I think we still have a long way to go on 9x9. 7x7 is played almost perfectly by computers but not 9x9. To compare, checkers is a game now where computers rarely make mistakes and in fact they have practically solved the game. But 7x7 is probably even farther away this kind of master than computer checkers.
9x9 is already too profound for computers but computers do play it quite well. Personally, I would like to see 11x11 become popular as the next step, but it isn't going to happen - it's just not popular with people. 13x13 is a great stepping stone on the way to bigger boards, but it will keep computers challenged for many years to come. It's certainly a matter of preference and taste, but I would like to see one board size "mastered" thoroughly before proceeded to the next that is 2 larger, i.e. 9x9, 11x11, 13x13, 15x15 etc. By "mastered" I don't mean anything close to perfect play or even as good as the best player in the world, but just something that approaches the top players in strength - perhaps the top 1000 in the world. Don On Wed, Jul 14, 2010 at 9:39 AM, <[email protected]> wrote: > Quoting Nick Wedd <[email protected]>: > > I see that many people now find 13x13 Go more interesting than 9x9. >> The next two KGS bot tournaments, on August 15th and on September 5th, >> are both scheduled to be 9x9. How would people feel about my changing >> one or both to be 13x13 instead? >> > > Well there was some 13x13 activity on this list recently. For a while some > program showed up on CGOS 13x13 so interest increased from zero to more than > zero. But I would be careful how to interpret that. > > Meanwhile people are testing new versions of their programs on CGOS 9x9. So > activity wise it seems 9x9 is most popular, then 19x19 and less so 13x13. > But looking at CGOS can be biased because people may only test on 9x9 just > because they get a lot of games quickly and a greater variance of opponents. > > Still right now only two programs are active on CGOS13x13 in addition to > the GnuGo anchor. > > I would like to run at least one tournament on 9x9 because that is what I > have been personally working on lately. > > Also I tend to think that developing go programs for 13x13 is similar to > 19x19. What is interesting with 13x13 is that one should perhaps be able to > make stronger programs on 13x13 before 19x19. So it is a natural stepping > part in test against humans. Still I think working on 19x19 will naturally > generalize to 13x13 as well. > > The difference between 9x9 and (13x13/19x19) is that the former has whole > board patterns/opening whereas the latter one needs to focus on local > fighting and evaluating sequences of moves strategically. This then have > consequences for how large scale patterns should be used in the programs. > > It is also premature to stop working on 9x9. The best programs are strong > indeed but most are very buggy in the sense they play very weak in certain > situations. Best opening play is just mere guessing and mimicry. I would > define 9x9 as sort of mature when top programs wins with one color and loses > only with the other. I do not see that happening yet. > > Anyway I know there are very strong programs for 9x9 that has not played > for a while on CGOS since the all time rating list was reset some months > ago. So it now looks like Valkyria is the strongest. Who dares to disprove > that?! ;-) > > Best > Magnus > > _______________________________________________ > 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
