Computer Chess Club Archives

Search

Messages

Subject: Re: ICGA_J (June) self-play information

Author: guy haworth

Date: 11:31:38 09/05/01

Go up one level in this thread

Ernst Heinz used Fritz (with help from ChessBase to set up the experiment
repeatably and carefully).

Clearly, advances in search strategies are reducing the effective fan-out:  pick
your favourite number in the range 3-6.  Certainly, 3 gives 3^3 = 27 < 35 and an
improvement of 3 plies.

'All other things being equal', EH's results suggest that you are not going to
see more than 240 ELO improvement by upping guaranteed-search-depth from say 14
to 17 plies.

Results for other chess engines will be less predictable.

Interesting comparison between SSDF results and EH's.  EH ran 3,000 games per
match which is by a long way, more than others have done.

G

• Re: ICGA_J (June) self-play information Uri Blass 11:50:00 09/05/01
  • EHeinz self-play experiment guy haworth 15:58:09 09/05/01
    • Re: EHeinz self-play experiment Uri Blass 22:42:43 09/05/01

This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.