[Computer-go] Bayeisan/Probablistic Playouts in Computer Go

Dave Dyer ddyer at real-me.net
Thu Sep 25 15:39:55 PDT 2014


In my non-go MCTS games, I usually score playouts on a continuous 
-1:1 scale rather than as +1 or -1.   I use the same arithmetic to
update UCT values, and it seems to work at least as well as strict
win/loss scoring.   

The motivation for this is to allow the playouts to be stoped at any
chosen point, not just at the end of game; and for actually ended
games, to allow the score to give feedback about "better" wins.




More information about the Computer-go mailing list