In two recent versions of gnubg, ChangeLog 1.175 (16 July 2003), and ChangeLog
1.165 (14 July 2003) I am getting bogus statistics from
cube decision rollouts. I tried rolling out the cube decision after 21:
13-11, 6-5 22:
24-20*,6-4,6-4 66: and
noticed that the last three columns of "win statistics" and last two
columns of "cube statistics", for both possible cube locations,
display what appears to be non-zeroed garbage frequently. With the 2 July 2003 archive
version I get no problem. The most recent cube decision rollouts do not
seem correct, either.
Robert Eberlein