On Sun, 2007-04-08 at 00:44 -0400, [EMAIL PROTECTED] wrote: > I question whether it's valid to make this kind of comparison when > Gnugo scales so differently from UCT. If you froze one of the UCT > prgrams at 1 million playouts/move and then tried to scale gnugo until > it matched that level of strength, that might not even be possible. I > think you need to compare 2 curves (as you're doing with the lite and > heavy UCT programs), not one curve and one fixed point.
I'm not really comparing the scalability of GnuGo but I'm treating it as a fixed point. I'm also trying to "abstract" the concept of strength/time as a unified concept whether a program scales or not. In other words, the strength of a program is almost meaningless without a consideration of time. When people compare 2 programs and say this one is stronger than that one, there is this unspoken assuption that they are running on similar hardware with similar time being used. (And this is how it should be, but they are almost oblivious of the time factor.) So I think it's more correct to say Gnugo is stronger at 9x9 GO even though Lazarus beats it on CGOS. No one would question this if I were running Lazarus on a 486, they would just say GnuGo is stronger. Of course I agree that GnuGo has different scalability properties so it's a slippery concept. I don't know if GnuGo would win if I cranked up the level on it. But at least at fast time controls, GnuGo has a more efficient underlying engine and "in principle" it might scale better. I'm not saying we should throw out our concept of ELO strength - but I am saying that it's not 1 dimensional. In other words, you have to give 2 parameters e.g. "Gnugo is stronger than Lazarus playing at 1 minute per move on a pentium 2.4", etc. If a program is known to be "fixed" without the ability to adjust parameters or levels, it is sufficient to speak as if it's an absolute thing without any regard to time, because the time factor is implicit and understood. Therefore, you will occasionally hear statements like, "program X is pretty strong, but it takes quite a bit of time for it to play a move." - Don _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
