[computer-go] thoughts on 100,000 cgos games

[David G Doshay]

Sorry for the lack of clarity on my part. I know that I can ask for  
all of
the sgf files for SlugGo's games, and that the files sent will include
the opponent name. What is more accurate is that I need the info in
the sgf file that we generate that has all of the branching information
that SlugGo generates and considers. I have doubts that I can write
a parser that finds our sgf file (with all of the branching info) that
corresponds to the sgf file sent me with the opponent name. I don't
doubt that it is possible, just my ability to do it. sgf is hard  
enough to
read when it describes a plain vanilla game. When all of SlugGo's
moves have a dozen options, and in particularly when doing multiple
levels of branching in the lookahead, the sgf is just full of nested
brackets and references to comments that wind up all over the file.

What I can do is rerun every one of the games against a strong
opponent and regenerate the SlugGo evaluation to within the
variation in the random number seed. But that sure will take a
long time. I would rather have that info the first time I do the
calculations. Can I ask in the process of setting up the game
for the name and/or rating of the opponent program?

With respect to learning, about the best idea so far is that if I  
note that
some evaluation of the board goes from SlugGo winning to SlugGo
loosing, and SlugGo does eventually loose, then it is likely that the
error was before that. I am not at all sure that this must be true. But
that does narrow it down to "some move near here." In my opinion
at this time, that is not enough to do automated learning.


Cheers,
David



On 7, May 2006, at 5:44 PM, Don Dailey wrote:

> But Davids original complaint was that he couldn't do training because
> he didn't know who the opponent was (I don't understand why he thinks
> this but he does - I can send him the games WITH opponent  
> information in
> the SGF files.)
>