Yes, it is definitely possible to engineer an improvement that is mostly effective against a specific opponent and may even hurt the play in general.
The kind of in-transitivity I have seen the most is where you make a change that helps a lot in self-testing but only a little or not at all against other opponents. Usually it is a matter of common sense. Different kinds of improvement may call for more skepticism about the results. A common effect when doing global searching is that a small improvement in speed helps a lot more in self-testing than against other opponents - but again, it's not a BAD thing that you sped up the program. Somehow it seems that the extra speed is more fatal when you know exactly what your opponent is thinking! In the experiments I'm doing, I'm not specifically tuning the evaluation to beat itself and apparently I'm not seeing a problem with transitivity. - Don On Wed, 2006-11-29 at 15:15 +0100, alain Baeckeroot wrote: > Le mercredi 29 novembre 2006 14:21, Don Dailey a écrit : > > I have heard this many times - but it doesn't always apply. In fact I > > have heard that IMPROVEMENTS always look better against your > > twin-brother but if that were true, I would always want to test against > > my twin since it makes improvements stand out. It's hard to measure > > small improvements so this would be like using a microscope to help me. > > > > But unfortunately a change can help you beat your twin but make the > > program worse against other opponents - but I have only occasionally > > seen this be a big factor although I admit it does happen. I just > > think the effect is exaggerated by people. A general rule is that if > > you are better against Joe, you are "probably" better against Fred. > > Agreed. > Some version ago a "cosmic-gnugo" option was introduced (playing huge > moyo following Takemiyia Masaki 9p cosmic style), and it was > significantly stronger against standard GNU Go, but weaker against other > opponents. > > As it is not an improvement (lose against other), this "microscope effect" > at least showned a weakness in the standard version (not abble > to counter a cosmic strategy). This is usefull. > > Alain > _______________________________________________ > 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/
