Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Learning and doing research

Author: Omid David Tabibi

Date: 04:23:26 03/20/03

Go up one level in this thread


On March 20, 2003 at 02:52:39, Travers Waker wrote:

>http://www.kozachenko.com/dotbishop/  (used to be Yellow Bishop)
>
>Paper 67: Andreas Junghanns, "Are there practical alternatives to alpha-beta?"
>(ICCA Journal, 21(1): 14-32, March 1997), rev. January 29, 1998, 19p.

This was just the paper I was going to recommend. Beating AlphaBeta is no easy
task, as it is the result of decades of research. There might be alternatives,
but I'm quite skeptical about revolutionary different methods managing to
outperform AlphaBeta. As Aske Plaat once mentioned, today we are not dealing
with the simple AlphaBeta, but we have all the enhancements (NegaScout/PVS,
null-move pruning, transposition tables, history heuristic, etc) which make the
system almost impossible to beat within a short period.

I believe that a more worthwhile research should focus on improvements within
the AlphaBeta framework, mostly null-move pruning, futility pruning, and other
pruning methods which do have major impacts. I recommend you read Heinz' book
"Scalable Search in Computer Chess", or at least his papers dealing with various
pruning methods, available at his website (pdf versions also available at
http://www.cs.biu.ac.il/~davoudo/articles/heinz).

Computer chess research is now mainly focused on the very minute details of
search trees. This might be considered a curse, as just by looking at the recent
issues of ICGA Journal you can see the scarce number of (published) new ideas in
computer chess (the latest issue 25(4) had no paper on computer chess), but I'd
rather consider it a blessing, since as the "most researched field in computer
science" computer chess can allow itself to deal with such profound details
other fields can only dream of.

Omid.


>
>His conclusion is that there aren't practical alternatives to alpha-beta, but it
>might still be worth reading to get an idea of what alternatives have been
>looked at.



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.