A tesuji missing from a joseki line

I found what seems to be a pretty key omission from an OGS joseki line that emerges from one of those newfangled 3-3 invasions. I tried to leave some comments but for some reason the comments got attached to this position instead of becoming a question about the new move. Linked here:

This is the missing tesuji, which greatly improves this line for white, so long as white has a working ladder to the lower left. Depending on the board situation, suddenly this branch may become quite much worse for black.

Why is this a much better move for white if whte has a ladder? Well this is the baseline we are comparing against that OGS has as the main variation:

Okay so after the new move, if black naively tries the same thing, white captures directly. Clearly a big improvement compared to baseline:

Black would better take once first, and take sente. But white later #5 capture still means that black has to come back with #6 to S19 anyways. So we get this. White is still much thicker than in baseline, due to having captured the stones off the board.

The obvious question - what if black runs? Then black gets into trouble too. White plays out some ladderish moves first then captures:

White capture is sente against the corner, black has to come back to S19, it’s too good to allow white to save and connect everything. But if black S19, then white gets 6 stones in a ladder.

So if ladder to lower left is good for white, this tesuji works.


Before examining the question, has T16 (at that sequence point) even been seen in professional play?
I can’t find a reference for it in Waltheri.

OGS Joseki policy is that sequences without professional play should, if added at all, bear some sort of marker to indicate that.

I don’t think that OGS Joseki should even be hosting sequences like Ke Jie (9p) vs. An Sungjoon (7p) | Waltheri's go pattern search which only have a single game of play. The job of supplying that level of detail belongs properly to Waltheri, which packages the information in the context of players, event, and most importantly date. Explaining lines with less than ten games of play is really not our purview.

What’s the difference between a “fighting-style joseki” and just a “fight”? It’s the quantity of professional play in the line. I’d like to see every obscure position in the explorer tagged with a game link, eg.

following Shin Jinseo – Park Jungwan 2020 game link


To elaborate, it’s not sufficient that it has been “played in professional play” … that might be one game.

To be considered Joseki and therefore a candidate for OJE, it needs to have been established as Joseki by a pro-level human source.

This is the policy. I’m aware that there are exceptions in OJE, for various reasons - mostly reasons like “we hadn’t figured out how to apply this policy at that time”.


… that being said, discussion like this is absolutely welcome, and if we can establish credibility for new lines, I’d love to see them added.


Wiktionary’s policy for addition of a new word is that it has to appear at least three times in published material, with a total timespan of at least a year.

We could have a similar first entry point for what can be considered for joseki status:

“A sequence to be considered for joseki status should have appeared in at least three professional games, with a total timespan of at least a year.”

If it meets this criterion, it can pass to the second round of evaluation.

I don’t have a strong opinion on whether the line from T16 should be there in the first place or not. :slight_smile:

I do think that if it’s included, then it’s misleading and a disservice to omit an important tactic. (And if the sequence up to before T16 does occur in pro play but T16 doesn’t, perhaps, the existence of this tesuji could be a reason why it doesn’t, although of course having a ladder wouldn’t be absolutely consistent).


@shinuito are you able to take a look at this and comment?

The line in OP was a questionable variation published in a Bouzlich book that hadn’t seen professional play. I added some more common variations at prior points in that sequence but it is a fairly obsolete joseki already and never saw much play.


This has been about my criteria as well.

But then how do you handle situations like this where you have ~10 games up until the 35th move in the joseki and then it diverges into 5 branches? Do you include the line at all? Some exceptions are made for this kind of situation and that is where many inaccuracies and incomplete variations come from.

Though as said, this case is slightly different because it was just copied from a Richard Bozulich book so kind of skipped any further criterion.


Richard Bozulich, though a very established English-language author, is an amateur.

I don’t think he has the authority to confer joseki status on a sequence.


Agreed and it had been discussed after some of the variations were put up. I am not sure what the conclusion was exactly but it was something along the lines of double checking any variations from his books with pro games.

Also note that the sequence in question isn’t actually marked as joseki. Joseki will usually* end with a tag saying Joseki in the heading or below the description with other tags and also involve nothing but Ideal moves.

Edit: Also more on topic, I will add the tesuji in OP when I look at my larger database of pro games. I know I have seen this ladder variation played at least in a top pro game on Fox. I also want to find out exactly which move should be marked as doubtful. If the ladder is working, I suspect that R19 is actually the branching point rather than the linked position.


You might be right, but looking at Walteri and seeing one game and concluding that it’s only one game doesn’t seem right. It’s an incomplete database of games, that has a search function. There’s other databases like smartgokifu and go4go, although admittedly searching through go4go would be tedious. “At least one game” I believe is more correct unless you’re 100% certain that it’s just one game.

However, if the sequence up to a point has 20+ games played, and you try to show where a certain move ends up, is it surprising that a 30 or more move sequence ends up at low game count by the end?

I’m not sure which part to comment on. I suppose the tesuji in the op is better if the ladder is good. I didn’t look for all possible variations at the time.

Yes the sequence came from Bozulich, I usually tag the lines with the source.

I can (help) search for whichever lines/variations and link/upload the games I find if you want. I could probably use Smartgo as a way to use go4go too if I can find the event/player/year. Should I exhaustively look for the line all the way to the OP’s J16. Smartgo also throws up the ke jie game. Should I look up the line up to a much earlier point?

In any case remove/relabel whatever you want.


True, but the burden of proof is to find those other games and supply them, not just assume they exist.

But yeah, I should have said “one obvious game”.

True but a distinction also needs to be made. Maybe the sequence up to a point is a joseki, and after a point there’s too many lines to consider to be part of the joseki. Some other joseki we have just say “fighting variation” and leave it at that.

So after a point they can just be either trimmed out entirely, or they could be labelled mistake and refutation, or questionable move etc.

I don’t think I’m the person to evaluate that, but I was trying to help populate the joseki explorer with variations I had access to.

I’m certainly not strong enough to evaluate these 3-3 variations as good for one player or even.


I favor removing the sequence from 2 here, because it bucks the inclusion criteria. Here’s the inclusion criteria from the Policy/Guidelines:

For starters, the sequence from 2 here is not joseki, because it is not “credible” or “established” under the guidelines. Mr. Bozulich is an amateur, and the sequence has only one professional game as far as I can tell (which does not continue to the position first linked above but follows at different path).

As far as I can tell, the path is not a mistake or refutation either.

So in my view, we’d need to decide if the sequence is interesting and valuable enough to include. I’m not sure it is.


It could be interesting and maybe valuable(debatable considering how niche this pattern is) to know the ladder sequence and while 2 isn’t credible/established as a joseki line, it can revert to the position in these three games.

edit: This is the post that probably spawned this discussion. It might be interesting because it is usually AI’s top choice but they have the blind spot for the ladder. It is kind of a funny move that even when you correct for the ladder blindspot, it is basically the best move (+/- 0.2 points) in the games I’ve looked at. So you’re right it’s not a mistake/refuation either. As I said, I doubt its usefulness because it is such an obscure joseki and it wont be missed if removed.

Removing it probably sets precedent to clean up a handful of other variations. As mentioned above, some joseki branch rapidly after a bunch of moves and we either have to choose between excluding moves that were played only a couple of times and leaving things unfinished or adding them despite us not calling them Joseki.


We certainly owe shinuito thanks for taking the time to help get OJE off the ground by supplying the information that he had :bowing_man: . He was one of the notably few people that really put in the effort that made it come to life by providing content, and I’d like to take particular care to make sure that the discussion centres around “do we want to make a change in this case?” rather than “Should this have been put in, in the first place?” - the latter having the undesireable result of asking shinuito to defend something that does not need to be defended.

Note that when this was added by shinuito we were still in the process of deciding what was in and out, and the fact that it was in a published book definitely met the criterion of “Credible”.

I think the fact that a book might be published by an amateur wasn’t even a consideration.

Hence, the fact that it came from a book was “more than sufficient” at the time - there was no additional burden of proof.

Being able to actually source professional games to back up information from books or other credible sources is a rather newer idea. I wonder if we’d actually be able to do this for random sequences found in books by actual pros … especially older books?

It sounds like we could do with some criteria for at what point “being in a published book” amounts to Credible for Inclusion (I for one had certainly assumed naively that it was in itself enough).