Serendipity - still room for the human

You come across odd reactions - including your own - when you are creating databases.

For example, it is easy to convince yourself that the production of an easily accessible mountain of games is one of the most important developments in go history. We have the example of chess to sustain that view. Yet go pros, and Orientals in general, have shown remarkably little interest in them. That's odd, isn't it?

It doesn't mean pros aren't interested in previous games, of course. But the system they prefer at the Nihon Ki-in is simply to ask the clerical staff to stick photocopies of the latest few games by Yi Ch'ang-ho or whoever in their pigeon holes. Also, when the pros get the games, they rarely play right through them. Indeed, they often don't play them over at all. Just looking at the game record is enough. They usually just want the first few moves, and they are easy enough to play over mentally. They appear to be not the slightest bit interested in whether Black played move X 50 times but move Y 30 times. They want ideas, not moves; theory, not statistics. And they have the infrastructure to be able to pick up or pass on these ideas - teachers, study groups and much more printed material than we have in the west.

Still, playing over a game from a database or an sgf record is more convenient, isn't it? We were struck by how vehemently a senior official at the Nihon Ki-in poured scorn on that idea. The view there, he told us (despite their own 65,000 game database!), is that playing over games yourself from a written record is the way to become strong. Presumably, on that view, you get an even hairier chest if you play over a game from a single diagram. Well, odd as it may sound, that has worked for GoGoD's main database contributor, T. Mark Hall. At any rate he can claim at least two stones improvement just from inputting games, though the extra hairs may have gone grey when doing 380-move monsters with Chinese numerals and smudgy printing and misprints!

But databases have certainly been able to shoot down even pro preconceptions about when certain moves were first played, and, as we have shown ourselves in our "White 8 is bad" seminar (a mini-version will be given at the 2006 US Congress) they can elucidate the development of pro ideas as well as moves.

Probably what it all boils down to is that the value of a database depends mainly on the human element - the quality of input and the quality of the questions you ask of it. We're happy enough with the quality of our input as we have orignal sources and can read them. But we still find the questions tough to frame. What we are presenting here are a few recent examples that came up during inputting and that the GoGoD team thought intriguing enough to discuss over their weekly lunch. They are examples where the interesting moment (we believe) could be highlighted only by a human editor. In other words, someone at some point has to have enough awareness of go history or be strong enough at go to be able to say, "Hey, that's interesting/unusual/new!"

Naturally we pile in all the examples we find into our encyclopaedia which supports our database. But here are those recent novelties which we believe so far are beyond the abilities of computers alone to pick up.

Next >>