Hard talkchess 2018 set - work in progress

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Vinvin
Posts: 4058
Joined: Thu Mar 09, 2006 8:40 am

Re: Hard talkchess 2018 set - work in progress

Post by Vinvin » Tue Jul 17, 2018 9:33 am

Paloma wrote:
Tue Jul 17, 2018 8:13 am

I'm interrested in these 18 special positions you mentioned above.
Here are they :

Code: Select all

1b6/2p5/1kB5/1PN5/8/3K4/8/q7 w - - bm Nb3; c1 "only winning move";
1k3b1q/pP2p1p1/P1K1P1Pp/7P/2B5/8/8/8 w - - bm Kd5 Bd5; c1 "diff=8";
1r4Q1/7p/2N1B2k/7P/3Pp1P1/K1N1P2R/PP2PR2/8 b - - ; c0 "eval +-, no repetition";
2r5/6pk/7p/8/8/7P/1rbR1RPK/8 b - - ;  c0 "eval = Draw ! No progress possible because of the pin";
4k2b/7P/4KP1P/8/8/pB6/P7/8 w - - bm f7; c0 "eval=mate"; 
6k1/8/8/8/q5p1/3p2P1/3N1R2/6K1 b - - 1 45 ; c0 "eval = draw by fortress";
6r1/3q1npk/2b1pbnp/Rp1p4/1N1P3P/2PQ1pP1/1K3B2/5B1R w - - bm Qxg6; c1 "mate";
8/1p6/1p6/kPp2P1K/2P5/N1Pp4/q2P4/1N6 w - -; c0 "eval = draw"; c1 "diff=9";
8/2b2k1K/1pPp1p2/1P1P1P2/5B2/8/8/8 w - - bm Be3 Bd2 Bc1; c1 "the point to play Bxf6 on move 3";
8/2pPpP2/1P1qk1p1/1p4P1/1P4p1/1p1QK1P1/2PpPp2/8 w - - bm f8=Q; c1 "shortest mate";
8/8/1P1k3p/3P2pP/4KpP1/8/8/8 w - - ; c0 "eval = draw";
8/8/5k2/8/2B4p/5RbP/5pP1/7K b - -; c0 "eval = draw";
8/p1K5/6Bk/PP5P/8/8/5b2/8 w - - 0 1; c0 "eval = draw";
8/pp3p1k/6pp/7P/P5P1/3K4/1PP5/8 b - -; c0 "+- (SF pawn EG bug)";
brkrqbqn/qnpNqqbr/bnNqRbrn/qNrrnRrb/BrqnbqRr/bqRNnprR/QQbNNrNn/bqrnRbKb w - - Nbxa7;  c0 "eval = mate in 8";
kB5K/2P4B/P4rP1/4r3/2N1n3/1RP4N/3b4/6RQ b - -; c0 "eval +- no perpet";
r1b1k3/1p1p1pp1/1p1P3p/pP6/P7/8/5PPP/4R1K1 b q -; c0 "eval +/-";
r1b3kr/pp1n2Bp/2pb2q1/3p3N/3P4/2P2Q2/P1P3PP/4RRK1 w - - bm Re5; id "arasan19.124"; c0 "Orcrist-Crafty, ICC 2011";

Dann Corbit
Posts: 8757
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: Hard talkchess 2018 set - work in progress

Post by Dann Corbit » Tue Jul 17, 2018 6:23 pm

This position is not possible. Too many major pieces.
brkrqbqn/qnpNqqbr/bnNqRbrn/qNrrnRrb/BrqnbqRr/bqRNnprR/QQbNNrNn/bqrnRbKb w - - Nbxa7; c0 "eval = mate in 8";
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

Vinvin
Posts: 4058
Joined: Thu Mar 09, 2006 8:40 am

Re: Hard talkchess 2018 set - work in progress

Post by Vinvin » Tue Jul 17, 2018 6:47 pm

Dann Corbit wrote:
Tue Jul 17, 2018 6:23 pm
This position is not possible. Too many major pieces.
brkrqbqn/qnpNqqbr/bnNqRbrn/qNrrnRrb/BrqnbqRr/bqRNnprR/QQbNNrNn/bqrnRbKb w - - Nbxa7; c0 "eval = mate in 8";
Yes, I know. That's why this position is in the "special" category. ;-)

Dann Corbit
Posts: 8757
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: Hard talkchess 2018 set - work in progress

Post by Dann Corbit » Tue Jul 17, 2018 7:04 pm

These positions all need a + for check on the bm.

r1k1n2n/8/pP6/5R2/8/1b1B4/4N3/1K5N w - -
r1b2rk1/2q1bppp/pp2p3/2npP3/1n3P1P/2NBBN2/PPPQ2P1/1K1R3R w - -
4B3/1Pb4p/8/4p1k1/3p1p2/5K2/2P5/8 b - -
r1n2N1k/2n2K1p/3pp3/5Pp1/b5R1/8/1PPP4/8 w - -
4r2r/pppkq1pp/2n1pn2/4p1B1/4N2Q/8/PPP3PP/4RRK1 w - -
1B1r4/rp2npkp/2b1pbp1/1qp5/nPN1R3/1P1P1QP1/2P2PBP/5R1K w - -
6r1/3q1npk/2b1pbnp/Rp1p4/1N1P3P/2PQ1pP1/1K3B2/5B1R w - -
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

Dann Corbit
Posts: 8757
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: Hard talkchess 2018 set - work in progress

Post by Dann Corbit » Tue Jul 17, 2018 7:05 pm

Vinvin wrote:
Tue Jul 17, 2018 6:47 pm
Dann Corbit wrote:
Tue Jul 17, 2018 6:23 pm
This position is not possible. Too many major pieces.
brkrqbqn/qnpNqqbr/bnNqRbrn/qNrrnRrb/BrqnbqRr/bqRNnprR/QQbNNrNn/bqrnRbKb w - - Nbxa7; c0 "eval = mate in 8";
Yes, I know. That's why this position is in the "special" category. ;-)
I don't mind them, but there are a few engines that will crash or have other undefined behavior because an array bound is exceeded.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

Vinvin
Posts: 4058
Joined: Thu Mar 09, 2006 8:40 am

Re: Hard talkchess 2018 set - work in progress

Post by Vinvin » Sun Jul 22, 2018 4:28 pm

jdart wrote:
Mon Jul 16, 2018 12:48 am
A few others I have collected recently:

1r1q1r2/1pbn1ppk/2n1b2p/pQ2pN2/P1NpP3/3P2P1/1P1B1PBP/R1R3K1 b - - bm g6; c0 "Arasan-Spyderchess, FICS 2018";

r5k1/p4p1p/1pq1r1pB/1N1bP3/P1p4Q/3nR2P/5PP1/3R2K1 w - - bm Nd4; c0 "Galanov-Oliveira, CADAP-RUS ICCF 2017";

2r3k1/2b1Rpp1/p2p2qp/P7/1PpB4/2P3PP/5Q1K/8 w - - bm b5; c0 "Porozovsky-Bißmann, S-Open/6-sf08 2017";

Not sure if these qualify as "hard".

--Jon
Hi Jon,
I ran the 3 positions 5 times with SF, 180 sec per position on an i7-6*4GHz.
Only number 3 is hard enough.
Here are time to solve :

Code: Select all

  1) .. g7-g6           Nc6-a7   
  1) .. g7-g6           g7-g6   * 20 Seconds
  1) .. g7-g6           g7-g6   * 22 Seconds
  1) .. g7-g6           g7-g6   * 29 Seconds
  1) .. g7-g6           g7-g6   * 34 Seconds
  1) .. g7-g6           g7-g6   * 9 Seconds
  
  2) Nb5-d4             Nb5-d4  * 14 Seconds
  2) Nb5-d4             Nb5-d4  * 14 Seconds
  2) Nb5-d4             Nb5-d4  * 16 Seconds
  2) Nb5-d4             Nb5-d4  * 19 Seconds
  2) Nb5-d4             Nb5-d4  * 3 Seconds
    
  3) b4-b5              Kg8-d8   
  3) b4-b5              Qf2-f3   
  3) b4-b5              Qf2-f3   
  3) b4-b5              Qf2-f3   
  3) b4-b5              Qf2-f3   
  3) b4-b5              Qf2-f3  
With more time, SF can find #3 -> b5

FEN: 2r3k1/2b1Rpp1/p2p2qp/P7/1PpB4/2P3PP/5Q1K/8 w - - 0 1

Code: Select all

Stockfish_180714_x64_modern:
...
 41/60-	04:35	3 087 110 880	11 194 675	+0,82	1.Qf3 Kf8
 41/68+	05:43	3 840 752 631	11 176 642	+0,95	1.b5
 41/75+	07:42	5 076 305 710	10 984 037	+1,62	1.b5

Stockfish_180714_x64_modern:
...
 44/88+	09:54	6 336 932 379	10 663 766	+1,17	1.Qf3
 44/88+	13:57	8 946 584 467	10 679 759	+1,39	1.b5
 44/88+	17:02	10 873 417 438	10 630 541	+1,68	1.b5

Vinvin
Posts: 4058
Joined: Thu Mar 09, 2006 8:40 am

Re: Hard talkchess 2018 set - work in progress

Post by Vinvin » Mon Jul 23, 2018 2:52 pm

Now, the set includes 199 positions :

107 from here : http://talkchess.com/forum3/viewtopic.p ... 16#p757316
87 from here (2 positions take out from 89) : http://talkchess.com/forum3/viewtopic.p ... 97#p768397
3 from here : http://talkchess.com/forum3/viewtopic.p ... 07#p761407
one from here (b5!) : http://talkchess.com/forum3/viewtopic.p ... 42#p769142
one from here (b4!) : http://talkchess.com/forum3/viewtopic.p ... 65#p767365

Still 1 position to find to succeed the first release of this project more than 12 years old :-)

Version from 2006 here : https://www.stmintz.com/ccc/index.php?id=475956
Version from 2008 here : http://talkchess.com/forum3/viewtopic.p ... 62#p201262

Dann Corbit
Posts: 8757
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: Hard talkchess 2018 set - work in progress

Post by Dann Corbit » Mon Jul 23, 2018 6:28 pm

Do you have a link the the current set?
It is hard for me to know which positions to include by looking at the message trails.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

Vinvin
Posts: 4058
Joined: Thu Mar 09, 2006 8:40 am

Re: Hard talkchess 2018 set - work in progress

Post by Vinvin » Mon Jul 23, 2018 8:39 pm

Dann Corbit wrote:
Mon Jul 23, 2018 6:28 pm
Do you have a link the the current set?
It is hard for me to know which positions to include by looking at the message trails.
Sure !
Here are 2 files :
The 199 positions, each of the 5 groups separated by a blank line (203 lines) : http://home.scarlet.be/vincentlejeune/c ... 18-199.epd
I found 11 duplicates after sorting the file. 188 positions left : http://home.scarlet.be/vincentlejeune/c ... licate.epd

Colin-G
Posts: 114
Joined: Mon Oct 31, 2016 5:30 pm
Location: England

Re: Hard talkchess 2018 set - work in progress

Post by Colin-G » Tue Jul 24, 2018 7:42 am

When the epd list is finalized, I would suggest adding a number for the id parameter in all lines to help indentification.
I note that some positions do not have any id value at all.

Post Reply