Life In 19x19

There is an apparent clamour for an up-to-date joseki dictionary. Many of us believe josekis are fool's gold.

But even if you say, "Yes, yes, understanding moves is better than memorising them" yet insist that it's good to have a work of reference beside you, you could still be buying a pig in a poke.

In writing the Go Seigen ten-game match books, I was astonished at how often I would encounter, in professional commentaries, phrases like "Black 9 was a novelty by Go" yet the move may have been played centuries before, or even just before by another pro. Of course, they were writing in the days before databases, but on the other hand they had far fewer games to keep track of.

There are similar dubious claims even in the best joseki books. I picked one book at random and looked for the first three "new" or "most/-est" type phrases I could find. It only took about five seconds, using Jungsuk in Our Time. Then I checked the facts. Similar results will be obtained with any other dictionary.

On page 172 it said "Up to 11, this is one of the most famous jungsuks". Well, it only occurred in 9 games out of 62,000 in the GoGoD database.

On page 153 it said "White 1 is an aggressive modern move". Well, only if you count 1972 as modern, and if you overlook that the move appeared once in 2009, no times in 2008, and three times in 2007 - out of 35 times all told.

On page 155 it said "Black 1 was one of Go Seigen's favourite moves during his heyday." Hey ho, but no it wasn't. He played it three times in 855 games and had it played against him three times.

This was prompted indirectly by a question from Jochen Fassbender about a joseki where some pros thought it was bad for Black and others thought it was good for Black. Who do you trust?

As many people now know, the only reliable advice on josekis is "White 8 is bad".

What is this "white 8 is bad" talk?

Kirby wrote:What is this "white 8 is bad" talk?

Take a look at the "New in Go" column no. 12 on the GoGoD website (I would link, but the website is composed of horribly organized frames).

John Fairbairn wrote:There is an apparent clamour for an up-to-date joseki dictionary. Many of us believe josekis are fool's gold.

But even if you say, "Yes, yes, understanding moves is better than memorising them" yet insist that it's good to have a work of reference beside you, you could still be buying a pig in a poke.

In writing the Go Seigen ten-game match books, I was astonished at how often I would encounter, in professional commentaries, phrases like "Black 9 was a novelty by Go" yet the move may have been played centuries before, or even just before by another pro. Of course, they were writing in the days before databases, but on the other hand they had far fewer games to keep track of.

There are similar dubious claims even in the best joseki books. I picked one book at random and looked for the first three "new" or "most/-est" type phrases I could find. It only took about five seconds, using Jungsuk in Our Time. Then I checked the facts. Similar results will be obtained with any other dictionary.

On page 172 it said "Up to 11, this is one of the most famous jungsuks". Well, it only occurred in 9 games out of 62,000 in the GoGoD database.

On page 153 it said "White 1 is an aggressive modern move". Well, only if you count 1972 as modern, and if you overlook that the move appeared once in 2009, no times in 2008, and three times in 2007 - out of 35 times all told.

On page 155 it said "Black 1 was one of Go Seigen's favourite moves during his heyday." Hey ho, but no it wasn't. He played it three times in 855 games and had it played against him three times.

This was prompted indirectly by a question from Jochen Fassbender about a joseki where some pros thought it was bad for Black and others thought it was good for Black. Who do you trust?

As many people now know, the only reliable advice on josekis is "White 8 is bad".

It seems that the authors of these books don't really do any research, they just write their own opinions as to good/bad, and the history of moves is mostly what they heard growing up or from other pros. I'm not surprised. What 9-dan pro is going to spend the time to do a thorough study, using databases or not, or investigate the opinions of many other pros? And what about proof reading? Proof-reading a large joseki dictionary would be a huge task. I don't know if it is true but I remember reading someplace that Ishida's dictionaries used a lot of amateur analysis. In the end we have to form our own opinions as to various move sequences, as do the pros.

BTW, my favorite part of the Ishida dictionaries is the reference figures showing game positions. Unfortunately I hear there won't be any of these in the new edition by Takao.

gowan wrote:BTW, my favorite part of the Ishida dictionaries is the reference figures showing game positions. Unfortunately I hear there won't be any of these in the new edition by Takao.

That's true indeed. You would have to refer to the also new edition (2008) of Kihon Fuseki Jiten.

John Fairbairn wrote:There is an apparent clamour for an up-to-date joseki dictionary. Many of us believe josekis are fool's gold...

You're spoiling the party, John! Thank you. We need more of this type of analyses. However, what this suggests about the value of learning Josekis is another matter entirely.

My original post was just a casual response to a query I'd had from someone. However, I have since responded to a query of my own when looking at a recent Weiqi Tiandi, and I ended up rather perturbed about the Ishida dictionary. I had found flaws in it before, but thought little of them. Today's investigation left me distinctly underwhelmed about the book, though.

The position that started me off was the one below.



Luo Xihe did a tewari analysis of the joseki (as he called it) in the upper right, and said the position was equivalent to the one in the lower left in which Black had then exchanged 13 for 6 (not awful, but ajikeshi-ish), then 9 for 12 (awful) then 3 for 4 (also awful, White's empty triangle notwithstanding). The conclusion has to be that the joseki status of this position could fairly be called into question.

No problems so far. I then checked this position in the GoGoD database. There were just three occurrences, and every one was by Hashimoto Shoji. That did not amount to joseki status in my opinion.

I therefore checked the position in Ishida. Much to my surprise it was given there with two-stars, meaning a STANDARD and BASIC joseki (Dia. 372 on page 177 of Vol. 1). No way. In the text roundabout there are other niggly problems. One is the claim that this pattern is seen more often than another basic pattern, which doesn't seem to stack up. Then there is the comment (page 173) that White 4 is "the modern method", which is not what I would really call a move played in 1928 (i.e. pre even New Fuseki) and often immediately thereafter. But leaving these aside, the core problem is that the Ishida book is claiming standard and basic joseki status for a position that we have seen just three times from one player and which can also be put under a cloud of suspicion in tewari terms. (Bear in mind that the GoGoD database is pretty comprehensive now in its coverage of published Japanese games going back to the 1920s.)

The next step was to check in another joseki dictionary. I selected the Joseki Shojiten (1966) which is really by Kitani but is attributed to then dead Suzuki Tamejiro as it was derived from his Great Joseki Dictionary. There it does not give a diagram for this supposedly basic joseki at all. Instead it gives the line where Black 13 is at 14. The commentary, however, says that "There have been occasions when Black has played [this alternative, i.e. 14] at 13". That strikes me as a perfectly adequate and accurate statement of the database obervations. Furthermore, the alternative line (i.e. Black 14) is described there simply as "even" and not as joseki (the position evaluations in this book follow a strict scale, in which even ranks below joseki, rather than the "elegant variation" style used in Ishida).

When you consider that the GJD, and thus the Kitani book, was based on massive studies by many professionals of the calibre of Kitani (Suzuki's pupils) at a time when they would have been glad of the extra income, whereas the Ishida book seems really to be the work of a couple of amateurs, this discrepancy is not surprising. But in my mind it definitely casts a dark shadow over the reliability of the Ishida book, and likewise of any future book that follows the same method of compilation.

PS How do you get numbers on stones?

John Fairbairn wrote:My original post was just a casual response to a query I'd had from someone. However, I have since responded to a query of my own when looking at a recent Weiqi Tiandi, and I ended up rather perturbed about the Ishida dictionary. I had found flaws in it before, but thought little of them. Today's investigation left me distinctly underwhelmed about the book, though.

The position that started me off was the one below.

(;SZ[19]
;B[qd];W[oc];B[ld];W[me];B[od];W[md];B[pc];W[ob];B[mc];W[nd];B[oe];W[lc];B[pb];W[kc]
;B[cp];W[eq];B[dq];W[er];B[ep];W[fp];B[eo];W[iq])

Luo Xihe did a tewari analysis of the joseki (as he called it) in the upper right, and said the position was equivalent to the one in the lower left in which Black had then exchanged 13 for 6 (not awful, but ajikeshi-ish), then 9 for 12 (awful) then 3 for 4 (also awful, White's empty triangle notwithstanding). The conclusion has to be that the joseki status of this position could fairly be called into question.

No problems so far. I then checked this position in the GoGoD database. There were just three occurrences, and every one was by Hashimoto Shoji. That did not amount to joseki status in my opinion.

I therefore checked the position in Ishida. Much to my surprise it was given there with two-stars, meaning a STANDARD and BASIC joseki (Dia. 372 on page 177 of Vol. 1). No way. In the text roundabout there are other niggly problems. One is the claim that this pattern is seen more often than another basic pattern, which doesn't seem to stack up. Then there is the comment (page 173) that White 4 is "the modern method", which is not what I would really call a move played in 1928 (i.e. pre even New Fuseki) and often immediately thereafter. But leaving these aside, the core problem is that the Ishida book is claiming standard and basic joseki status for a position that we have seen just three times from one player and which can also be put under a cloud of suspicion in tewari terms. (Bear in mind that the GoGoD database is pretty comprehensive now in its coverage of published Japanese games going back to the 1920s.)

The next step was to check in another joseki dictionary. I selected the Joseki Shojiten (1966) which is really by Kitani but is attributed to then dead Suzuki Tamejiro as it was derived from his Great Joseki Dictionary. There it does not give a diagram for this supposedly basic joseki at all. Instead it gives the line where Black 13 is at 14. The commentary, however, says that "There have been occasions when Black has played [this alternative, i.e. 14] at 13". That strikes me as a perfectly adequate and accurate statement of the database obervations. Furthermore, the alternative line (i.e. Black 14) is described there simply as "even" and not as joseki (the position evaluations in this book follow a strict scale, in which even ranks below joseki, rather than the "elegant variation" style used in Ishida).

When you consider that the GJD, and thus the Kitani book, was based on massive studies by many professionals of the calibre of Kitani (Suzuki's pupils) at a time when they would have been glad of the extra income, whereas the Ishida book seems really to be the work of a couple of amateurs, this discrepancy is not surprising. But in my mind it definitely casts a dark shadow over the reliability of the Ishida book, and likewise of any future book that follows the same method of compilation.

PS How do you get numbers on stones?

Thanks for posting this. Do you suppose errors like this will be cleaned up in the new Takao edition? And what about the untranslated Japanese revision, supposedly by Ishida? Nakayama-sensei, in his book Joseki Hazure, said that errors in joseki books tend to persist for 30 years or more.

There is an interesting Japanese book on the evolution of joseki. I have posted some info on it at my blog here.