Le Lundi 13 Mars 2006 21:22, Gunnar Farneb�ck a écrit : > Dan wrote: > > Perhaps the semeai nodes should be increased but 2000 is > > too much. The current default is 500. > > These are the results with semeai node limit 1000: > 4 PASS > 2 FAIL > > and with 1500: > 8 PASS > 10 FAIL > > With 1000 nodes I measure a regression slowdown by 9% and with 1500 a > slowdown by 17%. > > /Gunnar >
On CGOS 9x9 games i _feel_ that increasing the level gives really more strengh (twinbot at level 12 is ranked 30 ELO lower than at level 14, =4% of win probability). Maybe this is not due to semeai_nodes, but on a 9x9 it seems probable that this is important. I wonder if playing with more semeai nodes would not gives some other testcases where we see more differences than the actual regression suite. Alain _______________________________________________ gnugo-devel mailing list gnugo-devel@gnu.org http://lists.gnu.org/mailman/listinfo/gnugo-devel
