>> ... play those 4 moves as if a single move.... > > I'd just like to point out that you shouldn't do this if there is a ko > going on. ;-)
Yes, good point. It is possible to detect if you need a ko threat (the illegal ko move is the move you want to play most), and in that case I think I was widening the choice of moves. It is also possible to detect if a move was played as a ko threat when analyzing the pro games, and flag it as "possibly only a ko threat" (*), but I think I ran out of time long before I got that sophisticated. Darren *: I.e. it could then be put in a special database of "ko threat pattern moves". But, what I actually tried to do was assign the gote and sente value of each pattern. A ko threat would look like, say, a 15pt gote, but a 0pt sente move. But measuring exact move values is hard enough in the endgame (see mathematical endgames), let alone the middle game, and so this didn't work well. Why I think this pattern idea might have merit in MCTS is that all you need is a reasonable ordering of pattern urgency. The move selection then adds some noise to cover up the inaccuracies. -- Darren Cook, Software Researcher/Developer http://dcook.org/gobet/ (Shodan Go Bet - who will win?) http://dcook.org/mlsn/ (Multilingual open source semantic network) http://dcook.org/work/ (About me and my work) http://dcook.org/blogs.html (My blogs and articles) _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
