Hard-Talkchess-2020 set, final release

[peter]

Ng4 is easy to find, so the best thing to do is to remove the position.

100. HTC.185,r3r1k1/pp1q1p2/2p2npb/PPPp4/3Pp3/2NNPpP1/1R1B1P1K/3Q3R b - - bm Kg7
seems to have a dual (or almost a dual) with Ng4+ !?

its worth to reconsider Kg7 as the only best move imo !?

I left it in the suite and just count Ng4 as solved too, after all it's just transposing to Kg7 anyhow.
New Crystal is very strong in my personal SMP MultiPV=4 list.
Same old hardware of 12x3GHz Xeon CPU (popcnt compile) 8G hash, latest green SF NNUE, 1min/pos, MultiPV=4, 6men Syzygys

Crystal 3.2
Bisher gelöst: 88 von 114 ; 43:32m

Code: Select all

         1   2   3   4   5   6   7   8   9  10  11  12  13  14  15  16  17  18  19  20
 -------------------------------------------------------------------------------------
   0 |   2   1   -   -  43   4   0  15  25   1  31   -  35   0  14   0   6   9   2  37
  20 |   7   -   -   7   1   1   1  26   -   0   7  26   0  30  14   1   1   -   -   3
  40 |   -   0  39   1  48  57   0   1  15   2   -   1   -  45   3  13  60   1   6   0
  60 |   0   -   3   -   -   -   7   5  25   0   -   0   0   0   4   0  11   0   0  26
  80 |   0   0   -   1   -  13   0   6   -   0  10   -  60  13   -   2  37   1   -   0
 100 |   -   3   7   -   0   -   2   1  28  23   9   1   -  60

[Vinvin]

...
Crystal 3.2
Bisher gelöst: 88 von 114 ; 43:32m

Nice !
Where can I find Crystal 3.2 ?

[peter]

...
Crystal 3.2
Bisher gelöst: 88 von 114 ; 43:32m

Nice !
Where can I find Crystal 3.2 ?

Here:

https://github.com/jhellis3/Stockfish/releases/tag/v3.2

A fine compile I got at Ipman's site:

http://ipmanchess.yolasite.com/compiles.php

[Vinvin]

Some new results :
Crystal 3.2 : 89 solved
ShashChess17.1 with nn-7756374aaed3 : 89 solved
ShashChess17.1 with nn-8a08400ed089 : 85 solved
Stockfish_21061522 : 77 solved
Stockfish_21061411 : 79 solved

Current top engines + SF :

Code: Select all

On 114 positions ->
Crystal 3.2        : 89 solved
SugaR.AI.ICCF.1.90 : 89 solved
ShashChess17.1-7756374aaed3 : 89 solved
ShashChess 16      : 87 solved
ShashChess 17      : 86 solved
Crystal3.1         : 86 solved
Bluefish-v13.1     : 85 solved
Bluefish-v13       : 84 solved
Black-Diamond-v13  : 80 solved
SugaR.AI.ICCF.1.00 : 79 solved
Stockfish_21061411 : 79 solved
ShashChess15.1     : 78 solved
Stockfish_21051920 : 75 solved 
Stockfish 13       : 67 solved

The SugaR AI ICCF with the new (bigger) net will probably overpass all of them !
There are a lof of small improvement on the Stockfish side ( https://tests.stockfishchess.org/tests ) for a couple of weeks, this will probably lead to improvement in the test.

Note there's a comparison between Fat Fritz 2 and SF on a couple of test positions here : https://en.chessbase.com/post/solving-the-impossible

[peter]

Thanks for the new results!

The SugaR AI ICCF with the new (bigger) net will probably overpass all of them !

Pity Marco Zerbinati closed the project end of march.

http://outskirts.altervista.org/forum/v ... ost#p28694

[Vinvin]

Thanks for the new results!

The SugaR AI ICCF with the new (bigger) net will probably overpass all of them !

Pity Marco Zerbinati closed the project end of march.

http://outskirts.altervista.org/forum/v ... ost#p28694

You tell me bad news

[Vinvin]

Results for Stockfish 14 : 79 solved (same as Stockfish_21061411)

Current top engines + SF :

Code: Select all

On 114 positions ->
Crystal 3.2        : 89 solved
SugaR.AI.ICCF.1.90 : 89 solved
ShashChess17.1-7756374aaed3 : 89 solved
Crystal3.1         : 86 solved
Bluefish-v13.1     : 85 solved
Black-Diamond-v13  : 80 solved
SugaR.AI.ICCF.1.00 : 79 solved
Stockfish 14       : 79 solved
Stockfish 13       : 67 solved

I cleaned the sheet by removing some duplicates engines : https://www.dropbox.com/s/584upipsovf2p ... 4.ods?dl=0
Note : you can download the sheet if it doesn't appear well in your browser.

[peter]

Results for Stockfish 14 : 79 solved (same as Stockfish_21061411)

Current top engines + SF :

Code: Select all

On 114 positions ->
Crystal 3.2        : 89 solved
SugaR.AI.ICCF.1.90 : 89 solved
ShashChess17.1-7756374aaed3 : 89 solved
Crystal3.1         : 86 solved
Bluefish-v13.1     : 85 solved
Black-Diamond-v13  : 80 solved
SugaR.AI.ICCF.1.00 : 79 solved
Stockfish 14       : 79 solved
Stockfish 13       : 67 solved

I cleaned the sheet by removing some duplicates engines : https://www.dropbox.com/s/584upipsovf2p ... 4.ods?dl=0
Note : you can download the sheet if it doesn't appear well in your browser.

Maybe you'd like to give Dietrich Kappe's new Dark Horse 0.3bxl.nnue a try, scores best in my SMP MultiPV- list with Crystal solving 92/114.

http://talkchess.com/forum3/viewtopic.p ... 31#p899331

[Vinvin]

Position #144 is probably not good enough.
3 moves are clearly winning.
[d]r1b1qk1r/pn3pp1/1p2p3/3pP1Bp/1bPN4/4Q1R1/P2N1PP1/R5K1 w - - bm Bf6; id "Hard-Talkchess-2020.144" ;

Code: Select all

Stockfish_21081512_x64_avx2:
...
 44/70	13:21	 18.588.671k	23.191k	+1,75	1.Rd1 Nc5 2.Nb5 Qd7 3.Rf3 Bb7 4.Nf1 Na6 5.Qf4 Kg8 6.Ng3 Bf8 7.Qh4 Rh7 8.Nxh5 f5 9.exf6 Qf7 10.fxg7 Qxh5 11.Qxh5 Rxh5 12.gxf8R+ Rxf8 13.Rg3 Kh7 14.cxd5 Bxd5 15.Nc3 Nc5 16.Nxd5 exd5 17.Rg4 Rf5 18.Be3 Re5 19.Rgd4 Ne6 20.R4d3 Kg6 21.g3 Nc5 22.Rc3 Ne6 23.Kg2 Re4 24.Rcd3 Ra4 25.f4 Rxa2+ 26.Kf3 Ng7 27.Rxd5 Nf5 28.Bf2 Rh2 29.R1d2 Ra3+ 30.R5d3 Ra4 31.Rc3 Kf6 32.Rc6+ Ke7 33.Bc5+ bxc5 34.Rc7+ Ke6 35.Rxh2 Ra3+ 36.Kf2
 44/80	13:21	 18.588.671k	23.191k	+3,00	1.Re1 Kg8 2.Bf6 Bf8 3.cxd5 exd5 4.e6 Bxe6 5.Bxg7 Bg4 6.Qxe8 Rxe8 7.Rxe8 Kxg7 8.Nf5+ Kf6 9.Ne7 Kg7 10.Nxd5 Bd6 11.Rxh8 Kxh8 12.Re3 Be6 13.Nf6 Kg7 14.Ne8+ Kf8 15.Nxd6 Nxd6 16.a3 Ke7 17.Re5 h4 18.f3 Nf5 19.Ne4 a6 20.Ng5 Kf6 21.f4 Bd7 22.Nf3 a5 23.Kf2 b5 24.Rc5 Ke6 25.Ng5+ Ke7 26.Re5+ Kf6 27.Nf3 a4 28.Rc5 Ke6 29.Ng5+ Kd6 30.Ne4+ Ke7 31.Rc7 Ke6 32.Nc5+ Kd6 33.Rxd7+ Kxc5 34.Rxf7 Nd6 35.Rc7+ Kd4
 44/76	13:21	 18.588.671k	23.191k	+4,44	1.N2f3 Kg8 2.Bf6 Bf8 3.Nb5 Bd7 4.Nc7 Qc8 5.Nxa8 Nc5 6.Ng5 Qxa8 7.Rd1 Qc6 8.Qf3 Be8 9.cxd5 exd5 10.Rxd5 h4 11.Rg4 Nd7 12.Qd1 Nxf6 13.exf6 Qxf6 14.Rd8 Bc6 15.Qd4 Qxd4 16.Rgxd4 f6 17.Ne6 Kf7 18.Rc8 Bc5 19.Rdd8 Re8 20.Rxe8 Bxe8 21.Nxc5 bxc5 22.Rxc5 a6 23.Rc7+ Kg8 24.Ra7 Bb5 25.Kh2 g5 26.Kh3 Bc4 27.a4 Bf1 28.f3 Kf8 29.Kh2 Ke8 30.Kg1 Bc4 31.Kf2 Kf8 32.Rb7 Bd3 33.Rc7 Bb5 34.axb5 axb5 35.Rb7 b4 36.Kg1 Ke8 37.Kh2 b3 38.Rxb3
 44/87	13:21	 18.588.671k	23.191k	+5,01	1.Bf6 gxf6 2.exf6 Qd7 3.N2f3 Nd6 4.Ne5 Qc7 5.Rg7 Nxc4 6.Nxc4 dxc4 7.Rd1 Bd7 8.Qg5 Qd6 9.Rxf7+ Kxf7 10.Qg7+ Ke8 11.f7+ Kd8 12.Qxh8+ Ke7 13.Qxa8 Kxf7 14.Qf3+ Ke7 15.Qxh5 Qd5 16.Qh4+ Ke8 17.Qg4 Bc3 18.Nb3 Qf5 19.Qxc4 Be5 20.Qe2 Bc6 21.Nd4 Bxd4 22.Rxd4 Bd5 23.a4 Kd7 24.f3 Qb1+ 25.Rd1 Qb3 26.Qb5+ Kc7 27.Qxb3 Bxb3 28.Ra1 Bd5 29.Kf2 a6 30.g4 b5 31.axb5 axb5 32.Rb1 Bc4 33.Ke3 Kd6

[fishpov]

Hi,
Did you try SugaR AI 2.30 ICCF ?
https://workupload.com/file/hgx4B7GHyzh

Thanks