You don't have to see to the end of the game to play well. They have done studies in computer chess which show the number of times a deeper search changes it's mind. It becomes smaller and smaller with depth - presumably because most of the moves are already optimial.
Also, it is clear (in chess) that computers play the VERY BEST move most of the time, as good humans do. At the upper levels of skill, the difference between a good and great player is the occasional move. In my UCT study, you will notice that the highest level is signficantly stronger than the levels below it. However, MOST of the time, it plays the same move it would have selected after just 1000 play-outs. In other words, most of it's superiority is centered around a few key moves. The 2 million node version RARELY plays a different move than the 1 million node version, but it is perhaps just a couple of time per game on average. And yet that is enough to demonstrate signficant superiority. In most chess games between grandmasters, you can often identify ONE move that made the difference between winning and losing. - Don On Sun, 2007-04-08 at 11:46 -0400, [EMAIL PROTECTED] wrote: > The factorial of 81 is about 10^140. The number of legal positions may > be some roots of that. It's still a huge number. The number of > simulations used in UCT grograms is several hundred thousand, which is > tiny compared to the total possible number of plays. How can they > still be so successful? I think the only explanation is that the > evaluation values (at least for MC) is highly structured. The question > needed to answer is that is the scale of this structure proportional > to nxn? (n is the board length.) If it is, it's very hopeful for > computer Go in 19x19 using UCT-MC approach. > > > Daniel Liu > > ______________________________________________________________________ > AOL now offers free email to everyone. Find out more about what's free > from AOL at AOL.com. > > _______________________________________________ > computer-go mailing list > computer-go@computer-go.org > http://www.computer-go.org/mailman/listinfo/computer-go/ _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
