Re: [computer-go] SlugGo v.s. Many Faces update

[Mark Boon <tesuji@xxxxxxxxx>]

Compgo123@xxxxxxxxxxxxxxxxx wrote:

> In a message dated 3/31/05 4:25:24 PM Pacific Standard Time, 
> ddoshay@xxxxxxxxxxxxxxxxx writes:
>
>
> > For the branching factors I have been running on our new mini cluster
> > for the last many days, of 63 games with SlugGo playing white with a
> > 5.5 point komi, SlugGo won 42 and Many Faces won 21. 
>
>
> This results seems indicate that SlugGo is one stone or so stronger. 
> But if GnuGo was one or two stone weaker than MF, it's still a 3 
> stones improvement.
For dan-players a 2:1 win-loss ratio indicates about one stone 
difference in playing level. I'm not sure to what extent that holds up 
in the higher kyu regions, but if anything it indicates a slightly 
bigger difference in strength. Computers are not human however, so 
different effects come into play. The only real test is to change the 
handicap. I think 60+ games is already a considerable number. Depends of 
course on what you're trying to measure. Do you want to know the 
win-loss ratio in an even game? Or do you want to know how many stones 
stronger it plays? In the latter case I'd suggest you play with changing 
handicap. Change the handicap as soon as one side is up four games or 
so. You'll see it converge to a certain handicap in 60 games with no 
problem. My guess is after a while you're going to be at most one or two 
stones off of the real level at any particular time, and you'll get a 
pretty good idea of the level simply by the handicap at which the most 
games are played.

Another question though: it's asserted above that MF is one or two 
stones stronger than GnuGo. Is this backed up by any data?

   Mark Boon
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/