I am a twenty years computer professional, a master level Chess player and about a 95 percentile Backgammon player that has recently switched to Go. I realize the games present different problems in computing. Neural nets work great for Backgammon because of the element of chance and quasi-brute force works well for Chess as the effective branching factor is so low. As Go needs both approaches and combinations in Go can frequently be thirty moves or longer I am surprised at how well your program plays!
My question relates to analyzing positions using GG. I have recently been trying to get GG to find the killing move that concluded a game between Takagawa Shukaku 9d and Kato Masao 9d. Shukaku's last move was to try to connect two groups by playing under the stones on the second line, a common method. In case my terminology is incorrect, I would describe the position as three stones in the shape of a hanging connection, two stones on the third line and one on the second, surrounding a protruding stone of an opponent's live group on three sides. For example, White stones at C3, D2, E3 and Black stones at D3 and D4. When this connection is attempted the standard method of breaking it is to sacrifice a stone in one of the cutting points so that when you place a stone in the other cutting point it is atari. I have been making test runs with GG failing to find the move every time. In each successive run I increment all of the parameters by one. I do have a concern that I am missing a parameter as the documentation lists branch-depth twice. I am running on a 2.2G AMD 64FX box and the last run took about eight hours. While GG produces different results at different levels, the killing move is never in the list produced by the all_move_values query! This move is obviously the best move as it is about a sixty point move.
Now that all of the preliminary information is out of the way, here is the question. Does increasing the computational settings cause GG to look at more and or different initial moves? If not, why isn't a basic tesuji play in the list? I think it takes 33 moves to capture the group although it is obviously dead after five moves. I just started a last test run that I expect to take three days with settings that I think should enable GG to compute the death of the group to the end. But that is only a guess based upon the documentation and my feeble Go skills.
Thank You in advance for you time!
I also hope this helps with your development effort.
CTCampbell
PS
I remembered to count moves as in Go rather than Chess ;-) Move = ply.