There are ways to make go programs play more like a human in this regard, if you are willing to reduce the strength. You have to pick one, maximum strength or some level of compromise to create the illusion of playing stronger. There does not seem to be a way around this.
Here is a very simple analogy that I think explains why the current technique works so well and it's so difficult to improve on: Suppose you were a world class athlete hoping to win the Boston Marathon. That is a pretty ambitious goal, but maybe not so ambitious if you are the worlds best. However, if your goal is not win the marathon, but to finish 5 minutes ahead of the second place runner, you have a serious problem. As soon as you see that you are "behind" the pace to do this, you must start running much harder. And by doing this you almost guarantee you will not even place first. You will not win because the goal you set is actually not very compatible with the goal of winning the race, even though it seems like winning the race is a subset of the goal. In computer go, as it turns out, the goal of winning the game seems to be like the goal of winning the marathon. If the goal is more ambitious, you have to pay a price for that extra risk. It's like a law of nature I suppose. Do you want to win, or look good? If you want to look good, expect to win less. I think one solution that is used is to superimpose a more classical move algorithm over this, so that by default the program plays a reasonable move instead of a random move when it otherwise has no affect on the outcome. Don 2011/7/3 Andrés Domínguez <[email protected]> > 2011/7/3 Álvaro Begué <[email protected]>: > > On Sun, Jul 3, 2011 at 8:36 AM, Leon Matoh <[email protected]> > wrote: > >> [...] > >> One of problems (which I tested with gogui, thankyou very much) > >> was losing points in endgame when program is winning. > > > > Are you talking about positions where the program losses a won game by > > not playing the endgame correctly, or situations where the program > > ends up winning by a smaller margin that it could have? The former is > > certainly a problem, but it is arguable whether the latter is a > > problem or not. > > From a mathematical point of view (thinking a win is perfect score) maybe > is not a problem, but from a player wanting to play an endgame against > a good player of course it is. Nobody wants to play the endgame against > 30k level moves, it's extremely annoying. Thinking a perfect move should > take the maximum of the points is a ploblem anyway. > > Andrés > _______________________________________________ > Computer-go mailing list > [email protected] > http://dvandva.org/cgi-bin/mailman/listinfo/computer-go >
_______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
