Go Online

February 23, 2004
By Roy Laird, AGA Webmaster

When most people study go, they look at professional games, life-and-death problems and other technical commentary. But to an enthusiastic subset, "Studying Go" also means learning about it as a social, historical, and cultural artifact. Those who have crossed the "digital divide" love the fact that the strongest computer programs routinely lose to small children and mid-level amateurs. When we play go, we do something that cannot be programmed. A number of well-written articles have expanded on this theme, notably Katie Hafner in The New York Times at 
http://query.nytimes.com/gst/fullpage.html?res=9D01EEDC1F38F932A3575BC0A9649C8B63
and 
http://query.nytimes.com/gst/abstract.html?res=F00C1EF93B580C778EDDAB0994D0494D81,
David Mechner in The Sciences at http://www.cns.nyu.edu/~mechner/compgo/sciences/ and Michael Brooks in New Scientist. It was a bit of a struggle to get the Brooks article, but worth 
it. Go to www.newscientist.com and sign up for a free seven-day trail of the archive. Search for "reiss" (Go4++ programmer Michael Reiss is interviewed at length) and you'll find the article.

But for those who crave scholarly discourse about go, the "Mother Lode" is the semi-annual International Conference on Baduk, which publishes its proceedings. The first two conference have published a total of more than forty articles on a wide variety topics. To learn about how to get your copy of the proceedings book, go to http://www.usgo.org/yearbook/2003ICOB.asp. Copies are available to AGA members only while supplies last!

Now that I've shared some interesting links with you, I hope someone has a link I'm looking for. On GoTalk, the lively British forum, someone recently mentioned an article that supposedly demonstrates objectively that go is the most difficult of all games, but couldn't recall where it 
appeared. Apparently the author started from the assumption that if Player A can beat Player B in two out of three games, Player A is one level stronger than Player B. He/she then determined that in chess, there are about twenty-five such "degrees of difficulty," but that go contains nearly forty such "degrees." Does anyone know where to find that article? If so, 
please e-mail it to webmaster@usgo.org. Thanks!

Followup (March 1, 2004)

PROVING GO IS HARDER: Last week Online Go columnist Roy Laird asked about measuring the relative difficulty of go compared to other games. "Here is the math I've used to support this argument myself when comparing the two games,†responds Rick Mott, who has worked closely with AGA ratings Statistician Paul Matthews over the years. “According to 'Inside the AGA Rating System' (download from http://www.usgo.org/resources/downloads/aga rating.pdf), when Player A is one stone stronger than Player B, Player A will win five out of every six even games. Ratings in the AGA database run roughly from 40 to +9, with no ratings between 1 and 1, so the number of stacked "5 out of 6" win probability levels of playing strength in go is about 47. Beginners are more variable, and we have less data below 30 kyu, so the system may overestimate the spread among beginners to some degree. But even if a true beginner starts at 30K instead of 40K, there are still nearly forty degrees of difficulty. 
"To derive a comparable number of levels for chess, look at the US Chess Federation's algorithm at http://math.bu.edu/people/mg/ratings/rs/node5.html . From the formula for WE (winning expectancy), we find that if chess player A beats chess player B five out of six times, their ratings differ by 280 points. Since chess ratings top out near 2800, there are 10 such levels. The USCF compresses the top of the scale to keep the strongest players' ratings from increasing without bound, so perhaps there a few more levels in chess that it might appear at first glance. Still, by one specific measure of complexity, the playing strength spectrum in go appears to be about the cube of that in chess."

Back to Columns