5x1000 games, no opening book, Fuego svn version 1467.
Number of simulations varied from 3K (3000) per move to 300K.
I was interested in whether the recent fix to allow Fuego to play for draws
with integer komi worked, and whether the number of draws increases with longer
searches.
Here are the results:
3K simulations: 9.5% draws, 52.4% wins by Black
10K simulations: 8.8% draws, 47.7% wins by Black
30K simulations: 11.2% draws, 49% wins by Black
100K simulations: 10.6% draws, 45.9% wins by Black
300K simulations: 13.8% draws, 46.1% wins by Black
So there seems to be a slight increase in draws. It is interesting that Black
has an advantage, e.g. with 300K White wins only 100-13.8-46.1 = 40.1%.
I wonder how to do the similar test with an opening book. We now have a quite
large book for 7.5 komi, and a smaller book for 6.5 komi. Using the 7.5 book
with a komi of 7 for White will make the program equally happy with a
draw-or-win while in the book. Using it with Black will make it try to avoid
draws.
Of course, we could build a new book with 7.0 komi. But it would be nice to be
able to control between "must-win" and "happy-with-draw" situations.
Martin
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
