Hard-Talkchess-2020 set, final release

[Vinvin]

can anyone please post a few epds of the toughest ones with a brief comment about the right plan to follow?

Hi !
Here is a list of the 46 hardest positions from a previous test with an earlier Stockfish.
About the plan to follow, I will import all the positions in "Lichess Studies" (packs of 20 positions) with mainlines in following days.

Code: Select all

1k6/3p4/1B6/4Pp1p/1p5R/1p4p1/pP3n2/K6n w - - bm Rxh1;                                                                         
1nb3rr/2pp3p/p1k2P2/n1q1PN2/P1pR3N/b4PBp/1R2P3/1KQ5 w - - bm Rd6+;                                                            
1r3r2/4bpkp/1qb1p1p1/3pP1P1/p1pP1Q2/PpP2N1R/1Pn1B2P/3RB2K w - - bm Qf6+;                                                      
1r6/4k3/r2p2p1/2pR1p1p/2P1pP1P/pPK1P1P1/P7/1B6 b - - bm Rxb3+;                                                                
1rb4k/p5np/3p1rp1/1ppB4/2N2P2/1P2R1P1/P1P4P/4R1K1 w - - bm Re8+;                                                              
2b1r3/r2ppN2/8/1p1p1k2/pP1P4/2P3R1/PP3PP1/2K5 w - - bm Nd6+;                                                                  
2bqrr1k/p5b1/1p1p2pp/nPpBp3/P1P1N2P/3PN1P1/R4P1K/3Q1R2 w - - bm Nc3;                                                          
2r2r1k/qb3p1p/p2p1Np1/np1Pp1P1/2P5/3RP2P/5R2/4Q1K1 w - - bm Rf4;                                                              
3B4/1r2p3/r2p1p2/bkp1P1p1/1p1P1PPp/p1P1K2P/PPB5/8 w - - bm Ba4+;                                                              
3k4/2pq3p/pp5R/4P3/P6P/2PN4/1PN3K1/8 w - - bm Nd4;                                                                            
4K1k1/8/1p5p/1Pp3b1/8/1P3P2/P1B2P2/8 w - - bm f4;                                                                             
4R3/P7/2bB2p1/r3pk2/2p2p1P/2P2P1K/6P1/8 w - - bm Rxe5+;                                                                       
4b1k1/P4pPp/1R3P1P/2r5/8/1P6/1K6/8 w - - bm a8=B;                                                                             
4k1br/1K1p1n1r/2p2pN1/P2p1N2/2P3pP/5B2/P2P4/8 w - - bm Kc8;                                                                   
4q1kr/p6p/1prQPppB/4n3/4P3/2P5/PP2B2P/R5K1 w - - bm Qxe5;                                                                     
4r1k1/1r1np3/1pqp1ppB/p7/2b1P1PQ/2P2P2/P3B2R/3R2K1 w - - bm Bg5;                                                              
6br/1KNp1n1r/2p2p2/P1ppRP2/1kP3pP/3PBB2/PN1P4/8 w - - bm Bxc5+;                                                               
7k/4PBNp/7P/2p5/2P5/2rq2Pp/1P3P1K/8 b - - bm Qxg3+;                                                                           
8/2N4r/1p3pkp/8/5K1p/2P4N/P3Bn2/8 w - - bm Bh5+;                                                                              
8/3P3k/n2K3p/2p3n1/1b4N1/2p1p1P1/8/3B4 w - - bm Nf6+;                                                                         
N7/8/2KQ2rp/6k1/3p3p/2p4P/4PP2/5N2 w - - bm f4+;                                                                              
qrb5/b1Pp3p/1pkB4/1p2P3/1P1KPp2/R4P1p/P3pP1P/4N1n1 w - - bm Rc3+;                                                             
r1b2rk1/2q1bppp/p2pp3/2n3PQ/1p1BP3/1BN5/PPP2P1P/2KR2R1 w - - bm Bf6;                                                          
r1b2rk1/pp3ppp/1n2p3/3pP3/2P2B2/b1PB1qP1/P1Q1NP1P/1K1R3R w - - bm Nd4;                                                        
r2qrbk1/1b3p2/3p2pB/1p1P1N2/1np1P3/5NRP/1P3PP1/1B1Qn1K1 w - - bm N3h4;                                                        
r5k1/p1pb2bp/3p3r/P1pPp1p1/2B1Pq2/1R2QPp1/1P4PP/5RBK b - - bm Rxh2+;                                                          
rn3r1k/pn1p1ppq/bpp4p/7P/4N1Q1/6P1/PP3PB1/R1B1R1K1 w - - bm Bg5;                                                              
rnb2rk1/1pq1bppp/p3p3/2npP1P1/3N1P2/2NBB3/PPP4P/R2QK2R w KQ - bm Bxh7+;                                                       
rnbq1rk1/p3bp2/2p1p1p1/1p1nP1P1/2pP1P2/2N5/PP4P1/RBBQK2R w KQ - bm Bd2;                                                       
r1qn1rk1/p1pp1ppp/1p2p2b/1P2P3/P7/R2B1N2/1P4PP/3Q1R1K w - - bm Bxh7+;                                                         
r4rk1/pb3pp1/1p1qpR2/2ppN3/3P4/2P4P/PP1pQ1P1/3R2K1 w - - bm Qxd2 Ng4;                                                         
Nq3n1k/8/P1p2Npp/2Kp3R/pp6/n3p3/8/4Q3 w - - bm Nc7;                                                                           
r7/1q1r1p1k/3p1Qpp/pbpP4/1nB1P1P1/1P3NBP/5P1K/4R3 w - - bm Bf4;                                                               
r2q1r2/1b2bpkp/n3p1p1/2ppP1P1/p6R/1PN1BQR1/NPP2P1P/4K3 w - - bm Qf6+;                                                         
5rk1/pp3p1p/7b/2pR4/8/2P4P/P1PNr1P1/2K4R b - - bm Rxd2;                                                                       
4K2k/1pQp3p/1R4p1/1P4P1/8/3P3P/1q3r2/8 w - - bm Qd8;                                                                          
1B1r4/rp2npkp/2b1pbp1/1qp5/nPN1R3/1P1P1QP1/2P2PBP/5R1K w - - bm Qxf6+;                                                        
7q/b1p5/1p1Npkb1/pPP2ppP/P1P5/3B2P1/5P1R/K3R3 w - - bm hxg6;                                                                  
8/1B3b1p/7P/p2p4/P2pp1kp/Q7/1N1Pq2P/7K w - - bm Qh3+;                                                                         
1r3rk1/5p1p/p2pb1p1/3Np1P1/2pRPR2/qP2Q1N1/P1PK3P/8 w - - bm Nf5;                                                              
r1bq1b1r/pp1pp1pp/2n1k3/8/4nQ2/N1p1BN1P/PP3PP1/R3K2R w KQ - bm Nb5;                                                           
3nn3/2p2p1k/1p1pp1p1/p2B3p/r2B2N1/4P2N/3PKPP1/6q1 w - - bm Ng5+;                                                              
r4rk1/1bq1bppp/p1npp3/2nB1NP1/1p2P3/2N1B3/PPP1QP1P/2KR2R1 w - - bm Qh5;                                                       
r1b1qk1r/pn3pp1/1p2p3/3pP1Bp/1bPN4/4Q1R1/P2N1PP1/R5K1 w - - bm Bf6;                                                           
8/8/p2k1p2/1p1p3p/1P1P3p/P3NPP1/5K2/1b6 w - - bm Ng2;                                                                         
r1br2k1/1p1nqppp/p1p1pn2/3p4/1bPP4/PPN1PN2/2BB1PPP/R2Q1RK1 b - - am Bxa3;  

[peter]

Crystal 2019 is here not as good as Houdini Tac, but of course the second best engine. In some other testsuites it's better than Houdini. BTW when are engines tested at TCEC hardware 176 threads?

My latest Crystal- compile at same hardware-TC as Houdini 6 T and the others I showed above, again 1 minute/pos., 6MV (maybe I'll give it one more try at 8MV next time, looked like missing still some solutions for not having them in 6 best lines 1 minute long, as much as I saw watching from time to time)
130/213

Code: Select all

Bisher gelöst: 130 von 213  ;  117:05m

         1   2   3   4   5   6   7   8   9  10  11  12  13  14  15  16  17  18  19  20
 -------------------------------------------------------------------------------------
   0 |   6  30   -   4   1   -   -   -   8   0   0   -  30   -  50   -   7   -   -   -
  20 |   -   3   -   -   1   -   0  36  38   1   1  37   -   -  40   -  54   -  32   0
  40 |   -   2  49   -  29   4   -   -   0  35  46  39  11   -   0   3   -   -  26   0
  60 |  47   -  16  51   -  21   -   -   6   -   3   0   -   -  30  21   -   -  27   -
  80 |   8   0  31  12   8  10   -   0  10   -   -   4   -   0   -   4   -   2  20   6
 100 |   0   -   5   -   -   3   0  23  40   1   -   -   -   6  14  21   -   9   1  13
 120 |   0   5  16   4   -  16   2   1  15   3   -  26  30   0   -   -  47   -   -   0
 140 |  31  24  24  60   -   -  58   -   3   -   7  11   -   -  18  18  13   -  22  13
 160 |   3   -   6   -   3  11   -   -   -   -   -   0   9   1  11  58   -  13   0   7
 180 |   0  14   -   -   3   -   8   2   -   -   -  18   -  56  34   -  21   3   0   -
 200 |   -   7   5   -  14  51   0   -  16   -   3   0   -
 

[giovanni]

Thank you. Appreciated. I didn't know about this Lichess feature. I am really looking forward to look at the mainlines!

[Dann Corbit]

The attached file has my database analysis for the test set.
I only have one different solution, and the difference is partial, because the problem has multiple solutions:
[d]4r1k1/1r1np3/1pqp1ppB/p7/2b1P1PQ/2P2P2/P3B2R/3R2K1 w - - acd 51; acs 900; bm Bg5 Bg7 Bxc4+ Be3; c0 "Bg5 = 343, Bg7 = 170, Bxc4 = 128, Be3 = 123"; c1 "10 4 2 2"; c3 "Bg7"; ce 343; id "Hard-Talkchess-2020.080"; pm Bg5; pv Bg5 Nf8 Qh8+ Kf7 Rxd6 Qc5+ Kg2 fxg5 Rh7+ Nxh7 Qxh7+ Kf8 Qh6+ Kf7 Qxg6+ Kf8 Rc6 Bxe2 Rxc5 bxc5 Qf5+ Kg7 Qxg5+ Kf8 Qf5+ Kg7 Qe5+ Kf8 Qh8+ Kf7 Qh5+ Kf8 e5 Rb2 Qh8+ Kf7 Qh7+ Kf8 Kg3 e6 f4 Bxg4 Kxg4 Rg2+ Kf3 Rg7 Qe4 Rg1 Qe3 Rd1 Qxc5+ Kf7 Qc4 Kf8 Qc7 a4 Qc4 Re7 Ke2 Rdd7 Qxa4 Ra7 Qb3 Ra6 a4 Kf7 Qc4 Ra5 Kf3 Rea7 Qd3 Re7 Qh7+ Kf8 Qh8+ Kf7 Ke3 Raa7 Qh7+ Kf8 Qe4 Kg7 Qc4 Rac7 Qb3 Ra7 Kd3 Ra6 Qc4 Ra5 Kc2;

I have the following scores for various moves (in centipawns):
"Bg5 = 343, Bg7 = 170, Bxc4 = 128, Be3 = 123"

So Bg7 does indeed win, but Bg5 is a little better. Bxc4 and Be3 probably also win.
(apparently, I cannot attach the file because the board quota has been reached).

[giovanni]

The attached file has my database analysis for the test set.
I only have one different solution, and the difference is partial, because the problem has multiple solutions:
[d]4r1k1/1r1np3/1pqp1ppB/p7/2b1P1PQ/2P2P2/P3B2R/3R2K1 w - - acd 51; acs 900; bm Bg5 Bg7 Bxc4+ Be3; c0 "Bg5 = 343, Bg7 = 170, Bxc4 = 128, Be3 = 123"; c1 "10 4 2 2"; c3 "Bg7"; ce 343; id "Hard-Talkchess-2020.080"; pm Bg5; pv Bg5 Nf8 Qh8+ Kf7 Rxd6 Qc5+ Kg2 fxg5 Rh7+ Nxh7 Qxh7+ Kf8 Qh6+ Kf7 Qxg6+ Kf8 Rc6 Bxe2 Rxc5 bxc5 Qf5+ Kg7 Qxg5+ Kf8 Qf5+ Kg7 Qe5+ Kf8 Qh8+ Kf7 Qh5+ Kf8 e5 Rb2 Qh8+ Kf7 Qh7+ Kf8 Kg3 e6 f4 Bxg4 Kxg4 Rg2+ Kf3 Rg7 Qe4 Rg1 Qe3 Rd1 Qxc5+ Kf7 Qc4 Kf8 Qc7 a4 Qc4 Re7 Ke2 Rdd7 Qxa4 Ra7 Qb3 Ra6 a4 Kf7 Qc4 Ra5 Kf3 Rea7 Qd3 Re7 Qh7+ Kf8 Qh8+ Kf7 Ke3 Raa7 Qh7+ Kf8 Qe4 Kg7 Qc4 Rac7 Qb3 Ra7 Kd3 Ra6 Qc4 Ra5 Kc2;

I have the following scores for various moves (in centipawns):
"Bg5 = 343, Bg7 = 170, Bxc4 = 128, Be3 = 123"

So Bg7 does indeed win, but Bg5 is a little better. Bxc4 and Be3 probably also win.
(apparently, I cannot attach the file because the board quota has been reached).

Impressive, thank you. May be you could share a wetransfer link?

[peter]

The attached file has my database analysis for the test set.
I only have one different solution, and the difference is partial, because the problem has multiple solutions:
[d]4r1k1/1r1np3/1pqp1ppB/p7/2b1P1PQ/2P2P2/P3B2R/3R2K1 w - - acd 51; acs 900; bm Bg5 Bg7 Bxc4+ Be3; c0 "Bg5 = 343, Bg7 = 170, Bxc4 =
So Bg7 does indeed win, but Bg5 is a little better. Bxc4 and Be3 probably also win.

Thanks for pointing at that again, Dann!
That's Persson- Teichmeister corr. 2009, used by Walter Eigenmann in his Eret too.
In this one testset it's corrected already, as far as I know, at least in my database. I'll do so in HTC20 too, Bg5 ist best move.

Edit: Oops, at my version of HTC2020, copied from the starting- posting of Vincent's, it's already 1.Bg5, that's stored as solution.

[peter]

The final release is ready (213 positions with id) :

New leader in my personal list at 1 minute/pos:
CorChess 5.1, here again in 8MV, analysis contempt 100 both, as always 24 threads of 12x3GHz, 8G hash, no tbs:
146/213

Code: Select all

Engine: CorChess 5.1 160120 64 POPCNT0 (8192 MB)
von I. Ivec

Bisher gelöst: 146 von 213  ;  97:52m

         1   2   3   4   5   6   7   8   9  10  11  12  13  14  15  16  17  18  19  20
 -------------------------------------------------------------------------------------
   0 |  38  24   -  11   0  24   -   -   3   0   3   -  16   -  34   -   3   -  56   -
  20 |   -   5   -   -   1   -   0  12  14   1   1   8   -   -  59   -  50   -  55   0
  40 |   1   9   -   -   8   6   -  11   1   1  57   1   -   -   0   1   -  45  49   0
  60 |  35   -  31   4  12   2   -  24  40   -  12   0   -   -  17   2   -   -  15   -
  80 |   5   -   5   0  24   0   9   0   3   -   -  15   -   0   -  14   -   3  19  32
 100 |   0   -   7   3   -   4   0  14   6   2   -  43   -   6   7   4   -   6   -  15
 120 |   0   1  17   1   -   -   0   0  25   3   -  41   -   0   1   5   5   3  47   2
 140 |   -  17   8   -   0   3   3   0   1   -   3   3   -  35  41   5  55   0  25  18
 160 |   4   0  15   -   2   4   -   -   -   -   -   0   4   1   2   -   -  11   0   9
 180 |   2   3  22   -   8  29   8   3  39   1   -   3   -  17   -   -  11   1   5   -
 200 |  59   4  26   -   3   2   0   -  10  21   0   0   -

[Ovyron]

Contempt 100 is required to find the best moves? That's new

[Paloma]

Impressive, thank you. May be you could share a wetransfer link?

Go here http://rybkaforum.net/cgi-bin/rybkaforu ... #pid584904
and download at the bottom vhard-ana.7z

[peter]

Contempt 100 is required to find the best moves? That's new

To me it isn't.

Of course, it's not required, it helps a little in most tactical postions for the side to move, if it's about game changing best moves for the side in advantage.
And 100 (max. for most SFes) for both sides (analysis contempt) isn't so much difference to 24, (at least and e.g. not for SF), which is default for most SFes lately.

Give it a try, if you really didn't do so till now, have fun. Of course, if you are too much serious a tester for such disrespectful results, you have to compare to contempt 0 and to contempt 24 (default) and something in between too. Then you know, wich is best for sure and: how much difference it makes or doesn't make, of course for each and every of the many many engines we have nowadays.
It's not the holy cow of Elo, I'm measuring here, and it's just for my personal interest, you know? You need not take it too serious, from my side.