-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Jason House wrote: > I just looked at 133684. ego110_allfirst took all the way until move 18 > to play a move above the 2nd line. In fact, in the 8 moves it made > prior to that, 5 were on the first line. ErlyGo's random plays had only > 2 on the 2st line and 2 on the 2nd line. It seems like a random > strategy gives a better opening strategy than what ego110_allfirst ends > up with. > > I've been wondering what was done to ego110 to make ego110_allfirst, but > have been unable to find out yet. There's been lots of discussion > lately on what would make a good "all moves as first" bot with no final > conclusions. I know that my AMAF variant shares more in common with ego > than Don's, and Don's version. My bot is better than ego and Don's is > better than mine. I think it might be possible to create a non-searching very simple 9x9 bot based on all-as-first that is over 1600. > I've been toying around with time control lately (and failing to get it > right). The main principle in time control is that early moves are far more important than later moves. When 2 equal fairly weak bots play, the game is usually over (for all practical purposes) in the first very few moves. Often in the first 10 moves for example. So you might as well spend a lot more time on those moves. A very simple and yet effective strategy is to simply allocate a fraction of the remaining time (as reported by the GTP time_left command) to the next move. The fraction you use should be based on the board size. This will cause early moves to be allocated much more effort than later moves and ALSO has the desirable quality of speeding up the end-games. Arrange it so that the first move (that is not a book move) is allocated considerably more time than the 30th or 40th move. I put a lower bound on how low the allocated time can dip - because when it get's to the point where it is playing really fast, the game is very likely to be within a few moves of completion. But the lower bound is still pretty low - it's less than the 1/4 second CGOS gift and it's only purpose is to prevent a nearly random move. You can try to get more sophisticated, but you will probably be doing a lot of work for a gain you can't measure. I would be surprised if you could find a better time control that could be demonstrated to be better and it certainly won't be simpler. KISS is almost always best - if not KISS you better have a really good reason. - - Don I can take down housebot-xxx-ucb and put up housebot-xxx-amaf > if that helps with testing of erlygo. (housebot-xxx-shuff is needed for > my timing test since it seems to stress the time usage policy more than > my other variants) > > On 9/24/07, *Don Dailey* <[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>> wrote: > > Did you look at the games? > > I looked at game 133580.sgf and your program was clearly lost playing > the black pieces, however the game record says that your opponent > resigned in a dead won position. > > In fact, there were not moves left to play other than filling eyes. > > > - Don > > > > Urban Hafner wrote: >> Hej all, > >> I'm wondering if anybody knows how ego110_allfirst on CGOS works. The >> name seems to suggest that it does random playouts and uses the >> all-moves-as-first heuristic. Is this correct? And if so, how many >> playouts does it do per move? > >> I'm asking because I started developing my own bot. Right now the > only >> thing it does is play legal moves that aren't eyes of the bots color >> (according to libegos definition, as that's what the bot is/will be >> based on). The strange thing is that in the 9 games against >> ego110_allfirst it was able to win 5 [1]. This seems really strange to >> me. I would have guessed that even a small number of playouts would be >> enough to beat a random bot like this consistently. > >> Urban > >> [1] http://cgos.boardspace.net/9x9/cross/ErlyGo-0.0.1.html > > > > > > > > > > _______________________________________________ > computer-go mailing list > [email protected] <mailto:[email protected]> > http://www.computer-go.org/mailman/listinfo/computer-go/ > _______________________________________________ computer-go mailing list [email protected] <mailto:[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/ -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.6 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFG984IDsOllbwnSikRAuHRAKCPxXUlywNefTllCuXMpbGG76GYdACggDax Z8NSPcDD39LFAIQFKwPW7YI= =Mmjb -----END PGP SIGNATURE----- _______________________________________________ computer-go mailing list [email protected] http://www.computer-go.org/mailman/listinfo/computer-go/
