Re: [Bug-gnubg] Rotating rolls in rollout

[Morten Wang]

pardon me if this first post contains any errors or breaks any general
guidelines for bug-gnubg posts.  I googled for guidelines but couldn't
find any so I'm relying on my experience.

* Nis  Jorgensen
> (how many people actually do rollouts with 36^3 trials?)

actually, I did one a few days ago[1] and I'm expecting to do more if
there's actually anything to gain from doing them.  I'm not a Snowie
owner so coming from JellyFish 3 my preference is 1296 trial rollouts
with variance reduction for checker plays.  not having made the full
transition to gnubg yet I am not well travelled in its rollout options
but I'd probably be doing the same kind of rollouts there (unless
documentation suggests otherwise, I haven't actually read the FAQ yet).

if running a 36^3 trial rollout (with or without variance reduction)
provides me with more trustworthy results compared to the 1296 trial,
while not taking an eternity to complete I'll probably be doing more
of them.  my newly acquired server (AMD XP2000+) will be the testbed
for it, once I get gnubg running on it.


Morten!

1: I rolled out the first 7 moves listed by gnubg for the position in
   http://groups.google.com/groups?selm=m3k7l2qiwq.fsf%40tigger.plaindog.no
   I don't have the settings at hand but IIRC it was a
   non-truncated 36^3 trial rollout at expert level, no variance
   reduction, no rotation, cubeless.

-- 
"...a liquid which was almost, but not quite, entirely unlike tea."
  My opinions are not necessarily those of my provider,
  not necessarily mine, and probably not necessary.