Zen was indeed heading for a crushing win.
The first funky move by zen was 157 and after that it didn't get a thing
right.
Many truly separate battles are still the bane of mc bots.
I guess that's the natural price to pay for always considering the whole
board.
What is frightening about this example is that even "easy" semeais seem to
contribute substantially to the overall confusion. I have seen Zen perform
wonderfully in playing against 2 weak enemy groups at once.
But in that case it was really one big battle.
It's counterintuitive, but the problems start when fights are "obviously"
separated.
Imo, Zen 5.9 has reached 2 dan level. That's very impressive and is a
warning against dire prognostications of future stagnation. My guess is that
a lot of effort has already gone into avoiding gratuitous context switching
in the mc search. Maybe in the future old concepts like "groups" and
"connections" will sneak in through the back door of statistical mc search
evaluations.
Stefan
----- Original Message -----
From: "terry mcintyre" <[email protected]>
To: <[email protected]>
Sent: Monday, August 30, 2010 4:32 AM
Subject: Re: [Computer-go] Last night: Zen5.8 vs John Tromp
I just watched this Zen19N game:
http://files.gokgs.com/games/2010/8/30/Zen19N-atheist.sgf
If I am reading correctly, Zen could have killed the bottom right, but
played
its own liberties before taking away its opponent's liberties.
There is also a tail which zen lost at the top, due to negligence.
The bottom left group, I think, could have been saved.
This looks like a classic case of snatching defeat from the jaws of
certain
victory.
Perhaps this game can provide some useful test cases?
Terry McIntyre <[email protected]>
Unix/Linux Systems Administration
Taking time to do it right saves having to do it twice.
----- Original Message ----
From: Yamato <[email protected]>
To: [email protected]
Sent: Thu, August 26, 2010 9:33:16 AM
Subject: Re: [Computer-go] Last night: Zen5.8 vs John Tromp
I attach the image of Zen's evaluation in the game.
Look at the white groups at the top and top-left. Both of them are not
100% correctly evaluated. When this happens, Zen gets weaker because of
the divergence and inaccuracy of Monte-Carlo simulations.
I think it must be the current major problem of all MC programs.
Christian Nentwich wrote:
>From a quick look, I would say the bots' evaluation in this case is
entirely
>correct. Zen was quite far ahead, and then made some bad endgame
>mistakes.
>The trade around move 200 is particularly costly. Playing even a simple
>safe
>move at 176 (like B10) would win the game without trouble.
>
>Christian
>
>
>On 26 August 2010 10:01, Darren Cook <[email protected]> wrote:
>
>> > Tonight an interesting game was played in the
>> > cellars of KGS: Zen19N vs Tromp.
>> > (Zen is a bot, Tromp is John Tromp...)
>> >
>> > You can download the sgf from the KGS archives at
>> > http://www.gokgs.com/gameArchives.jsp?user=tromp&year=2010&month=8
>> >
>> > For a long time bots (Zen itself, Many Faces) believed
>> > that White (=Zen) was clearly ahead. But around move 200
>> > the evaluations swung over, until Tromp won after move 271.
>>
>> Thanks Ingo. What are people's thoughts: were the programs mistaken
>> in
>> their optimism, or did Zen make a mistake in the middle game?
>>
>> Darren
>>
>>
>> --
>> Darren Cook, Software Researcher/Developer
>>
>> http://dcook.org/gobet/ (Shodan Go Bet - who will win?)
>> http://dcook.org/work/ (About me and my work)
>> http://dcook.org/blogs.html (My blogs and articles)
>> _______________________________________________
>> Computer-go mailing list
>> [email protected]
>> http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
>>
>---- inline file
>_______________________________________________
>Computer-go mailing list
>[email protected]
>http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
--
Yamato
_______________________________________________
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
