SPCC: Testrun of Stockfish 171206 with Contempt=+15 finished

Discussion of computer chess matches and engine tournaments.

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
pohl4711
Posts: 1042
Joined: Sat Sep 03, 2011 5:25 am
Location: Berlin, Germany
Contact:

SPCC: Testrun of Stockfish 171206 with Contempt=+15 finished

Post by pohl4711 » Thu Dec 21, 2017 12:16 pm

Testrun of Stockfish 171206 with Contempt=+15 finished.

I will do 3 testruns of Stockfish 171206 with 3 different contempts (Contempt=+15, +25, +40). Lets see, if the new contempt-algorithm of Stockfish can gain some Elo...and will lower the draw-rate and the number of 3fold-draws. Stay tuned!

http://www.sp-cc.de/experiments.htm


(Perhaps you have to clear your browsercache or reload the website)

Jouni
Posts: 1960
Joined: Wed Mar 08, 2006 7:15 pm

Re: SPCC: Testrun of Stockfish 171206 with Contempt=+15 fini

Post by Jouni » Thu Dec 21, 2017 2:31 pm

Thanks for test. So c=15 is clear ELO gain vs. Fire and Shredder, but neutral against Houdini, Komodo and Fizbo(!).
Jouni

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

Re: SPCC: Testrun of Stockfish 171206 with Contempt=+15 fini

Post by JJJ » Fri Dec 22, 2017 11:38 am

This means contempt 15 is still in the error bar and doesn't seems to really help much, even vs weaker opponent.

pohl4711
Posts: 1042
Joined: Sat Sep 03, 2011 5:25 am
Location: Berlin, Germany
Contact:

Re: SPCC: Testrun of Stockfish 171206 with Contempt=+15 fini

Post by pohl4711 » Fri Dec 22, 2017 12:51 pm

JJJ wrote:This means contempt 15 is still in the error bar and doesn't seems to really help much, even vs weaker opponent.
Correct. But mention, that in a ratinglist, Stockfish is tested vesus many and much weaker engines, than in my testruns. It is possible, that there would be a bigger Elo-gain in an FGRL, CEGT or IPON Testrun...
And we see in my testing results, that the contempt measureable lowers the number of draws and the 3fold-draws.
I am looking forward to the testing results with Contenpt +25 and +40.

Stefan (SPCC)

Carlos777
Posts: 644
Joined: Sun Dec 13, 2009 5:09 pm

Re: SPCC: Testrun of Stockfish 171206 with Contempt=+15 fini

Post by Carlos777 » Fri Dec 22, 2017 2:37 pm

pohl4711 wrote:Testrun of Stockfish 171206 with Contempt=+15 finished.
Interesting test, thank you.

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

Re: SPCC: Testrun of Stockfish 171206 with Contempt=+15 fini

Post by JJJ » Sat Dec 23, 2017 10:26 am

pohl4711 wrote:
JJJ wrote:This means contempt 15 is still in the error bar and doesn't seems to really help much, even vs weaker opponent.
Correct. But mention, that in a ratinglist, Stockfish is tested vesus many and much weaker engines, than in my testruns. It is possible, that there would be a bigger Elo-gain in an FGRL, CEGT or IPON Testrun...
And we see in my testing results, that the contempt measureable lowers the number of draws and the 3fold-draws.
I am looking forward to the testing results with Contenpt +25 and +40.

Stefan (SPCC)
In a rating list maybe, but the purpose was to win more in closer encounter than in larger encounter in event like TCEC.

Some others test suggest than contempt 20 might be the best contempt for that.

Post Reply