I am not sure what you try to show here, but if we can't show that gnubg is individually better than bgb or sn, the likelihood it is better than both is still harder to show
On 7/6/07, Achim Mueller <[EMAIL PROTECTED]> wrote:
* Joseph Heled <[EMAIL PROTECTED]> [070705 14:25]: > Standard error may be informative, but statistical importance matters > more to me. I think it is clear you tell gnubg is better than Jelly > from those number, but nothing more. Well, I wrote a lot of nonsense today (my statistic lessons are 20 years ago), but what about this conclusion: gnubg 1623 over all wins bgblitz 1519 overall wins j.s.d = sqrt(27.3+27.3 + 27.38*27.38) ~ 54.6 (1623 - 1519)/54.6 ~ 1.90 1.90 --> 97.1% confidence that in the overall competition (1 bot against three bots) gnubg is better than bgblitz. That's one way we compare the reliability of move rankings, so it should be right here also, shoudn't it? Ciao Achim _______________________________________________ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
_______________________________________________ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg