ShashChess

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

Moderators: hgm, Rebel, chrisw

Eduard
Posts: 1439
Joined: Sat Oct 27, 2018 12:58 am
Location: Germany
Full name: N.N.

Re: ShashChess

Post by Eduard »

Dear Andrea! I noticed that ShashChess creates a learning file called experience.bin. But if I switch on MCTS (the setting Single) then experience.bin becomes experience_new.bin. When MCTS is switched off again, the learning file is called experience.bin again. Why this?
Eduard
Posts: 1439
Joined: Sat Oct 27, 2018 12:58 am
Location: Germany
Full name: N.N.

Re: ShashChess

Post by Eduard »

The engine plays impressive and very interesting. I played tournaments on PlayChess and InfinityChess for 2 days. MCTS was set for single. Yesterday on Infinity +1 and 5th place and in the Bullet on PlayChess 3rd place. Today on InfinityChess also +1 and 3rd place, tied with the winner. With 23.000 kns (start position) that is a very good result.
amchess
Posts: 323
Joined: Tue Dec 05, 2017 2:42 pm

Re: ShashChess

Post by amchess »

ShashChess 17.1
Solved an important bug on the search for Shashin theory.
The engine now is a lot stronger.
https://github.com/amchess/ShashChess/wiki/Matches
Aligned with Stockfish patch
Author: candirufish Date: Mon May 31 18:31:32 2021 +0200 Timestamp: 1622478692 Check Extension with Static Evaluation extension for checking moves, at higher depth and more decisive positions.
https://github.com/amchess/ShashChess/releases/tag/17.1
amchess
Posts: 323
Joined: Tue Dec 05, 2017 2:42 pm

Re: ShashChess

Post by amchess »

An extraordinary result also on hard positions:
https://github.com/amchess/ShashChess/w ... te-results
Paloma
Posts: 1167
Joined: Thu Dec 25, 2008 9:07 pm
Full name: Herbert L

Re: ShashChess

Post by Paloma »

What means eg. 1q1r3k/3P1pp1/ppBR1n1p/4Q2P/P4P2/8/5PK1/8 w - - bm Rxf6 :
the numbers 64 and 88 in your spreadsheet ?

Rf6 (64) 1 Rf6 (88)

For other positions there are also 100 and 100 for both engines displayed, so it whould not sec. ?
peter
Posts: 3186
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: ShashChess

Post by peter »

amchess wrote: Wed Jun 02, 2021 5:28 pm An extraordinary result also on hard positions:
https://github.com/amchess/ShashChess/w ... te-results
Looking forward to an update of Vincent Lejeune's tests with the small 114- positions subset of his Hard Talkchess suite:

http://talkchess.com/forum3/viewtopic.p ... 67#p893667

At the latest sheet I know

https://www.dropbox.com/s/qfz8u1oqmxodx ... s.ods?dl=0

ShashChess 16 was best version still.

Which setting do you prefer for such tactical best move suites, Andrea?

Tal only, Tal+Capablanca or default setting (all GM- options unchecked)?
MCTS?
Full depth threads?

Thanks for the update regards
Peter.
Eduard
Posts: 1439
Joined: Sat Oct 27, 2018 12:58 am
Location: Germany
Full name: N.N.

Re: ShashChess

Post by Eduard »

I won today with ShashChess 17.1 the Bullet Tour 19 rounds on PlayChess (Only with 12 Cores). :lol:

Here are the first 3:

Code: Select all

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19       
1 SolistaChess,ShashChess 17.1  2480 ◊ ½/3 ◊ 1/15 ♦ ½/4 ♦ ½/5 ◊ ½/2 ♦ ½/7 ◊ ½/6 ♦ ½/10 ♦ ½/8 ◊ ½/11 ◊ ½/5 ♦ ½/4 ♦ ½/6 ◊ 1/9 ◊ 1/3 ♦ ½/2 ◊ ½/7 ◊ 1/11 ♦ ½/10 11.5 / 19     
2 HerrBert,SugaR AI 2.10 sse41- 2386 ◊ ½/6 ♦ ½/14 ♦ ½/7 ◊ 1/11 ♦ ½/1 ◊ ½/4 ♦ ½/3 ◊ ½/5 ◊ ½/10 ♦ ½/9 ♦ ½/7 ◊ ½/10 ◊ ½/5 ♦ 1/8 ♦ ½/4 ◊ ½/1 ♦ ½/3 ◊ ½/6 ♦ ½/8  10.5 / 19     
3 Trampas,Stockfish 190521      2430 ♦ ½/1 ◊ ½/11 ◊ ½/5 ♦ ½/4 ♦ ½/10 ◊ 1/12 ◊ ½/2 ♦ ½/9 ♦ ½/6 ◊ ½/8 ◊ ½/4 ♦ ½/5 ♦ 1/10 ◊ ½/7 ♦ 0/1 ♦ ½/8 ◊ ½/2 ♦ ½/7 ◊ ½/11 10.0 / 19 103.00
My Settings: Capablanca and MCTS=Single
Vinvin
Posts: 5228
Joined: Thu Mar 09, 2006 9:40 am
Full name: Vincent Lejeune

Re: ShashChess

Post by Vinvin »

peter wrote: Wed Jun 02, 2021 9:02 pm
amchess wrote: Wed Jun 02, 2021 5:28 pm An extraordinary result also on hard positions:
https://github.com/amchess/ShashChess/w ... te-results
Looking forward to an update of Vincent Lejeune's tests with the small 114- positions subset of his Hard Talkchess suite:
...
I plan to run it very soon :wink:
peter
Posts: 3186
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: ShashChess

Post by peter »

Vinvin wrote: Thu Jun 03, 2021 1:40 am
peter wrote: Wed Jun 02, 2021 9:02 pm
amchess wrote: Wed Jun 02, 2021 5:28 pm An extraordinary result also on hard positions:
https://github.com/amchess/ShashChess/w ... te-results
Looking forward to an update of Vincent Lejeune's tests with the small 114- positions subset of his Hard Talkchess suite:
...
I plan to run it very soon :wink:
Thank you!
Which one setting will you choose? Default, I guess?
Just asking because of comparability to my SMP- run regards
Peter.
Eduard
Posts: 1439
Joined: Sat Oct 27, 2018 12:58 am
Location: Germany
Full name: N.N.

Re: ShashChess

Post by Eduard »

There is a new Eman 7.22. Eman 7.22 now analyzes exactly as he plays. I compared ShashChess 17.1 with Eman 7.22 with some positions. The network is the new green NNUE nn-7e66505906a6.nnue, and both engines analyzed without learning file. Eman with default settings, ShashChess with Style Capablanca and MCTS=Single.

It seems that Eman 7.22 calculates more selectively and sometimes overlooks something. There are positions where both programs are about the same speed, but there are clear differences in some.

Here are some examples under Ryzen 3900X with 22 threads (hash = 6 GB):

Diagram:
[d]5b2/5k1p/1pb2pp1/1r1p4/3P1B1P/P7/1P3PP1/2R3K1 w - - 0 1
Best move 28. b2-b4!

Analysis by Eman 7.22 64-bit AVX2:

28.Rxc6 Rxb2 29.Rc7+ Ke6 30.Rxh7 Bxa3 31.Rg7 Kf5 32.Be3 b5 33.g4+ Kxg4 34.Rxg6+ Kf5 35.h5 Rc2 36.Rg8 b4 37.h6 Rc7 38.Rg7 Rc8 39.Rb7 Ke4 40.Kf1 Bc1 41.Bxc1 Rxc1+ 42.Kg2 Rc8 43.Rxb4 Rh8 44.f3+ Ke3 45.Kg3 Rxh6 46.Kg4 Rg6+ 47.Kf5 Rg5+ 48.Kxf6 Rh5 49.Ra4 Kxf3
The position is equal: = (0.00) Depth: 54/45 00:00:57 1425MN, tb=2286634
28.b4
The position is equal: = (0.06 ++) Depth: 55/45 00:01:19 1966MN, tb=3351238
28.b4
The position is equal: = (0.13 ++) Depth: 55/45 00:01:23 2057MN, tb=3398074
28.b4
White has an edge: = (0.23 ++) Depth: 55/45 00:01:25 2131MN, tb=3431623

Analysis by ShashChess 17.1:

28.Rxc6 Rxb2 29.Rc7+ Ke6 30.Rxh7 Bxa3 31.Rg7 Rb1+ 32.Kh2 Bb2 33.Be3 Rd1 34.Rxg6 Bxd4 35.Bxd4 Rxd4 36.Kg3 Kf7 37.h5 b5 38.Kf3 Rc4 39.g4 b4 40.g5 fxg5 41.Rxg5 b3 42.h6 Rb4 43.h7 Rb8 44.h8Q Rxh8 45.Rxd5
The position is equal: = (0.00) Depth: 40/36 00:00:10 193MN, tb=180500
28.b4
The position is equal: = (0.08 ++) Depth: 41/38 00:00:13 257MN, tb=231107
28.b4
The position is equal: = (0.16 ++) Depth: 41/38 00:00:13 273MN, tb=233737
28.b4
White has an edge: = (0.28 ++) Depth: 41/38 00:00:14 287MN, tb=237084
28.b4
White is slightly better: +/= (0.46 ++) Depth: 41/38 00:00:15 308MN, tb=238908

Diagram:
[d]r5rk/1bq1bpp1/1p1ppn1p/pP5P/2P1P3/P1N2P2/1B1Q1P2/1K1R1BR1 w - - 0 1
Best move 19. Nd5

Analysis by Eman 7.22 64-bit AVX2:

19.Qe3 e5 20.Bd3 Nd7 21.f4 Qc5 22.Qh3 Rad8 23.Nd5 Bf6 24.Qf3 Rde8 25.Nxf6 Nxf6 26.f5 Rc8 27.Qe2 Nh7 28.Bc1 Ng5 29.Rg3 f6 30.f3 Rgd8 31.Ka2 Qc7 32.Be3 Rb8 33.Qd2 Bc8 34.Bf1 Bb7 35.Ka1 Bc8 36.a4 Bb7 37.Bd3 Bc8 38.Be2 Bb7 39.Kb2 Bc8 40.Kb1 Bb7 41.Bd3 Bc8 42.Bc2 Qxc4 43.Kb2
White is clearly better: +- (2.00) Depth: 35/49 00:01:21 1782MN, tb=3331
19.Nd5
White is clearly better: +- (2.08 ++) Depth: 36/34 00:01:33 2041MN, tb=3398
19.Nd5
White is clearly better: +- (2.16 ++) Depth: 36/34 00:01:34 2064MN, tb=3400
19.Nd5
White is clearly better: +- (2.29 ++) Depth: 36/34 00:01:35 2087MN, tb=3404
19.Nd5
White is clearly better: +- (2.47 ++) Depth: 36/34 00:01:36 2120MN, tb=3406

Analysis by ShashChess 17.1:

19.Bd3 e5 20.f4 Nd7 21.f3 Rac8 22.fxe5 dxe5 23.Nd5 Bxd5 24.exd5 Bc5 25.Rge1 Bd4 26.Bxd4 exd4 27.Bf5 Ne5 28.Rxe5 Qxe5 29.Bxc8 Rxc8 30.Qxd4 Qxh5 31.Qd3 Rd8 32.d6 Kg8 33.Kb2 Qe5+ 34.Qc3 Qc5 35.Rd5 Qf2+ 36.Kb3 h5 37.d7 h4
White is clearly better: +- (1.96) Depth: 31/45 00:00:45 881MN, tb=91
19.Nd5
White is clearly better: +- (2.04 ++) Depth: 32/47 00:01:00 1206MN, tb=179
19.Nd5
White is clearly better: +- (2.12 ++) Depth: 32/47 00:01:01 1217MN, tb=179
19.Nd5
White is clearly better: +- (2.25 ++) Depth: 32/47 00:01:01 1222MN, tb=185
19.Nd5
White is clearly better: +- (2.43 ++) Depth: 32/47 00:01:01 1224MN, tb=185
19.Nd5
White has a decisive advantage: +- (2.67 ++) Depth: 32/47 00:01:02 1238MN, tb=288
19.Nd5
White has a decisive advantage: +- (3.00 ++) Depth: 32/47 00:01:03 1262MN, tb=396

Diagram:
[d]4r3/2p3kp/R1Pp1q2/3P2p1/1P2rp2/7P/1Q3RP1/6K1 w - - 0 1
Best move 1. Qxf6+!

Analysis by Eman 7.22 64-bit AVX2:

1.Ra7 Re1+ 2.Kh2 Qxb2 3.Rxc7+ Kg6 4.Rxb2 h5 5.Rf2 Rc1 6.Rd7 Kf5 7.g3 Ree1 8.gxf4 Rh1+ 9.Kg2 Rcg1+ 10.Kf3 Rxh3+ 11.Ke2 g4 12.Kd2 Ra1 13.c7 Ra2+ 14.Ke1 Rh1+ 15.Rf1 Rhh2 16.Rf2 Ra1+ 17.Ke2 Ra2+
Black is slightly better: =/+ (-0.52 --) Depth: 44/43 00:02:36 3918MN, tb=3837081
1.Qxf6+
Black is slightly better: =/+ (-0.36 ++) Depth: 44/58 00:03:03 4620MN, tb=4685735

Analysis by ShashChess 17.1:

1.Ra7 Re1+ 2.Kh2 R8e5 3.Rxc7+ Kh6 4.Qc2 R5e4 5.Qb2
Black is slightly better: =/+ (-0.47 --) Depth: 23/35 00:00:06 99206kN, tb=19694
1.Qxf6+ Kxf6 2.Ra7 Rc8 3.b5 Rb4 4.Re2 Rxb5 5.Re6+ Kf5 6.Re7 Rxd5 7.Raxc7 Rxc7 8.Rxc7 Rc5 9.Kf2 Rc2+ 10.Kf3 h6 11.Rc8 h5 12.Rf8+ Ke5 13.Rg8 Rc3+ 14.Ke2 Ke4 15.Rxg5 h4 16.Rg6 Rc2+ 17.Ke1 Rxc6 18.Re6+ Kf5 19.Rh6 Rc2 20.Rxd6
Black is slightly better: =/+ (-0.57) Depth: 23/39 00:00:06 106MN, tb=22262
(---)
1.Qxf6+ Kxf6 2.Ra7 Rc8 3.b5 Rb4 4.Re2 Rxb5 5.Re6+ Kf5 6.Re7 Rxd5 7.Raxc7 Rxc7 8.Rxc7 Rc5 9.Kf2 Rc2+ 10.Kf3 Rc3+ 11.Ke2 Ke5 12.Re7+ Kf6 13.Rc7 h6 14.Rc8 Kg6 15.c7 Kh5 16.Kf2 d5 17.Rd8 Rxc7 18.Rxd5 Rc2+ 19.Kf3 Kh4 20.Rd8 Rc5
Black is slightly better: =/+ (-0.33) Depth: 31/40 00:00:16 342MN, tb=135277

I think ShashChess 17.1 is a very good engine for analysis, but I also won on PlayChess with it. The Setting Capablanca & MCTS=Single is great.