Why should the worst case be the most interesting?
In a program of this complexity worst case isn't the "true" strength of the
program.
Worst case is basically a bug.
What's wrong with looking average play to judge the program?
And in terms of "interesting" I must say that I find the programs best play
much more interesting than it's worst play.
With best play I don't mean some book play ofcourse, but a fine solution to
a tricky problem.
Granted, truly awesome play is currently mostly to be seen on 9*9.
But I've seen some great kills on the big board that any top amateur could
be proud of.
Stefan
----- Original Message -----
From: "Raymond Wold" <[email protected]>
To: <[email protected]>
Sent: Thursday, July 22, 2010 9:59 PM
Subject: Re: [Computer-go] Shodan Go Bet
On 21.07.2010 15:14, Darren Cook wrote:
One option I'm considering is a facade that is running two or more
programs beneath the surface.
This may result in a weaker program than one of the programs playing
alone
but consistently.
The setup is meant differently:
In each game only one program is used. But, before the
game a coin is flipped to decide which program from
a pool.
This is just a technique to stop a human exploiting a known weakness of
a program? (Does anyone knows a specific weakness of any particular
strong program? My feeling is all the MCTS programs have similar
strengths/weaknesses, so knowing you are playing Fuego or playing Zen
won't really help you much.)
Personally, I would not put much consideration into a result that occured
without he human player getting to play a lot of games against the program
(the same version he will play the final matches against) in advance,
where he can find its weaknesses.
I don't think a program that plays as 1dan only when the opponent don't
excessively fight, or only when there's not a lot of life-and-death
issues, or only when the opponent don't go for ippoji, and so on, really
qualifies as 1dan, even though it may get such a rank quite legitimately
on a go server where it plays almost all human at most a handful of games
each. You might not think this "fair" from a human point of view, as
humans don't play one another repeatedly to find flaws either, but I put a
different standard on programs that do not learn on their own. It's the
worst case that is interesting, not the best or the average.
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
