Hard-Talkchess-2020 set, final release

Vinvin · Post by **Vinvin** » Wed Jul 15, 2020 2:41 am

Vinvin wrote: ↑Thu Jul 02, 2020 10:38 pm 2 more results :

ShashChess 12.0 : 111 solutions (1842 seconds)
Not a lot of solutions but hold some records now. This engine has surely something special.
Another curious thing : the search changes when I restart with the same position. There should be random things in this engine.
I ask ed the question in the ShashChess thread : http://talkchess.com/forum3/viewtopic.p ... 37#p849637

ShashChess has been updated to version 12.1.
No more random results.
This new version scores well : 154 position (1177 seconds). A bit better than Stockfish_20062710.
It holds 3 records :

Code: Select all

191) Bd3xh7      18 seconds
193) .. c3xb2    17 seconds
195) .. Rh6xh2   6 seconds

Results sheet updated : http://home.scarlet.be/vincentlejeune/c ... 2020v2.ods

MMarco · Post by **MMarco** » Wed Jul 15, 2020 7:59 am

Thanks for the results. Do you plan on testing the SFNNue engines? It might be interesting. I think this position is in your test, and StockFiNN 0.2 seems to do well on it (only one try though, 8 threads):

[d]r5k1/p1pb2bp/3p3r/P1pPp1p1/2B1Pq2/1R2QPp1/1P4PP/5RBK b - - 0 28

Code: Select all

1: Aaron Summerscale - Gawain Jones 0-1, British Championship 2009
r5k1/p1pb2bp/3p3r/P1pPp1p1/2B1Pq2/1R2QPp1/1P4PP/5RBK b - - 0 1

Analysis by Stockfish NNUE halfkp-384 x64:

28...Qxe3 29.Bxe3 Rxh2+ 30.Kg1 g4 31.fxg4 Rh4 32.Rb7 Rxg4 33.Rxc7 Rxe4 34.Rxd7 Rxc4 35.Rxd6 
  Black has an edge: = (-0.24)  Depth: 10/14   00:00:00
28...Qxe3 29.Bxe3 Rxh2+ 30.Kg1 Bf6 31.Rb7 Bd8 32.Rc1 
  Black is clearly better: -/+ (-1.34)  Depth: 11/18   00:00:00
28...Qxe3 29.Rxe3 Rb8 30.b3 Bb5 31.Bxb5 Rxb5 32.Ra1 Kf7 33.Rd3 a6 34.h3 Bf6 35.Rc3 Rh4 36.Be3 
  Black is better: -/+ (-1.07)  Depth: 12/18   00:00:00
28...Bc8 29.Rb1 Bf6 30.Rc1 Qxe3 31.Rxe3 Rb8 32.Rc2 Bd7 33.Rec3 Rh4 34.b3 Kg7 35.h3 
  Black is better: -/+ (-0.89)  Depth: 13/23   00:00:00  435kN
28...Bf6 29.Rc1 Qxe3 30.Rxe3 Rb8 31.Rb3 Rxb3 32.Bxb3 gxh2 33.Be3 Bb5 34.Bc4 Bxc4 35.Rxc4 
  Black is better: -/+ (-1.14)  Depth: 14/24   00:00:00  858kN
28...Bc8 29.Qe1 a6 30.Be2 Bf6 31.Bd3 Rh5 32.Bc4 Kg7 33.Be2 Bd8 34.Bc4 h6 35.Re3 Rb8 36.b3 Qh4 37.Qc3 
  Black is better: -/+ (-0.81)  Depth: 15/24   00:00:01  1589kN
28...Bc8 29.Qe1 a6 30.Be2 Bf6 31.Bc4 Kg7 32.Be2 Qh4 33.Bc4 Bd8 34.Qa1 Bd7 35.Qc1 Kg8 36.Qc3 
  Black is slightly better: =/+ (-0.68)  Depth: 16/24   00:00:01  1694kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Qh4 32.Bc4 Bd7 33.Qc3 Bd8 34.Qe1 Kh8 35.Be2 Bc8 36.Qa1 g4 37.Qa4 Rg6 38.fxg4 Bxg4 
  Black is slightly better: =/+ (-0.56)  Depth: 17/30   00:00:01  2080kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Qh4 32.Bc4 Be7 33.Bd3 Kh8 34.Bc4 Qf4 35.Be2 Bd8 36.Bc4 Bd7 37.Be2 Qh4 38.Bc4 
  Black is slightly better: =/+ (-0.58)  Depth: 18/30   00:00:01  2226kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Qh4 32.Be2 Bd8 33.Bc4 Bd7 34.Qc3 Kh8 35.Qe1 Be7 36.Qc3 Bc8 37.Bd3 Kg7 38.Qe1 Kg8 39.Bc4 Bd7 40.Be2 Bd8 
  Black is slightly better: =/+ (-0.48)  Depth: 19/34   00:00:01  2664kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Qh4 32.Be2 Kg7 33.Bd3 Be7 34.Bc4 Bd7 35.Be2 Kg8 36.Rb7 Bd8 37.Rb3 Kh8 38.Bd3 Bc8 39.Be2 Kg7 40.Bd3 Bf6 41.Bc4 
  Black is slightly better: =/+ (-0.49)  Depth: 20/34   00:00:01  2739kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Be7 32.Bc4 Kh8 33.Bd3 Bd7 34.Bc4 Qh4 35.Bd3 Bf6 36.Be2 Bd8 37.Bc4 Kg8 38.Be2 Qf4 39.Bd3 Kh8 40.Bc4 
  Black is slightly better: =/+ (-0.48)  Depth: 21/34   00:00:01  3020kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Be7 32.Bc4 Kh8 33.Bd3 Bd7 34.Bc4 Qh4 35.Bd3 Bf6 36.Be2 Bd8 37.Bc4 Kg8 38.Be2 Qf4 39.Bd3 Kh8 40.Bc4 Bc8 41.Bd3 
  Black is slightly better: =/+ (-0.48)  Depth: 22/34   00:00:01  3326kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Be7 32.Bc4 Kh8 33.Bd3 Bd7 34.Bc4 Qh4 35.Bd3 Bf6 36.Be2 Bd8 37.Bc4 Kg7 38.Bd3 Bc8 39.Be2 Bf6 40.Bd3 Kh8 41.Bc4 Be7 42.Be2 
  Black is slightly better: =/+ (-0.48)  Depth: 23/34   00:00:01  3585kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Be7 32.Bc4 Kh8 33.Bd3 Bd7 34.Bc4 Qh4 35.Bd3 Bf6 36.Bc4 Kg7 37.Be2 Bd8 38.Bd3 Kg8 39.Be2 Bc8 40.Bc4 Bf6 41.Be2 Kh8 42.Bd3 
  Black is slightly better: =/+ (-0.48)  Depth: 24/34   00:00:01  3776kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Be7 32.Bc4 Bd8 33.Be2 Qh4 34.Bc4 Bf6 35.Be2 Be7 36.Bc4 Kh8 37.Be2 Bd7 38.Bc4 Qf4 39.Be2 Bf6 40.Bd3 Qh4 41.Be2 Bd8 42.Bc4 Be7 
  Black is slightly better: =/+ (-0.48)  Depth: 25/34   00:00:03  8400kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Be7 32.Bc4 Bd8 33.Be2 Qh4 34.Bc4 Bf6 35.Be2 Be7 36.Bc4 Kh8 37.Be2 Bd7 38.Bc4 Qf4 39.Be2 Kg8 40.Bd3 Qh4 41.Bc4 Bf6 42.Rb7 Bd8 
  Black is slightly better: =/+ (-0.48)  Depth: 26/34   00:00:03  8407kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Be7 32.Bc4 Bd8 33.Be2 Qh4 34.Bc4 Bf6 35.Be2 Be7 36.Bc4 Kh8 37.Be2 Bd7 38.Bc4 Qf4 39.Be2 Kg8 40.Bd3 Qh4 41.Bc4 Bf6 42.Rb7 Bd8 43.Rb3 
  Black is slightly better: =/+ (-0.48)  Depth: 27/34   00:00:03  8438kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Be7 32.Bc4 Bd8 33.Be2 Qh4 34.Bc4 Bf6 35.Be2 Be7 36.Bc4 Kh8 37.Be2 Bd7 38.Bc4 Qf4 39.Be2 Kg8 40.Bd3 Kg7 41.Be2 Bd8 42.Bc4 Qh4 43.Bd3 Bf6 
  Black is slightly better: =/+ (-0.48)  Depth: 28/33   00:00:03  8466kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Be7 32.Bc4 Bd8 33.Be2 Qh4 34.Bc4 Bf6 35.Be2 Be7 36.Bc4 Kh8 37.Be2 Bd7 38.Bc4 Qf4 39.Be2 Kg8 40.Bd3 Kg7 41.Bc4 Bd8 42.Be2 Qh4 43.Bd3 Bf6 44.Bc4 Bc8 
  Black is slightly better: =/+ (-0.48)  Depth: 29/33   00:00:03  9033kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Be7 32.Bc4 Bd8 33.Be2 Qh4 34.Bc4 Bf6 35.Be2 Be7 36.Bc4 Kh8 37.Be2 Bd7 38.Bc4 Qf4 39.Be2 Kg8 40.Bd3 Kg7 41.Bc4 Bd8 42.Be2 Qh4 43.Bc4 Bf6 44.Bd3 gxh2 
  Black is slightly better: =/+ (-0.48)  Depth: 30/32   00:00:04  10681kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Be7 32.Bc4 Bd8 33.Be2 Qh4 34.Bc4 Bf6 35.Be2 Be7 36.Bc4 Kh8 37.Be2 Bd7 38.Bc4 Qf4 39.Be2 Kg8 40.Bd3 Kg7 41.Bc4 Bd8 42.Be2 Qh4 43.Bc4 Bf6 44.Bd3 Bc8 
  Black is slightly better: =/+ (-0.48)  Depth: 31/38   00:00:05  14069kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Be7 32.Bc4 Bd8 33.Be2 Qh4 34.Bc4 Bf6 35.Be2 Be7 36.Bc4 Kh8 37.Be2 Bd8 38.Bc4 Qf4 39.Bd3 Be7 40.Be2 Bf6 41.Bd3 Qh4 42.Bc4 Kg7 43.Be2 Bd7 44.Rb7 
  Black is slightly better: =/+ (-0.48)  Depth: 32/39   00:00:07  18638kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Be7 32.Bc4 Bd8 33.Be2 Qh4 34.Bc4 Bf6 35.Be2 Be7 36.Bc4 Qf4 37.Bd3 Kh8 38.Be2 Bf6 39.Bd3 Qh4 40.Bc4 Rh5 41.Be2 Kg7 42.Bc4 Be7 43.Be2 Bd8 44.Bd3 Kh8 45.Bc4 
  Black is slightly better: =/+ (-0.48)  Depth: 33/34   00:00:07  19227kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Be7 32.Bc4 Bd8 33.Be2 Qh4 34.Bc4 Bf6 35.Be2 Be7 36.Bc4 Qf4 37.Bd3 Kh8 38.Be2 Bf6 39.Bd3 Qh4 40.Bc4 Rh5 41.Be2 Kg7 42.Bc4 Be7 43.Be2 Bd8 44.Bd3 Kh8 45.Bc4 Be7 
  Black is slightly better: =/+ (-0.48)  Depth: 34/35   00:00:08  23391kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Be7 32.Bc4 Bd8 33.Be2 Qh4 34.Bc4 Bf6 35.Be2 Be7 36.Bc4 Qf4 37.Bd3 Kh8 38.Be2 Bf6 39.Bc4 Bd8 40.Bd3 Qh4 41.Be2 Rh5 42.Bc4 Kg7 43.Be2 Be7 44.Bc4 Qf4 45.Bd3 Bf6 46.Be2 
  Black is slightly better: =/+ (-0.48)  Depth: 35/36   00:00:11  29229kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Be7 32.Bc4 Bd8 33.Be2 Qh4 34.Bc4 Bf6 35.Be2 Be7 36.Bc4 Qf4 37.Bd3 Kh8 38.Be2 Bf6 39.Bc4 Bd8 40.Bd3 Qh4 41.Be2 Rh5 42.Bc4 Kg7 43.Be2 Kg8 44.Bc4 Kh8 45.Be2 Be7 46.Bc4 Bd7 
  Black is slightly better: =/+ (-0.48)  Depth: 36/37   00:00:12  34088kN
28...a6 29.Qe1 Bc8 30.Be2 Bf6 31.Bd3 Be7 32.Bc4 Bd8 33.Be2 Qh4 34.Bc4 Bf6 35.Be2 Be7 36.Bc4 Qf4 37.Bd3 Kh8 38.Be2 Bf6 39.Bc4 Bd8 40.Be2 Be7 41.Bd3 Qh4 42.Bc4 Rh5 43.Bd3 Bd8 44.Bc4 Kg7 45.Be2 Bf6 46.Bd3 Bd7 47.Be2 Be7 48.Bc4 
  Black is slightly better: =/+ (-0.48)  Depth: 37/40   00:00:13  34829kN
28...Rxh2+ 
  Black is slightly better: =/+ (-0.57 ++)  Depth: 38/40   00:00:17  46715kN, tb=2
28...Rxh2+ 
  Black is slightly better: =/+ (-0.66 ++)  Depth: 38/40   00:00:17  47596kN, tb=2
28...Rxh2+ 
  Black is better: -/+ (-0.80 ++)  Depth: 38/40   00:00:17  48522kN, tb=2
28...Rxh2+ 
  Black is better: -/+ (-0.99 ++)  Depth: 38/40   00:00:18  49416kN, tb=2
28...Rxh2+ 
  Black is better: -/+ (-1.26 ++)  Depth: 38/40   00:00:18  50522kN, tb=2
28...Rxh2+ 
  Black is clearly better: -+ (-1.61 ++)  Depth: 38/40   00:00:18  52317kN, tb=2
28...Rxh2+ 
  Black is clearly better: -+ (-2.08 ++)  Depth: 38/40   00:00:19  53809kN, tb=8
28...Rxh2+ 
  Black has a decisive advantage: -+ (-2.68 ++)  Depth: 38/40   00:00:20  56152kN, tb=28
28...Rxh2+ 
  Black has a decisive advantage: -+ (-3.47 ++)  Depth: 38/40   00:00:21  58876kN, tb=33
28...Rxh2+ 
  Black is clearly winning: -+ (-4.47 ++)  Depth: 38/46   00:00:23  64995kN, tb=33
28...Rxh2+ 
  Black is clearly winning: -+ (-5.74 ++)  Depth: 38/51   00:00:29  83644kN, tb=35
28...Rxh2+ 29.Bxh2 Qh4 30.Qg1 Rf8 31.Re1 Qh5 32.f4 exf4 33.e5 Bg4 34.a6 Kh8 35.Bf1 Bxe5 36.Rd3 Rb8 37.Rd2 Bd4 38.Rxd4 cxd4 39.Re7 Kg8 40.Bd3 f3 41.Rxh7 f2 42.Rxh5 fxg1Q+ 43.Bxg1 Bxh5 44.Bxd4 Rb3 45.Bxa7 
  Black is clearly winning: -+ (-6.11)  Depth: 38/60   00:00:35  105MN, tb=60

Vinvin · Post by **Vinvin** » Tue Jul 21, 2020 5:31 pm

MMarco wrote: ↑Wed Jul 15, 2020 7:59 am Thanks for the results. Do you plan on testing the SFNNue engines? It might be interesting.

I did some pretest with 20200721-1907.bin from here https://www.comp.nus.edu.sg/~sergio-v/nnue/
and it's very promising.
I'll run over the set this night.

Vinvin · Post by **Vinvin** » Mon Jul 27, 2020 1:18 am

Test finished for Stockfish.sse42.halfkp_256x2-32-32.profile-nnue.2020-07-19 with NN file : 20200721-1907.bin
Very good results for a young engine with a young NN.
155 solution found (1154 seconds).

That's 2 more than Stockfish_20062710.

This new engine get a lot of interesting records for 14 positions !

Code: Select all

 85) h5-h6       : 3   seconds
 97) .. Qh5-f5   : 12  seconds
100) a5-a6       : 1   seconds
133) Bh5-f3      : 38  seconds
139) .. Ne8-c7   : 6   seconds
146) Rd1-d8      : 0   seconds
155) Bd3-g6      : 7   seconds
160) Ne4-g5      : 0   seconds
184) .. Bf5-g4   : 123 seconds
185) .. Kg8-g7   : 1   seconds
186) Rf3-f6      : 4   seconds
189) Qf3xf4      : 4   seconds
193) .. c3xb2    : 0   seconds
197) .. Qb2-c2   : 25  seconds

Results sheet updated : http://home.scarlet.be/vincentlejeune/c ... 2020v2.ods

MOBMAT · Post by **MOBMAT** » Tue Aug 04, 2020 8:08 pm

I'm looking for the solution file for the Hard-Talkchess that has the best move mating depth, if known.

Vinvin · Post by **Vinvin** » Tue Aug 04, 2020 10:51 pm

MOBMAT wrote: ↑Tue Aug 04, 2020 8:08 pm I'm looking for the solution file for the Hard-Talkchess that has the best move mating depth, if known.

You can find them here : http://talkchess.com/forum3/viewtopic.p ... 17#p850417
You can download the 4 files from the "share" tab.

cc2150dx · Post by **cc2150dx** » Mon Aug 24, 2020 10:13 pm

Off topic:

When it comes to tactical test suites, what's recommended for cores and time for each position ? tablebases yes/no ?

Thanks

Vinvin · Post by **Vinvin** » Tue Aug 25, 2020 12:20 am

cc2150dx wrote: ↑Mon Aug 24, 2020 10:13 pm Off topic:

When it comes to tactical test suites, what's recommended for cores and time for each position ? tablebases yes/no ?

Thanks

30 minutes per position with 1 core.
As many EGTB as you want.

Vinvin · Post by **Vinvin** » Sun Dec 27, 2020 11:09 pm

Results for 6 new engines (NN file used : nn-62ef826d1a6d.nnue) :

Code: Select all

SugaR.AI.ICCF.1.00 : 166 solved
ShashChess15.1     : 166 solved
Bluefish-12-W      : 162 solved
Stockfish_20121407 : 153 solved
Eman 6.61          : 153 solved
SugaR.AI.1.00      : 149 solved

The 3 first ones are new records if we omit the 2 tricky version of Eman and BlueFish who are very weak in play.
These 3 engines combined, find 193 (!!!) solutions !
If we add Crystal-2020-06-29, they find 199 solutions !
To solved 11 remaining positions, you need to add Sting-SF (more specialized in fortress detection and positions without progress). A winning pack of 5 to not overlook tricky tactics and singular shots.

It's still 3 positions not found with this very clever pack :

Code: Select all

1rb1qrk1/2b2pp1/p3pBn1/3pP1Pp/1ppP4/2P1QN2/PP3P1P/R2BR1K1 w - - bm Nh4; id "Hard-Talkchess-2020.023" ;
2bqrr1k/p5b1/1p1p2pp/nPpBp3/P1P1N2P/3PN1P1/R4P1K/3Q1R2 w - - bm Nc3; id "Hard-Talkchess-2020.038" ;
r2r2k1/pqp1bppp/1pn1pn2/8/2NP1B2/2N1P1P1/PPQ2P1P/R4RK1 b - - bm Rxd4; id "Hard-Talkchess-2020.179" ;

Some engines that will be tested in the following days : Crystal_x64_110920, Cfish NUMA 261020 and Sting-sf-25.

Link to the sheet of results : https://www.dropbox.com/scl/fi/rmt10d1y ... pnp1c6kvmq

Ferdy · Post by **Ferdy** » Mon Dec 28, 2020 7:27 am

Vinvin wrote: ↑Sun Dec 27, 2020 11:09 pm Results for 6 new engines (NN file used : nn-62ef826d1a6d.nnue) :
Code: Select all
SugaR.AI.ICCF.1.00 : 166 solved
ShashChess15.1     : 166 solved
Bluefish-12-W      : 162 solved
Stockfish_20121407 : 153 solved
Eman 6.61          : 153 solved
SugaR.AI.1.00      : 149 solved
The 3 first ones are new records if we omit the 2 tricky version of Eman and BlueFish who are very weak in play.
These 3 engines combined, find 193 (!!!) solutions !
If we add Crystal-2020-06-29, they find 199 solutions !
To solved 11 remaining positions, you need to add Sting-SF (more specialized in fortress detection and positions without progress). A winning pack of 5 to not overlook tricky tactics and singular shots.

It's still 3 positions not found with this very clever pack :
Code: Select all
1rb1qrk1/2b2pp1/p3pBn1/3pP1Pp/1ppP4/2P1QN2/PP3P1P/R2BR1K1 w - - bm Nh4; id "Hard-Talkchess-2020.023" ;
2bqrr1k/p5b1/1p1p2pp/nPpBp3/P1P1N2P/3PN1P1/R4P1K/3Q1R2 w - - bm Nc3; id "Hard-Talkchess-2020.038" ;
r2r2k1/pqp1bppp/1pn1pn2/8/2NP1B2/2N1P1P1/PPQ2P1P/R4RK1 b - - bm Rxd4; id "Hard-Talkchess-2020.179" ;
Some engines that will be tested in the following days : Crystal_x64_110920, Cfish NUMA 261020 and Sting-sf-25.

Link to the sheet of results : https://www.dropbox.com/scl/fi/rmt10d1y ... pnp1c6kvmq

How to interpret the table? I understand there are 213 positions in this test.

Hard-Talkchess-2020 set, final release

Re: Hard-Talkchess-2020 set, final release

Re: Hard-Talkchess-2020 set, final release

Re: Hard-Talkchess-2020 set, final release

Re: Hard-Talkchess-2020 set, final release

Re: Hard-Talkchess-2020 set, final release

Re: Hard-Talkchess-2020 set, final release

Re: Hard-Talkchess-2020 set, final release

Re: Hard-Talkchess-2020 set, final release

Re: Hard-Talkchess-2020 set, final release

Re: Hard-Talkchess-2020 set, final release