Pros and computers

[dfan]

John Fairbairn's Joseki jocosity post, about databases contradicting statements made by pros, got me wondering: are there pros who are fluent with using Go databases and who are doing interesting research with them?

In chess, pretty much any player of International Master strength or above (and many who are much weaker) is constantly doing database research on computers, looking for opening ideas and looking up his opponents' predilections. But I don't get the sense that Go professionals do any of this.

Of course, joseki/fuseki study is very different from chess opening study, and there is a real qualitative difference in how much information you can usefully mine from a database. But it seems to me that a computer-fluent professional armed with a good database and good pattern-matching software could do some pretty interesting research. Does this occur and I haven't heard of it, or is it pointless, or is it just not the kind of thing that pros consider worth exploring?

[Harleqin]

One professional, I believe it was Hane Yasumasa, studied winning percentages for certain fuseki patterns in the 1980ies. I think that today, most players, especially in Korea, try to keep up with the newest joseki developments. I guess that most will also take a look at the games of their next opponent, but Go is not a game where you can gain too much from that.

[GoCat]

I guess that most will also take a look at the games of their next opponent, but Go is not a game where you can gain too much from that.

Really? This surprises me. I always hear about the "style" of this pro or that -- it seems that it would certainly be beneficial to study an opponent's games at least to be prepared for how he/she is likely to play, and possibly even use that knowledge to advantage.

But, what do I know -- I can barely make sense of my own games!

[Tommie]

In chess, pretty much any player of International Master strength or above (and many who are much weaker) is constantly doing database research on computers, looking for opening ideas and looking up his opponents' predilections.
But I don't get the sense that Go professionals do any of this.

Of course, joseki/fuseki study is very different from chess opening study, and there is a real qualitative difference in how much information you can usefully mine from a database. But it seems to me that a computer-fluent professional armed with a good database and good pattern-matching software could do some pretty interesting research. Does this occur and I haven't heard of it, or is it pointless, or is it just not the kind of thing that pros consider worth exploring?

If KingHunt game opening libraries are perhaps 18 plies deep than that covers 1/3 of the length of an average chess game.
The remaining 2/3 of the game are much more linear (I guess) than the respective parts chuban & yose in Go.

Usually I compare the immediate, instinct knowledge of chess openings with the basic instincts for shapes of a Go player.

To learn about the preferences of a specific opponent will be as valuable as in chess, yet the Go game is decided after av. 240 moves..

[hyperpape]

Actually, I have read quite a bit about schools of players in China and Korea studying popular lines of play, looking for innovative moves or refutations. I can't remember who it was, but an elderly Japanese pro recently lectured the younguns on the importance of preparing moves for competition.

That's not the same as using database analysis, but it does suggest that many professionals have the necessary frame of mind to use databases. That said, I've seen relatively few examples of database analysis that obviously hold value.

[pwaldron]

Actually, I have read quite a bit about schools of players in China and Korea studying popular lines of play, looking for innovative moves or refutations.

I've noticed this in a number of Korean books that I have. The best example is probably "Research From the Chung-Am Institute" (something like that). It's a three volume collection that presents a wrinkle in a joseki or fuseki position, and then spends several dozen diagrams examining what happens. It's a lot like Lee Chang Ho's Novel Plays and Shapes--most of the diagrams don't show one side getting destroyed (you fill those variations in yourself), but might rather conclude that one player is falling behind in development in a particular variation. You need a fair bit of strength to follow the analysis, especially if you don't understand the Korean commentary. I put the books on the shelf and figured they would be something I would have to grow into.

Regardless, it's clear that these pros were engaging in a fairly systematic study of joseki/fuseki positions. It may be that this also happens in Japan and the results don't get published in the same way. The old Go Worlds used to have a column on joseki innovations as well, but the analysis wasn't as deep or systematic.

[hyperpape]

I can't point to any right now, but I've read a lot of hand-wringing about how Japanese players do not study in groups the way that the Korean and Chinese players do. I think this is seen cited as one factor that affects performance in international tournaments.

[tapir]

I was told that some years ago when MasterGo was new and presented in Korea, that there wasn't anything similar in existence there but still there was no big interest in databases. I guess from that, that the people who are not satisfied with the available games and reviews are those professionals or would-be-professionals who just replay a thousand games if they have to prepare for something particular. So they don't need a research database as well.

In the audiogolessons Guo Juan repeatedly refers to databases she uses, but obviously it is something she became aware of those in Europe not in China. I guess people really use computers for storing kifu only in the East (or playing on the net).

Regards Tapir

PS I believe joseki innovation doesn't proceed by winning statistics but by inventing a single new sequence turning some usual pattern bad etc. You can spread the message without statistics then.