Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Go programming

Author: Janosch Zwerensky

Date: 12:33:43 02/13/04

Go up one level in this thread



>What is the meaning of not much weaker?

The author of Go++ recently published some data on rec.games.go suggesting that
his program plays at around 7 kyu european. GnuGo, on the other hand, is 9 kyu
on the kiseido go server(kgs), which would probably put it closer to 7 than to 8
kyu on the european rating scale. So the difference in strength should be one
handicap stone or less.

>What is the result in a match between GnuGo and Go++?

If the difference is one handicap stone, GnuGo should win around 30 percent of
its games against Go++ .

>(...)
>What time control and what hardware do you use in your games?

I usually have GnuGo play at level ten, which is the default, because higher
levels take too long to respond for my tastes. However, I have also played
versions of GnuGo that play on the kgs and on faster hardware, and I don't feel
the difference in strength is substantial.

>Do humans improve with more time per move relative to programs or after some
>time more time does not change much for most humans?

As far as I know, hard data on this one is lacking except for tsume-go (go
problem) solvers, which are said to gain less from more thinking time than
humans. It is my understanding that giving the top programs of today more time
to think will mostly help them to resolve local tactical fights more precisely
and not mend their general strategical deficits, so I would be very surprised if
simply giving them a lot more time led to an increase in playing strength
relative to humans who also get more time to think.

Regards,
Janosch



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.