Re: [gnugo-devel] small plays at the end of the game

[David G Doshay]

On 2, Dec 2004, at 6:01 PM, Gunnar Farnebäck wrote:

> I still don't understand where you're observing the move value.

From GNU Go 3.4:

extern struct move_data move[BOARDMAX]

move[i].value

> Isn't the code above enough to set the value to 1.0?

I have to admit that I spend very little time inside GNU Go code.
I spend my time in the SlugGo code. We have a few hooks into
GNU Go to maintain when you folks have a new version, and
that is all I ever see. I was trying to learn more by understanding
this problem.

I changed SlugGo back to allow the mistake, and tried multiple
times to generate the problem. The situation came up, but in all
cases the correct move was played (GNU Go always plays the
correct move). However, both the correct move and the incorrect
move did have .value = 0.0. This is what allows SlugGo to make
its mistake.

Should I be using .final_value instead?

In this .sgf file

myTwoGtp001.sgf
Description: Binary data

White move 26 is correct at A2 but has move[i].value = 0.0.
In the comment field it is labeled gg. Not in the file is that B1
also has value 0.0 and SlugGo would sometimes choose
that move. I have a set of games running to see if SlugGo
will generate that move again.


Thanks,
David