Don Dailey you need to teach underpromotion to Komodo!

[pichy]

Don Dailey you should teach Underpromotion to Komodo since your engine does not realize the White to play could win if it promote to a Rook instead of a Queen, like I say before Engines without tablebase are NOT as good as Top GM or even Master with the knowledge of the famous Saavedra endgame study. Here your program think that it is a draw after depth search 33 and I decided to play against it and mated it.

[d]8/8/1KP5/3r4/8/8/8/k7 w - - 0 1
FEN: 8/8/1KP5/3r4/8/8/8/k7 w - - 0 1

Komodo3-64:
1 00:00 52 520,000 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4
2 00:00 203 203,111 +1.40 c6-c7 Ka1-b2 c7-c8Q
3 00:00 620 206,648 +1.44 c6-c7 Rd5-d2 c7-c8Q Rd2-b2+ Kb6-a5
4 00:00 1,043 260,768 +0.01 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1
5 00:00 1,278 319,523 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q
5 00:00 1,284 256,745 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q
6 00:00 1,629 271,476 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q
6 00:00 1,635 233,572 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q
7 00:00 2,290 254,422 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q
7 00:00 2,296 229,594 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q
8 00:00 3,374 281,142 +0.16 c6-c7
8 00:00 5,199 346,614 +0.63 c6-c7
8 00:00 6,804 378,007 +1.34 c6-c7
8 00:00 17,827 379,290 +1.78 c6-c7 Rd5-d3 c7-c8Q Ka1-b2 Qc8-h8+ Rd3-c3 Kb6-b5 Kb2-c2 Qh8-d4 Rc3-b3+ Kb5-c5
8 00:00 17,835 371,559 +1.78 c6-c7 Rd5-d3 c7-c8Q Ka1-b2 Qc8-h8+ Rd3-c3 Kb6-b5 Kb2-c2 Qh8-d4 Rc3-b3+ Kb5-c5
9 00:00 32,693 510,819 +1.79 c6-c7 Rd5-d3 c7-c8Q Ka1-b2 Qc8-c4 Rd3-c3 Qc4-e2+ Kb2-b3 Qe2-d2 Rc3-c4 Qd2-d1+ Kb3-a3 Qd1-e1 Ka3-b2
9 00:00 32,701 503,088 +1.79 c6-c7 Rd5-d3 c7-c8Q Ka1-b2 Qc8-c4 Rd3-c3 Qc4-e2+ Kb2-b3 Qe2-d2 Rc3-c4 Qd2-d1+ Kb3-a3 Qd1-e1 Ka3-b2
10 00:00 63,316 666,475 +1.81 c6-c7 Rd5-d3 c7-c8Q Ka1-b2 Qc8-c4 Rd3-c3 Qc4-d5 Kb2-a3 Qd5-d4 Ka3-b2 Kb6-a5 Kb2-c2 Qd4-e5
10 00:00 63,324 659,620 +1.81 c6-c7 Rd5-d3 c7-c8Q Ka1-b2 Qc8-c4 Rd3-c3 Qc4-d5 Kb2-a3 Qd5-d4 Ka3-b2 Kb6-a5 Kb2-c2 Qd4-e5
11 00:00 122,183 749,588 +1.84 c6-c7 Rd5-d3 c7-c8Q Ka1-b2 Qc8-e6 Rd3-c3 Kb6-b5 Kb2-a3 Qe6-d6+ Ka3-b2 Qd6-d5 Kb2-a3 Qd5-d4 Rc3-b3+ Kb5-c5
11 00:00 122,191 745,069 +1.84 c6-c7 Rd5-d3 c7-c8Q Ka1-b2 Qc8-e6 Rd3-c3 Kb6-b5 Kb2-a3 Qe6-d6+ Ka3-b2 Qd6-d5 Kb2-a3 Qd5-d4 Rc3-b3+ Kb5-c5
12 00:00 310,610 949,878 +1.89 c6-c7 Rd5-d3 c7-c8Q Ka1-b2 Qc8-h8+ Kb2-b3 Qh8-e5 Rd3-c3 Kb6-b5 Rc3-d3 Qe5-f5 Kb3-c3 Qf5-c5+ Kc3-d2 Qc5-f2+ Kd2-c1 Kb5-b6 Rd3-b3+ Kb6-c5
12 00:00 310,618 947,003 +1.89 c6-c7 Rd5-d3 c7-c8Q Ka1-b2 Qc8-h8+ Kb2-b3 Qh8-e5 Rd3-c3 Kb6-b5 Rc3-d3 Qe5-f5 Kb3-c3 Qf5-c5+ Kc3-d2 Qc5-f2+ Kd2-c1 Kb5-b6 Rd3-b3+ Kb6-c5
13 00:00 375,703 973,324 +1.89 c6-c7 Rd5-d3 c7-c8Q Ka1-b2 Qc8-h8+ Kb2-b3 Qh8-e5 Rd3-c3 Kb6-b5 Rc3-d3 Qe5-f5 Kb3-c3 Qf5-c5+ Kc3-d2 Qc5-f2+ Kd2-c1 Qf2-f4+ Rd3-d2 Qf4-c4+ Rd2-c2 Qc4-d4
13 00:00 375,711 970,826 +1.89 c6-c7 Rd5-d3 c7-c8Q Ka1-b2 Qc8-h8+ Kb2-b3 Qh8-e5 Rd3-c3 Kb6-b5 Rc3-d3 Qe5-f5 Kb3-c3 Qf5-c5+ Kc3-d2 Qc5-f2+ Kd2-c1 Qf2-f4+ Rd3-d2 Qf4-c4+ Rd2-c2 Qc4-d4
14 00:00 696,868 1,035,464 +1.73 c6-c7 Rd5-d6+
14 00:00 700,879 1,036,802 +1.26 c6-c7 Rd5-d6+
14 00:00 704,063 1,035,386 +0.55 c6-c7 Rd5-d6+
14 00:00 715,763 1,037,336 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
14 00:00 715,771 1,037,348 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
15 00:00 729,837 1,039,654 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
15 00:00 729,845 1,039,665 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
16 00:00 756,853 1,046,823 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
16 00:00 756,861 1,045,386 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
17 00:00 790,306 1,060,812 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
17 00:00 790,314 1,059,401 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
18 00:00 837,605 1,076,613 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
18 00:00 837,613 1,075,239 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
19 00:00 903,764 1,102,151 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
19 00:00 903,772 1,100,817 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
20 00:00 1,017,902 1,137,320 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
20 00:00 1,017,910 1,137,329 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
21 00:01 1,163,781 1,180,305 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
21 00:01 1,163,789 1,179,118 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
22 00:01 1,395,637 1,233,984 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
22 00:01 1,395,645 1,233,991 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
23 00:01 1,719,305 1,292,710 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
23 00:01 1,719,313 1,291,745 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
24 00:01 2,177,298 1,339,052 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
24 00:01 2,177,306 1,338,232 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
25 00:02 2,906,827 1,394,830 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
25 00:02 2,906,835 1,394,164 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
26 00:02 4,088,550 1,453,965 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
26 00:02 4,088,558 1,453,451 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
27 00:04 7,522,955 1,527,193 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
27 00:04 7,522,963 1,526,885 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
28 00:08 12,686,734 1,573,060 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
28 00:08 12,686,742 1,572,866 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
29 00:15 24,621,324 1,616,845 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
29 00:15 24,621,332 1,616,739 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
30 00:32 55,173,795 1,684,130 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
30 00:32 55,173,803 1,684,130 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
31 00:54 91,575,585 1,677,823 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
31 00:54 91,575,593 1,677,792 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
32 02:37 271,414,833 1,723,651 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
32 02:37 271,414,841 1,723,640 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
33 07:23 778,049,905 1,753,849 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4
33 07:23 778,067,178 1,753,829 0.00 c6-c7 Rd5-d6+ Kb6-b5 Rd6-d5+ Kb5-b4 Rd5-d4+ Kb4-c3 Rd4-d1 Kc3-c2 Rd1-d4 c7-c8Q Rd4-c4+ Qc8xc4

====================================================================================================================

And here is the famous Saavedra ending study game that I played versus Komodo:

[Event " Rook vs Pawn-Ending"]
[Site "?"]
[Date "2012.03.29"]
[Round "?"]
[White "Pichy"]
[Black "Komodo3-64"]
[Result "1-0"]
[BlackElo "3000"]
[WhiteElo "2200"]
[SetUp "1"]
[FEN "8/8/1KP5/3r4/8/8/8/k7 w - - 0 1"]
[WhiteType "human"]
[BlackType "program"]

1. c7 Rd6+ 2. Kb5 Rd5+ 3. Kb4 Rd4+ 4. Kb3 Rd3+ 5. Kc2 Rd4 6. c8=R Rd2+ 7.
Kxd2 Kb2 8. Rc3 Kb1 9. Rb3+ Ka1 10. Kc3 Ka2 {Black resigns} 11. Kc2 Ka1
{Black resigns} 12. Ra3# *

[pichy]

In this Lasker Trap position Komodo does not have a problem promoting to a Knight, The problem only occured in the previous position that Komodo always promote to a Queen without seeing the consequence of losing the game.

[d]rnbqk1nr/ppp2ppp/8/4P3/1BP5/8/PPP1KpPP/RN1Q1BNR b kq - 0 1

FEN: rnbqk1nr/ppp2ppp/8/4P3/1BP5/8/PPP1KpPP/RN1Q1BNR b kq - 0 1

Komodo3-64:
1 00:00 411 205,416 +4.47 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2
2 00:00 1,193 198,816 +5.29 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Bc8-g4 Qd1-e1
3 00:00 2,284 207,642 +3.24 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 a7-a5 Kd2-c1 a5xb4
4 00:00 3,465 203,821 +3.58 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Qh4-f4+ Kd2-c3 Nb8-c6 Rh1xg1 Qf4xe5+
5 00:00 4,298 171,922 +3.58 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Qh4-f4+ Kd2-c3 Nb8-c6 Rh1xg1 Qf4xe5+ Kc3-b3 Nc6xb4
5 00:00 4,986 160,837 +3.58 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Qh4-f4+ Kd2-c3 Nb8-c6 Rh1xg1 Qf4xe5+ Kc3-b3 Nc6xb4
6 00:00 5,548 168,116 +3.75 f2xg1N+
6 00:00 7,250 195,942 +3.74 f2xg1N+ Ke2-e1
6 00:00 9,814 233,667 +3.87 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Nb8-c6 Kd2-c1 Nc6xb4 Rh1xg1 Bc8-f5 Qd1-d2 Nb4xc2
6 00:00 9,880 229,761 +3.87 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Nb8-c6 Kd2-c1 Nc6xb4 Rh1xg1 Bc8-f5 Qd1-d2 Nb4xc2
7 00:00 15,844 282,929 +3.71 f2xg1N+ Ke2-e1
7 00:00 21,468 330,274 +3.58 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Nb8-c6 Kd2-c1 Nc6xb4 Qd1-d2 Ng8-e7 Rh1xg1 Qh4xh2 Qd2-e3 Bc8-f5
7 00:00 21,862 331,242 +3.58 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Nb8-c6 Kd2-c1 Nc6xb4 Qd1-d2 Ng8-e7 Rh1xg1 Qh4xh2 Qd2-e3 Bc8-f5
8 00:00 25,969 346,249 +3.87 f2xg1N+
8 00:00 41,430 410,198 +3.89 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Nb8-c6 Kd2-c1 Nc6xb4 Qd1-d2 Ng8-e7 Rh1xg1 Qh4xh2 Qd2-e3 Bc8-f5 c2-c3
8 00:00 43,834 417,468 +3.89 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Nb8-c6 Kd2-c1 Nc6xb4 Qd1-d2 Ng8-e7 Rh1xg1 Qh4xh2 Qd2-e3 Bc8-f5 c2-c3
9 00:00 89,572 500,399 +4.09 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Nb8-c6 Kd2-c1 Nc6xb4 Qd1-d2 Ng8-e7 Qd2xb4 Qh4-f2 Rh1xg1 Qf2-e3+ Qb4-d2 Qe3xg1 Qd2-f4 Ne7-g6
9 00:00 92,572 505,856 +4.09 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Nb8-c6 Kd2-c1 Nc6xb4 Qd1-d2 Ng8-e7 Qd2xb4 Qh4-f2 Rh1xg1 Qf2-e3+ Qb4-d2 Qe3xg1 Qd2-f4 Ne7-g6
10 00:00 194,575 557,520 +3.86 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Nb8-c6 Kd2-c1 Nc6xb4 Qd1-d2 Ng8-e7 a2-a3 Nb4xc2 Kc1xc2 Bc8-f5+ Kc2-b3 Ra8-d8 Qd2-e3 Rd8-d1 Qe3xg1 Bf5xb1
10 00:00 202,808 560,243 +3.86 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Nb8-c6 Kd2-c1 Nc6xb4 Qd1-d2 Ng8-e7 a2-a3 Nb4xc2 Kc1xc2 Bc8-f5+ Kc2-b3 Ra8-d8 Qd2-e3 Rd8-d1 Qe3xg1 Bf5xb1
11 00:00 451,561 599,682 +3.65 f2xg1N+ Ke2-e1
11 00:01 635,729 608,353 +3.66 f2xg1N+
11 00:01 734,210 614,916 +3.65 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Qh4-f4+ Kd2-c3 Nb8-c6 Rh1xg1 Qf4xe5+ Kc3-b3 Nc6-d4+ Kb3-a3 a7-a5 Bb4-d2 Qe5-c5+ b2-b4 Nd4xc2+ Qd1xc2 Qc5xg1 Qc2-d3 a5xb4+ Ka3-b3 Bc8-e6 Bd2xb4
11 00:01 734,257 613,414 +3.65 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Qh4-f4+ Kd2-c3 Nb8-c6 Rh1xg1 Qf4xe5+ Kc3-b3 Nc6-d4+ Kb3-a3 a7-a5 Bb4-d2 Qe5-c5+ b2-b4 Nd4xc2+ Qd1xc2 Qc5xg1 Qc2-d3 a5xb4+ Ka3-b3 Bc8-e6 Bd2xb4
12 00:01 1,211,611 619,433 +3.69 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Qh4-f4+ Kd2-c3 Nb8-c6 Rh1xg1 Qf4xe5+ Kc3-b3 Nc6-d4+ Kb3-a3 a7-a5 Bb4-d2 Qe5-c5+ b2-b4 Nd4xc2+ Ka3-b2 Qc5xg1 Kb2xc2 a5xb4 Bd2xb4 Ng8-e7 Qd1-e2 Bc8-f5+ Kc2-b2 c7-c5 Bb4-c3 O-O
12 00:02 1,235,646 618,131 +3.69 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Qh4-f4+ Kd2-c3 Nb8-c6 Rh1xg1 Qf4xe5+ Kc3-b3 Nc6-d4+ Kb3-a3 a7-a5 Bb4-d2 Qe5-c5+ b2-b4 Nd4xc2+ Ka3-b2 Qc5xg1 Kb2xc2 a5xb4 Bd2xb4 Ng8-e7 Qd1-e2 Bc8-f5+ Kc2-b2 c7-c5 Bb4-c3 O-O
13 00:02 1,516,411 619,448 +3.85 f2xg1N+
13 00:03 2,115,231 631,412 +3.84 f2xg1N+ Ke2-e1
13 00:04 2,566,529 636,697 +3.97 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Qh4-f4+ Kd2-c3 Nb8-c6 Rh1xg1 Qf4xe5+ Kc3-d2 Bc8-f5 Bf1-d3 Bf5xd3 Qd1-e1 Bd3-e4 Kd2-c1 Nc6xb4 Qe1xb4 O-O-O Rg1-e1 Ng8-f6 Nb1-c3 Kc8-b8 Kc1-b1 a7-a5 Qb4-b5 Qe5xb5
13 00:04 2,566,578 636,236 +3.97 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Qh4-f4+ Kd2-c3 Nb8-c6 Rh1xg1 Qf4xe5+ Kc3-d2 Bc8-f5 Bf1-d3 Bf5xd3 Qd1-e1 Bd3-e4 Kd2-c1 Nc6xb4 Qe1xb4 O-O-O Rg1-e1 Ng8-f6 Nb1-c3 Kc8-b8 Kc1-b1 a7-a5 Qb4-b5 Qe5xb5
14 00:06 4,148,689 638,947 +4.00 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Qh4-f4+ Kd2-c3 Nb8-c6 Rh1xg1 Qf4xe5+ Kc3-d2 Qe5-d4+ Kd2-c1 Qd4xg1 Bb4-c3 Qg1-e3+ Qd1-d2 Qe3xd2+ Nb1xd2 f7-f6 b2-b3 Bc8-f5 Kc1-b2 Ra8-d8 Bf1-e2 Ng8-e7 Ra1-d1 Nc6-d4 Be2-h5+ g7-g6
14 00:06 4,194,709 638,464 +4.00 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Qh4-f4+ Kd2-c3 Nb8-c6 Rh1xg1 Qf4xe5+ Kc3-d2 Qe5-d4+ Kd2-c1 Qd4xg1 Bb4-c3 Qg1-e3+ Qd1-d2 Qe3xd2+ Nb1xd2 f7-f6 b2-b3 Bc8-f5 Kc1-b2 Ra8-d8 Bf1-e2 Ng8-e7 Ra1-d1 Nc6-d4 Be2-h5+ g7-g6
15 00:10 7,107,013 647,740 +3.84 f2xg1N+ Ke2-e1
15 00:11 7,506,937 646,982 +3.85 f2xg1N+
15 00:13 8,528,957 651,413 +3.99 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Qh4-f4+ Kd2-c3 Nb8-c6 Qd1-e1 Bc8-g4 Bb4-c5 Ng1-e2+ Bf1xe2 Qf4xe5+ Kc3-d2 Bg4xe2 Nb1-c3 O-O-O+ Kd2-c1 Qe5xc5 Qe1xe2 Ng8-f6 Qe2-f1 Rh8-e8 Kc1-b1 Kc8-b8 a2-a3 Rd8-d4 Rh1-g1 Rd4-h4
15 00:13 8,529,008 651,268 +3.99 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Qh4-f4+ Kd2-c3 Nb8-c6 Qd1-e1 Bc8-g4 Bb4-c5 Ng1-e2+ Bf1xe2 Qf4xe5+ Kc3-d2 Bg4xe2 Nb1-c3 O-O-O+ Kd2-c1 Qe5xc5 Qe1xe2 Ng8-f6 Qe2-f1 Rh8-e8 Kc1-b1 Kc8-b8 a2-a3 Rd8-d4 Rh1-g1 Rd4-h4
16 00:17 11,222,370 659,053 +3.83 f2xg1N+ Ke2-e1
16 00:18 11,903,036 657,771 +3.84 f2xg1N+
16 00:20 13,745,532 661,224 +3.91 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Qh4-f4+ Kd2-c3 Nb8-c6 Qd1-e1 Bc8-g4 Bf1-d3 a7-a5 Bb4-c5 Ng1-e2+ Bd3xe2 Qf4xe5+ Kc3-d2 Bg4xe2 Nb1-c3 O-O-O+ Kd2-c1 Qe5-g5+ Kc1-b1 Be2xc4 Bc5-e3 Qg5xg2 Rh1-g1 Qg2xh2 Rg1xg7 Ng8-e7 Qe1-c1 Ne7-f5 Be3-f4 Qh2-f2 Rg7-g5
16 00:20 13,745,582 661,131 +3.91 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Qh4-f4+ Kd2-c3 Nb8-c6 Qd1-e1 Bc8-g4 Bf1-d3 a7-a5 Bb4-c5 Ng1-e2+ Bd3xe2 Qf4xe5+ Kc3-d2 Bg4xe2 Nb1-c3 O-O-O+ Kd2-c1 Qe5-g5+ Kc1-b1 Be2xc4 Bc5-e3 Qg5xg2 Rh1-g1 Qg2xh2 Rg1xg7 Ng8-e7 Qe1-c1 Ne7-f5 Be3-f4 Qh2-f2 Rg7-g5
17 00:28 19,017,980 667,836 +3.90 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Qh4-f4+ Kd2-c3 Nb8-c6 Qd1-e1 Bc8-g4 Bf1-d3 a7-a5 Bb4-c5 Ng1-e2+ Bd3xe2 Qf4xe5+ Kc3-d2 Bg4xe2 Nb1-c3 O-O-O+ Kd2-c1 Qe5-g5+ Kc1-b1 Be2xc4 Bc5-e3 Qg5xg2 Rh1-g1 Qg2xh2 Rg1xg7 Ng8-e7 Qe1-f2 Qh2xf2 Be3xf2 Ne7-f5 Rg7-g5 Nc6-e7 Rg5-g4
17 00:28 19,229,249 668,006 +3.90 f2xg1N+ Ke2-e1 Qd8-h4+ Ke1-d2 Qh4-f4+ Kd2-c3 Nb8-c6 Qd1-e1 Bc8-g4 Bf1-d3 a7-a5 Bb4-c5 Ng1-e2+ Bd3xe2 Qf4xe5+ Kc3-d2 Bg4xe2 Nb1-c3 O-O-O+ Kd2-c1 Qe5-g5+ Kc1-b1 Be2xc4 Bc5-e3 Qg5xg2 Rh1-g1 Qg2xh2 Rg1xg7 Ng8-e7 Qe1-f2 Qh2xf2 Be3xf2 Ne7-f5 Rg7-g5 Nc6-e7 Rg5-g4

[Rebel]

[d]8/2P5/8/8/3r4/8/2K5/k7 w - -

Look more carefully and notice that K3 does play 6.c8=R here.

So it finds minor promotions from the root but not inside the search.

Is this a bug?

Not necessarily.

For game play removing all minor-promotions from the search gives a considerable speed-up in positions were promotions matter. It can be seen as a small ELO improvement.

Look at Rybka 1|2|3|4. Not one of those supports minor-promotions, not even at the root. I don't think that's a bug, it's deliberate.

[lucasart]

[d]8/2P5/8/8/3r4/8/2K5/k7 w - -

Look more carefully and notice that K3 does play 6.c8=R here.

So it finds minor promotions from the root but not inside the search.

Is this a bug?

Not necessarily.

For game play removing all minor-promotions from the search gives a considerable speed-up in positions were promotions matter. It can be seen as a small ELO improvement.

Look at Rybka 1|2|3|4. Not one of those supports minor-promotions, not even at the root. I don't think that's a bug, it's deliberate.

Yes and no. I tested the following with DoubleCheck:
1/ remove under promotions alltogether: it was actually weaker!? not much so, but still significant
2/ remove underpromotions only in the qsearch: that actually was an improvement.

So now I consider all promotions in the search and only queen promotions in the qsearch.

[BubbaTough]

1/ remove under promotions alltogether: it was actually weaker!? not much so, but still significant

Try only removing B/R promotions. I doubt it will be weaker (unless there is some bug like resigning if the opponent under promotes.

-Sam

[mar]

So now I consider all promotions in the search and only queen promotions in the qsearch.

Exactly what I do. I also sort underpromotions last in main search.

[Rebel]

[d]8/2P5/8/8/3r4/8/2K5/k7 w - -

Look more carefully and notice that K3 does play 6.c8=R here.

So it finds minor promotions from the root but not inside the search.

Is this a bug?

Not necessarily.

For game play removing all minor-promotions from the search gives a considerable speed-up in positions were promotions matter. It can be seen as a small ELO improvement.

Look at Rybka 1|2|3|4. Not one of those supports minor-promotions, not even at the root. I don't think that's a bug, it's deliberate.

Yes and no. I tested the following with DoubleCheck:
1/ remove under promotions alltogether: it was actually weaker!? not much so, but still significant
2/ remove underpromotions only in the qsearch: that actually was an improvement.

So now I consider all promotions in the search and only queen promotions in the qsearch.

Exactly what I am doing since my early days, skip them in QS only. And it gave a considerable saving because basically you remove 3 captures. Never tried it in the main search because I could not afford it anyway, see the user complaints at Rybka forum

[Mike S.]

Look at Rybka 1|2|3|4. Not one of those supports minor-promotions, not even at the root.

Only the bishop underpromotion is missing (I am not sure about every Rybka version), which was largely debated. Knight u.p. is ok, in search and even without immediate check:

Babushkin - Postnikow, Corr. 1970 (from Wikipedia)
[d]2K5/4P1p1/kp6/1p6/1P6/2P1P3/7p/1r6 w - - 0 1

Analysis by Rybka 2.3.2a mp 32-bit:

1.e8N gives a perpetual. Ka7 2.Nc7 Rxb4 3.cxb4 h1Q 4.Nxb5+ Ka6 5.Nc7+ Ka7 6.Nb5+ Ka6 7.Nc7+ Ka7
= (0.00) Depth: 7 00:00:00 5kN
= (0.00) Depth: 23 00:00:08 272kN

(1.e8Q loses.)

I think the missing bishop u.p. has to do with the size or effectiveness of a related variable, and/or the number of possible moves a pawn on the 7th rank can make, or similiar programmer's stuff

I think rook u.p. is more important, like shown here. Bishop u.p. really is "studies only" and my proposal was to offer it as an option, probably causing a little slowdown, which could be off by default.

[pichy]

I also tested Critter 1.4, Houdini, Spark, Spike 1.4 and none of those underpromote to Rook in the given position of the Saavedra study( Rook versus pawn). I believe that the main reason is that the author of those programs thought that the Queen covers all the moves that the rook covers plus the diagonals, but for this case promoting to a Queen only get the engine to draw (stalemate) instead of winning

[d]8/8/1KP5/3r4/8/8/8/k7 w - - 0 1

Look at Rybka 1|2|3|4. Not one of those supports minor-promotions, not even at the root.

Only the bishop underpromotion is missing (I am not sure about every Rybka version), which was largely debated. Knight u.p. is ok, in search and even without immediate check:

Babushkin - Postnikow, Corr. 1970 (from Wikipedia)
[d]2K5/4P1p1/kp6/1p6/1P6/2P1P3/7p/1r6 w - - 0 1

Analysis by Rybka 2.3.2a mp 32-bit:

1.e8N gives a perpetual. Ka7 2.Nc7 Rxb4 3.cxb4 h1Q 4.Nxb5+ Ka6 5.Nc7+ Ka7 6.Nb5+ Ka6 7.Nc7+ Ka7
= (0.00) Depth: 7 00:00:00 5kN
= (0.00) Depth: 23 00:00:08 272kN

(1.e8Q loses.)

I think the missing bishop u.p. has to do with the size or effectiveness of a related variable, and/or the number of possible moves a pawn on the 7th rank can make, or similiar programmer's stuff

I think rook u.p. is more important, like shown here. Bishop u.p. really is "studies only" and my proposal was to offer it as an option, probably causing a little slowdown, which could be off by default.

[rvida]

I also tested Critter 1.4, Houdini, Spark, Spike 1.4 and none of those underpromote to Rook in the given position of the Saavedra study( Rook versus pawn). I believe that the main reason is that the author of those programs thought that the Queen covers all the moves that the rook covers plus the diagonals, but for this case promoting to a Queen only get the engine to draw (stalemate) instead of winning

[d]8/8/1KP5/3r4/8/8/8/k7 w - - 0 1

Are you sure that underpromotion is the right move? The line from tablebase shows a queen promotion.

Code: Select all

position fen 8/8/1KP5/3r4/8/8/8/k7 w - - 0 1
go
1/ 1  00:00         0       0   +M26 c7 Rd3 c8=Q Kb2 Qc4 Rd8 Qe4 Kc1 Qf4+ Kd1 Kb5 Ke2 Qh2+ Kd1 Qh5+ Kc1 Kc4 Rd2 Qg6 Kb2 Qf5 Rc2+ Kd3 Rc3+ Kd2 Rb3 Qf6+ Kb1 Qe5 Rb6 Qe4+ Ka2 Qc4+ Kb1 Kc3 Rb2 Qa4 Rb8 Qd1+ Ka2 Qd5+ Ka1 Qh1+ Ka2 Qh2+ Kb1 Qxb8+ Kc1 Qb2+ Kd1 Qd2#