As I mentioned before, I'm running Lazarus at 7x7 8.5 komi. I now have 139 games in and white is only winning by a small margin. The game are being played at about 4 minutes per game (2 minutes per player.)
The current result is 52.5 percent for white. Of course if the programs were playing perfectly it should be zero percent for white if 9 is the correct komi. - Don On Thu, 2006-10-12 at 17:26 +0200, Magnus Persson wrote: > Quoting Don Dailey <[EMAIL PROTECTED]>: > > How clear is it that the correct Komi is 9 and not another value? Has > > it been proved, or is just strongly believed? And how strongly? > > It is very clear in my opinion. This citation comes from sgf file that > has been > circulating for some time > > " > In the American GO Journal (vol. 29, number 3, summer '95), James Davies had > this to say: "On a 7x7 board, Go becomes challenging for even the strongest > human players. The challenge (find the best moves for both sides) was taken up > in 1977 by a group of Japanese amateurs including Kiga Yasuo, Nebashi > Teruichi, > Noro Natsuo and Yamashita Isao. In 1989, with some professional help from Kudo > Norio and Nakayama Noriyuki, they reached the conclusion that Black wins by > nine points. What follows is a tip-of-the-iceberg summary of the evidence. It > should be added that the analysis is still somewhat tentative, and future > discoveries may lead to further revisions." > " > > My guess is that these amateurs very strong, because you need to be. > The reason > I strongly believe in it is that is very consistent with my experience of 7x7. > Everytime a short game is played with good logical moves from both side the > result just happens to support 9.5 komi most of the time. Note that the > japanese team spent 12 years on this analysis indicating that it was far from > easy. Further revisions I interprete as possible changes to the list of moves > that are correct and incorrect and not the komi itself. > > So my conclusion is > a)7x7 is far from trivial > b) The analysis done to find the solution was some really hard work and needed > professional strength for the finishing touch. > c) Because we now can apply this knowledge in an opening library strong > programs > can get even stronger, but this fact might hide how much knowledge is hidden > behind this simple opening library. > > Still 7x7 is very useful for me. I simply turn the opening library off. It is > much quicker and easier to get an overall impression of new versions of > Valkyria playing 7x7 rather than 9x9. It also more fun because with some > thinking time Valkyria can go very deep on 7x7. Thus one can get a feeling for > the dynamics of the program when it searches very deep. > > So having the solutions is a mixed blessing: it is ideal as a benchmark > to test > our programs with but on the other hand it makes it harder to make meaningful > tournaments for 7x7 although one might start with 4 random ply having both > programs play both colors or something like that. > > -Magnus > _______________________________________________ > computer-go mailing list > [email protected] > http://www.computer-go.org/mailman/listinfo/computer-go/ _______________________________________________ computer-go mailing list [email protected] http://www.computer-go.org/mailman/listinfo/computer-go/
