Don't we know that truncated rollouts for cube decisions are
susceptible to considerable evaluation systematic uncertainty?
(I never trust a truncated rollout on a cube decision, but I'm
pretty hardcore. :) By making that one of the default options
does it build a false sense of security for the user who may
not know this?
Chuck
On Wed, 25 Jul 2007, Achim Muller wrote:
Date: Tue, 24 Jul 2007 23:22:07 +0200
From: Achim Mueller <[EMAIL PROTECTED]>
Subject: Re: [Bug-gnubg] Rollout presets
* Massimiliano Maini <[EMAIL PROTECTED]> [070724 16:30]:
I saved
quick:
set rollout truncation plies 12
set rollout trials 324
set rollout jsd minimumgames 216
set rollout jsd limit 1.9600
set rollout player 0 chequerplay plies 0
set rollout player 0 cubedecision plies 0
quick-cube-2ply:
set rollout truncation plies 12
set rollout trials 648
set rollout jsd minimumgames 216
set rollout jsd limit 1.9600
set rollout player 0 chequerplay plies 0
set rollout player 1 cubedecision plies 2
quick-2ply:
set rollout truncation plies 12
set rollout trials 324
set rollout jsd minimumgames 108
set rollout jsd limit 1.9600
set rollout player 0 chequerplay plies 2
set rollout player 0 cubedecision plies 2
cube-2ply:
same as quick-cube-2ply but without jsd and truncation
full-2ply:
set rollout truncation enable off
set rollout trials 324 (or more)
set rollout jsd move off
set rollout player 0 chequerplay plies 2
set rollout player 0 cubedecision plies 2
Pruning for all 2ply (default)
Ciao
Achim
_______________________________________________
Bug-gnubg mailing list
Bug-gnubg@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-gnubg
