Elo measurement of contempt in SF in self-play.

Discussion of chess software programming and technical issues.

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
Michel
Posts: 1964
Joined: Sun Sep 28, 2008 11:50 pm

Elo measurement of contempt in SF in self-play.

Post by Michel » Sat Mar 10, 2018 8:15 am

http://tests.stockfishchess.org/tests/v ... 0297cb64af

Currently (after 180.000 games) it is 3.1 +- 1.0.

So the common wisdom that contempt cannot possibly work in self-play (an opinion that has been voiced by many people, including myself) seems to have been soundly refuted.... At least for SF.

Now one has to understand this....
Ideas=science. Simplification=engineering.
Without ideas there is nothing to simplify.

User avatar
lucasart
Posts: 2958
Joined: Mon May 31, 2010 11:29 am
Contact:

Re: Elo measurement of contempt in SF in self-play.

Post by lucasart » Sat Mar 10, 2018 9:20 am

Michel wrote:http://tests.stockfishchess.org/tests/v ... 0297cb64af

Currently (after 180.000 games) it is 3.1 +- 1.0.

So the common wisdom that contempt cannot possibly work in self-play (an opinion that has been voiced by many people, including myself) seems to have been soundly refuted.... At least for SF.

Now one has to understand this....
The test is bogus. See my comment on the diff.
Theory and practice sometimes clash. And when that happens, theory loses. Every single time.

Post Reply