From: utzoo!decvax!ittvax!wexel Newsgroups: net.games Title: Re: Programming Go - (nf) Article-I.D.: ittvax.590 Posted: Wed Jan 26 09:01:18 1983 Received: Fri Jan 28 06:44:00 1983 References: sri-unix.5061 There's an interesting story about that "guy at Stanford" who did a Go program. First of all, he is Bob(?) Ryder, the son of Robert Ryder, ex-BTL, the first Yank to get a Japanese Pro Go rating. Anyway, the Ryder, Jr. program was being debugged one night at Stanford when Don Knuth came along and offered to try the program out. Since Knuth had but a rudimentary knowledge of Go, the program, lousy as it was, creamed him, which allowed Ryder to announce that his program was good enough to beat Knuth. This impressed many people. This story was told to me as true. I can't vouch for its veracity. When I called the program lousy, that reflects the dire complexity of Go strategy, not the competence of Ryder. The problem with Go is the difficulty of recognizing strategic situations. For those familiar with the terms, joseki is trivial, fuseki is near impossible. The book runs out early and you are left with trying to recognize strategic significance of stones very distant on the board and very distant in time sequence of play. The most trivial example of this is in playing ladders. More advanced strategy adds orders of complexity. The first dissertation on Go playing, by the way, was that of Dave Lefkowitz at the Moore School, U.of P. in 1963(+-1). I think the title was "Strategic Pattern Recognition in the Game of Go."