Why is Stockfish dominating the TCEC so far?

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.
Leo
Posts: 816
Joined: Fri Sep 16, 2016 4:55 pm
Location: USA/Minnesota
Full name: Leo

Re: Why is Stockfish dominating the TCEC so far?

Post by Leo » Sat Apr 07, 2018 10:05 pm

Informative comments. There was a lot of talk (and some whining) about computer chess having draw death in the last year.
Advanced Micro Devices fan.

carldaman
Posts: 1705
Joined: Sat Jun 02, 2012 12:13 am

Re: Why is Stockfish dominating the TCEC so far?

Post by carldaman » Sun Apr 08, 2018 12:32 am

Leo wrote:Informative comments. There was a lot of talk (and some whining) about computer chess having draw death in the last year.
The draw death problem existed only because the top three engines were very close in rating at the time.

A better question for me would be why Stockfish plays so much better. For me it is the superior search, which helps with its top-notch attacking ability, and that is now enhanced by the addition of contempt. Lately, a lot of evaluation factors have been added that may be plugging some holes in its eval.

On the downside, there are some occasional blind spots based on what I've seen posted, and the old untrustworthy 0.00 eval scores are still there.

User avatar
Laskos
Posts: 9408
Joined: Wed Jul 26, 2006 8:21 pm
Full name: Kai Laskos

Re: Why is Stockfish dominating the TCEC so far?

Post by Laskos » Sun Apr 08, 2018 12:56 am

carldaman wrote:
Leo wrote:Informative comments. There was a lot of talk (and some whining) about computer chess having draw death in the last year.
The draw death problem existed only because the top three engines were very close in rating at the time.
I am not following this TCEC, but if you see draw rates oscillating here and there due to openings, contempt and other factors, draw death will BE a problem in some future. If a domineering AlphaZero had IIRC +29 -0 =71 (71% draw rate) against Stockfish, a completely different paradigm, much weaker engine, then imagine the draw rate between AlphaZero and equal in strength super-Stockfish (probably draw rate above 80% in the same conditions), or AlphaZero self-games (probably above 90%). A much weaker LCZero has draw rates against normal engines of similar strength of about 10%, so the trend is extremely clear.
A better question for me would be why Stockfish plays so much better. For me it is the superior search, which helps with its top-notch attacking ability, and that is now enhanced by the addition of contempt. Lately, a lot of evaluation factors have been added that may be plugging some holes in its eval.

On the downside, there are some occasional blind spots based on what I've seen posted, and the old untrustworthy 0.00 eval scores are still there.

JJJ
Posts: 1285
Joined: Sat Apr 19, 2014 11:47 am

Re: Why is Stockfish dominating the TCEC so far?

Post by JJJ » Sun Apr 08, 2018 2:30 am

I hope we re gonna have a big surprise for the next TCEC , otherwise the next winner is very clear.

Seems lot of power helped a lot Stockfish making improvement lately. Is he now uncatchable ?

corres
Posts: 1558
Joined: Wed Nov 18, 2015 10:41 am
Location: hungary

Re: Why is Stockfish dominating the TCEC so far?

Post by corres » Sun Apr 08, 2018 6:24 am

[quote="Dann Corbit"]

Contempt against lesser opponents is worth maybe 15 Elo.
Contempt against peers has almost no value.
An eval tweak is likely more important than a good contempt setting.

[quote="Dann Corbit"]

I agree.
Nowadays for the developers of Stockfish the most important thing the enhancement of the chess knowledge of Stockfish. Earlier the main task was the increasing of search speed and depth.
When the developers of other engines use the results of the open source Stockfish the gap between them will diminish again.
Ones who support the development of Stockfish they support other engines, too.


And a search improvement is ten times more important.

SF is winning more because H held still and SF moved ahead.[/quote]

Post Reply