In the context of Monte Carlo, a win or loss by a large margin is quite likely (at least in any close game) to be due to large blunders. (For example, allowing a large, safe group of stones to be captured.) Given this, it does not make sense to weight it more strongly than any other win or loss. It would not even surprise me if a better evaluation could be obtained by _penalizing_ wins by greater margins. (Something like 100-k for a win and 100+k for a loss, exactly contrary to what you propose.)
As a hopefully concrete example, consider the following 5x5 position, with black to play: .XO.. .XOO. .XXOO XXXO. .XO.O Komi is such that black can win by taking and filling the ko. Black _could_ still capture the whole board, by playing on the upper-right corner, but only if white fails to defend its group. One needs to avoid making this move appear overly attractive, because the winning move does not win by such a large margin. Weston On Thu, Oct 9, 2008 at 3:23 AM, "Ingo Althöfer" <[EMAIL PROTECTED]> wrote: > Some of you may want to stone me for this heresy, > but read carfully before. > > When you have MCST-/UCT for Go that can work for > real-valued scores (or at least a version that can > work for three-valued scores: winm, draw, loss), > you may look at Go with different scoring systems. > > Example: > A win by 0.5+k points is worth 100+k. > A loss by 0.5+k points gives score -100-k. > Let's call b=100 the base score. > Question: How does the playing style change with b? > (Of course, for very large B you should have almost the > normal playing style.) > > Ingo. > > PS: Such evaluations may also help to reduce the problem > of MC's laziness. > -- > Der GMX SmartSurfer hilft bis zu 70% Ihrer Onlinekosten zu sparen! > Ideal für Modem und ISDN: http://www.gmx.net/de/go/smartsurfer > _______________________________________________ > computer-go mailing list > [email protected] > http://www.computer-go.org/mailman/listinfo/computer-go/ > _______________________________________________ computer-go mailing list [email protected] http://www.computer-go.org/mailman/listinfo/computer-go/
