Re3 Bd3 Ke5 Be4 Kd4 Rd2+ Kc5 h3 h5 Rd3 Rxd3 Kxd3 Kb4 Kc2 h4 Bf5 Kc5 Be6 Kb4 Bd7 Kc5 Bc6 Kd4
20 00:13 20.002.671 1.531.432 -4.23 Kxg7 Re2 Rfe8 Rxf2 Kf6 Ra1 Rb6 Rxa6 Rxa6 Bxa6 Re3 Bd3 Ke5 Be4 Kd4 Rd2+ Kc5 h3 h5 Rd3 Rxd3 Kxd3 Kb4 Kc2 h4 Bf5 Kc5 Be6 Kb4 Bd7 Kc5 Bc6 Kd4
21 00:17 27.528.687 1.551.041 -4.23 Kxg7 Re2 Rfe8 Rxf2 Kf6 Ra1 Rb6 Rxa6 Rxa6 Bxa6 Re3 Bd3 Ke5 Be4 Kd4 Rd2+ Kc5 h3 h5 Rd3 Rxd3 Kxd3 Kb4 Kc2 h4 Bf5 Kc5 Be6 Kb4 Bd7 Kc5 Bc6 Kd4 b4 Kc4 b5
21 00:19 30.598.261 1.563.010 -4.23 Kxg7 Re2 Rfe8 Rxf2 Kf6 Ra1 Rb6 Rxa6 Rxa6 Bxa6 Re3 Bd3 Ke5 Be4 Kd4 Rd2+ Kc5 h3 h5 Rd3 Rxd3 Kxd3 Kb4 Kc2 h4 Bf5 Kc5 Be6 Kb4 Bd7 Kc5 Bc6 Kd4 b4 Kc4 b5
22 00:34 53.706.890 1.547.042 -4.30 Kxg7 Re2 Rfe8 Rxf2 Kf6 Ra1 Rb6 Rxa6 Rxa6 Bxa6 Re3 Bd3 Ke5 Be4 Kd4 Rd2+ Kc5 h3 h6 Rd3 Re2+ Kc3 Rh2 b4+ Kb6 Bf5 Re2 Kb3 Re5 Bd7 Re7 Bc6 Re2 Rc3
22 00:38 60.407.493 1.551.525 -4.30 Kxg7 Re2 Rfe8 Rxf2 Kf6 Ra1 Rb6 Rxa6 Rxa6 Bxa6 Re3 Bd3 Ke5 Be4 Kd4 Rd2+ Kc5 h3 h6 Rd3 Re2+ Kc3 Rh2 b4+ Kb6 Bf5 Re2 Kb3 Re5 Bd7 Re7 Bc6 Re2 Rc3
23 01:00 95.505.181 1.576.693 -4.21 Kxg7 Re2 Rfe8 Rxf2 Kf6 Ra1 Re3 Rxa6 Ke5 Ra7 Rc8 h4 h5 Rf1 Kd4 Kb1 Kc5 Ra6 Kd4 Rc6 Rce8 Rxc7 Re1+ Rxe1 Rxe1+ Ka2 Rh1 Rc6 Ke3 Rxd6 Rxh4 Ra6 Kxf3 d6
23 01:04 101.400.481 1.583.741 -4.21 Kxg7 Re2 Rfe8 Rxf2 Kf6 Ra1 Re3 Rxa6 Ke5 Ra7 Rc8 h4 h5 Rf1 Kd4 Kb1 Kc5 Ra6 Kd4 Rc6 Rce8 Rxc7 Re1+ Rxe1 Rxe1+ Ka2 Rh1 Rc6 Ke3 Rxd6 Rxh4 Ra6 Kxf3 d6

[OliverBr]

OliThink finds the underpromotion at every sd:

Code: Select all

 ./olithink64 -sd 16 "1r3rk1/2p3Pp/p2p4/3P4/2B2p2/1P3P2/1PK2p1P/3RR3 b - - "
go
 1   123      0        70  f2e1n d1e1
 2   123      0       314  f2e1n d1e1 g8g7
 3   107      0      1287  f2e1n d1e1 g8g7 e1g1 g7f6
 4   109      0      3914  f2e1n d1e1 f8f7 h2h4 f7g7
 5   108      0     10306  f2e1n d1e1 g8g7 c4a6 f8e8 e1e6
 6   122      1     21580  f2e1n d1e1 f8f7 e1g1 f7g7 g1g7 g8g7 c4a6
 7   113      2     40236  f2e1n d1e1 f8f7 c4a6 f7g7 a6f1 g7g5 f1c4
 8   111      5     82859  f2e1n d1e1 g8g7 c4a6 f8e8 e1e8 b8e8 a6d3 e8e1
 9   109     12    185584  f2e1n d1e1 f8f7 c4a6 f7g7 a6d3 g7g5 e1e7 g5g2 c2c3 h7h5
10   119     25    374780  f2e1n d1e1 g8g7 c4a6 f8e8 e1e8 b8e8 a6d3 e8e3 b3b4 e3f3
11   127     43    634766  f2e1n d1e1 g8g7 c4a6 f8e8 e1g1 g7f6 g1g4 f6e5 g4g7 e5d5 g7c7 e8e3
12   127     72   1050807  f2e1n d1e1 g8g7 c4a6 f8e8 e1g1 g7h8 a6d3 e8e3 g1f1 e3e5 d3e4 e5h5 f1h1
13   135    127   2101497  f2e1n d1e1 g8g7 c4a6 f8e8 e1g1 g7h8 a6d3 e8e3 g1f1 e3e5 d3e4 e5h5 f1h1 h5g5
14   148    268   4904475  f2e1n d1e1 g8g7 c4a6 f8e8 e1g1 g7h8 a6d3 e8e3 g1g5 e3f3 g5h5 h8g8 d3h7 g8f7 h7d3 f3f2 c2c3
15   132    461   9005451  f2e1n d1e1 g8g7 c4a6 f8e8 e1g1 g7h8 a6d3 e8e3 g1g5 b8g8 g5h5 g8g2 c2c3 e3f3 c3c4 h8g8 d3h7 g8f8
16   147   1167  23630711  f2e1n d1e1 g8g7 c4a6 f8e8 e1g1 g7h8 a6c4 e8e3 g1f1 b8a8 h2h4 a8g8 b3b4 g8g2 c2b1 g2h2 f1g1 h2h4
1. ... f2e1n

kibitz W: 147 Nodes: 18363430 QNodes: 5267281 Evals: 10939029 cs: 1167 knps: 2024

[P. Villanueva]

I suppose that means that Doch 1.3.4 doesn't support underpromotion in the search

It is possible that underpromotions are sorted late in the movelist and are reduced much more than queen promotions and/or quiet moves. Maybe Doch requires more depth to see them.

Richard

That is not posible. Doch should find the underpromotion in the first iteration, (where there are not reductions). It looks like the move generator skips underpromotions.