Stockfish 10 beta

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.
Damir
Posts: 1918
Joined: Mon Feb 11, 2008 2:53 pm

Stockfish 10 beta

Post by Damir » Mon Nov 19, 2018 11:30 am

has just been released.

Link: http://abrok.eu/stockfish

mibere
Posts: 6
Joined: Sun Nov 12, 2017 3:14 pm
Location: Germany
Contact:

Re: Stockfish 10 beta

Post by mibere » Mon Nov 19, 2018 6:36 pm

For macOS also available on https://stockfishchess.de

Eduard
Posts: 28
Joined: Fri Oct 26, 2018 10:58 pm
Location: Germany
Full name: Eduard Nemeth
Contact:

Re: Stockfish 10 beta

Post by Eduard » Mon Nov 19, 2018 11:51 pm

Thank You!

Here is a little testposition from latest WCC game today:

Caruana,F - Carlsen,M 1/2-1/2, World Chess Championship 2018


Analysis by Stockfish 191118 64 POPCNT:

23.Tad1 Ld6 24.Dh5 De8 25.Dxg5+ Dg6 26.Dxg6+ hxg6 27.Sc4 Tbd8 28.Sxd6 Txd6 29.Le5 Tdd8 30.Txf4 Tfe8 31.Lc3 Te4 32.Txe4 Lxe4 33.d6 Kf7 34.Te1 Lf5 35.Te7+ Kf8 36.Txb7 Txd6 37.Kf2 g5 38.Tc7 Td5 39.Ke3 Td3+ 40.Ke2 Td5 41.Kf2 Ke8 42.Kf3 Ld3 43.Ke3 Lf1 44.g3 Ld3 45.Ta7 Lf1 46.Tg7 Kf8 47.Tc7 Ke8 48.Kf3
+- (1.83 ++) Tiefe: 43/58 00:09:47 1622MN, tb=12193

Much better move here is 23.Rae1! with clear better chances to win. 23.Rad1 is good too, but not the best way. On my smartphone Stockfish 9 finds very quickly 23. Rae1! Why not Stockfish 10 beta too?

User avatar
Guenther
Posts: 2515
Joined: Wed Oct 01, 2008 4:33 am
Location: Regensburg, Germany
Full name: Guenther Simon
Contact:

Re: Stockfish 10 beta

Post by Guenther » Tue Nov 20, 2018 12:00 am

Eduard wrote:
Mon Nov 19, 2018 11:51 pm
Thank You!

Here is a little testposition from latest WCC game today:

Caruana,F - Carlsen,M 1/2-1/2, World Chess Championship 2018


Analysis by Stockfish 191118 64 POPCNT:

23.Tad1 Ld6 24.Dh5 De8 25.Dxg5+ Dg6 26.Dxg6+ hxg6 27.Sc4 Tbd8 28.Sxd6 Txd6 29.Le5 Tdd8 30.Txf4 Tfe8 31.Lc3 Te4 32.Txe4 Lxe4 33.d6 Kf7 34.Te1 Lf5 35.Te7+ Kf8 36.Txb7 Txd6 37.Kf2 g5 38.Tc7 Td5 39.Ke3 Td3+ 40.Ke2 Td5 41.Kf2 Ke8 42.Kf3 Ld3 43.Ke3 Lf1 44.g3 Ld3 45.Ta7 Lf1 46.Tg7 Kf8 47.Tc7 Ke8 48.Kf3
+- (1.83 ++) Tiefe: 43/58 00:09:47 1622MN, tb=12193

Much better move here is 23.Rae1! with clear better chances to win. 23.Rad1 is good too, but not the best way. On my smartphone Stockfish 9 finds very quickly 23. Rae1! Why not Stockfish 10 beta too?
You must have missed that this position was already posted today, just a few topics below.
23. g4! is winning... (has a much higher score - probably analysed with backward, forward hash filling >+4)

http://talkchess.com/forum3/viewtopic.php?f=2&t=68976

Eduard
Posts: 28
Joined: Fri Oct 26, 2018 10:58 pm
Location: Germany
Full name: Eduard Nemeth
Contact:

Re: Stockfish 10 beta

Post by Eduard » Tue Nov 20, 2018 12:15 am

Thank You!

Nevertheless, here Rae1 is better than Rad1, and Stockfish 9 (Smartphone version) finds. A diagram pictures (live from game with the Playchess app) can you see here on my homepage: https://en-chess.jimdofree.com/wcc2018/

Thanks anyway
Best wishes!

schack
Posts: 113
Joined: Thu May 27, 2010 1:32 am
Contact:

Re: Stockfish 10 beta

Post by schack » Tue Nov 20, 2018 12:42 am

For those of us who plan on compiling our own, any tips / tricks / flag configurations to maximise speed? Using MinGW + GCC 8.1

Eduard
Posts: 28
Joined: Fri Oct 26, 2018 10:58 pm
Location: Germany
Full name: Eduard Nemeth
Contact:

Re: Stockfish 10 beta

Post by Eduard » Tue Nov 20, 2018 12:44 am

Can someone explain me? I have set the contempt to 0. Now Stockfish 10 beta plays a little bit better.
It seems that contempt 24 is to much. :)

Caruana,F - Carlsen,M 1/2-1/2, World Chess Championship 2018
1r1q1rk1/1p2b2p/pN6/P1pP1bp1/5p2/2B2Q2/1P4PP/R4RK1 w - - 0 1

Analysis by Stockfish 191118 64 POPCNT:

23.Tae1 Lf6 24.h4 Lxc3 25.bxc3 g4 26.Dxf4 Dxh4 27.Dh2 Dxh2+ 28.Kxh2 Tf7 29.Te5 Lg6 30.Td1 Tf5 31.Te7 Tf7 32.Tee1 Tf2 33.d6 Td8 34.d7 Kf7 35.Te5 Tf5 36.Te3 Kg7 37.Kg3 h5 38.Tde1 Tf7 39.Kh4 Lf5 40.Te8 Tdxd7 41.Sxd7 Lxd7 42.T8e7 Kf6
+/- (1.60 ++) Tiefe: 31/47 00:01:05 189MN, tb=668
23.Tae1 Lf6 24.h4 Lxc3 25.bxc3 g4 26.Dxf4 Dxh4 27.Dh2 Dxh2+ 28.Kxh2 Tf7 29.Te5 Lg6 30.Td1 Tf5 31.Te7 Tf7 32.Tee1 Tf2 33.d6 Td8 34.d7 Kf7 35.Te5 Tf5 36.Te3 Kg7 37.Kg3 h5 38.Tde1 Tf7 39.Kh4 Lf5 40.Te8 Tdxd7 41.Sxd7 Lxd7 42.T8e7 Kf6
+- (1.69 ++) Tiefe: 31/47 00:01:09 200MN, tb=698

Dann Corbit
Posts: 9059
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: Stockfish 10 beta

Post by Dann Corbit » Tue Nov 20, 2018 1:01 am

Contempt has value when you have an inferior opponent. Suppose, for instance, that you come out of an opening book and your opponent is a little bit ahead of you (perhaps a quarter of a pawn). If you have an equal opponent, and you can find nothing better, it is a good idea to strive for a draw.

But suppose that your opponent is much weaker than you are. Now it is probably not a good idea to strive for a draw because you will probably win the game if you play on and avoid a draw.

So if (for instance) you have contempt on during self-play, you will see absurd things like both sides being ahead at the same time, according to the evaluation.

Unfortunately, with the standard Stockfish, you cannot turn off all of the contempt, but only part of it.

So I modify Stockfish so that I can use it for analysis and get the right answer.

Now, there is clearly a real affect of having contempt when it comes to game play. So it is a good idea for a chess engine. But when analyzing, it is telling you a small lie, which is bad.

IMO-YMMV
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.

Eduard
Posts: 28
Joined: Fri Oct 26, 2018 10:58 pm
Location: Germany
Full name: Eduard Nemeth
Contact:

Re: Stockfish 10 beta

Post by Eduard » Tue Nov 20, 2018 1:30 am

Thank You!

User avatar
Rebel
Posts: 4192
Joined: Thu Aug 18, 2011 10:04 am

Re: Stockfish 10 beta

Post by Rebel » Thu Nov 22, 2018 6:30 am

Played 3 matches with the latest development version against SF9

Code: Select all

  Match      games    Time   Perc  ELO
sf10 - sf9  10,000  15+0.1  60.2%  +72
sf10 - sf9  10,000  30+0.2  59.9%  +69
sf10 - sf9  10,000  60+0.4  59.5%  +67
Impressive!
People stalk because they unconsciously love you.

Post Reply