On the compgo list people have been saying there is basically no difference between level 1 and level 10. (On 9x9 at least.)
This seems strange to me, as the difference is how much tactical search is being done, and it looks quite significant. Does gnugo show a distinct difference between levels in self-play? Has anyone done any studies? I was just wondering if perhaps there might even be a bug stopping gnugo making proper use of the extra knowledge it is working so hard to gather? Or, can anyone explain in high-level terms, or with a board example, why the extra tactical work does not help much? Are there certain situations where it helps, and others where it makes no difference, and those useful situations don't come up often enough? Darren -- Darren Cook, Software Researcher/Developer http://dcook.org/gobet/ (Shodan Go Bet - who will win?) http://dcook.org/work/ (About me and my work) http://dcook.org/blogs.html (My blogs and articles) _______________________________________________ gnugo-devel mailing list gnugo-devel@gnu.org http://lists.gnu.org/mailman/listinfo/gnugo-devel