Then is this the condition? If a move X by your color is immediately after a Ko capture by the opponent and immediately before a Ko capture by your color, then do not count X in the RAVE update.
_____ From: [email protected] [mailto:[email protected]] On Behalf Of Aja Sent: Saturday, August 13, 2011 10:23 PM To: [email protected] Subject: Re: [Computer-go] Aja's PhD thesis: Question about RAVE+ Ko Threat Honestly, I have no good method to detect a ko threat now. I guess that's why I got only 30 Elo in my experiment. Anyway, all we know for sure is that a ko threat must exist between two ko captures (the latter one might be resolving a ko). The principle is that these ko threats must not be updated in the nodes before opponent's ko capture. Aja From: Brian Sheppard <mailto:[email protected]> Sent: Saturday, August 13, 2011 5:29 PM To: [email protected] Subject: [Computer-go] Aja's PhD thesis: Question about RAVE+ Ko Threat In your heuristic for adjusting RAVE for Ko threats, how is the "ko threat" identified? The following conditions seem required: 1) You cannot capture the Ko on this turn because the opponent captured last turn. 2) You make a move somewhere. But are these the only conditions? For example, - no requirement that a Ko exist in the current node (that is, the node in which RAVE is being updated) - no requirement that the "ko threat" actually threaten anything - no requirement that the "ko threat" occurs within the tree (rather than the playout) Thanks, Brian _____ _______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
_______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
