For ages now, I’ve been aware that we need to change the OJE position tags.
I keep procrastinating about actually doing it because I’m not sure what the right answer is.
Maybe we can thrash it out here?
One criteria I have is that a person should be able to set a filter that says “Show me positions that meet the key criteria of OJE: that are solid joseki (in the sense that they are an even outcome for both players) with a proper source”
Over time, we’ve discovered that we need to be able to distinguish other things like “out of date”. @aesalon has some solid ideas about this.
It may be that we also need a more sophisticated filter - one that can say “don’t include” as well as the basic one we have now that says “only include”.,
That could be a contender. Rather than “pre-AI”, I think it might be more helpful to be able to say “outdated by AI” … because some pre-AI joseki still survive?
Another problem to be fixed is the standard tag “Joseki - Done”. It is trying to say that “this is a point in the sequence where the outcome is balanced and you could tenuki or play situational moves”.
It’s “Done” in the sense that you’ve completed a Joseki sequence, but often there are follow on sequences, so “Done” doesn’t seem right.
Just call it “Joseki” instead of done. I guess there are some cases where you could continue an exchange or two and it still be “Joseki” but then anything after that “continuation”.
To me, I’m not sure that having continuations is quite right for a “Joseki dictionary”. Isn’t the point that Joseki are sequences that are played in one go usually because once started they are urgent to settle. But continuations might happen many (many) moves later given the rest of the board position.
Maybe ideally you wouldn’t have continuations at all or have them in a while separate place but I guess that ship has sailed!
The language you use depends to some extent on who you think the audience is. OGS is reputedly oriented toward kyus and beginners, and it is highly international. In that light, “Done” has the advantage of being short and clear, meaning the position is even and the next move can be tenuki.
Audience is especially important in regard to AI inputs. Dwyrin has long argued that kyus are better off playing simple, solid moves. “Old” moves typically lose less than a point. What is that to a DDK? If this consideration is regarded as important, it would be better to avoid language that implies that AI moves rule out all others all the time.
“Old” or “outdated” really describe only the frequency of contemporary use (some moves were old even before AI came along). In contrast, “outdated by AI” implies an inappropriate force of authority (outdated for whom?) and suggests the intimidating thought, “Gee, I don’t want to be outdated.” The most nearly neutral phrase I can think of, which at the same time brings AI into the evaluation, would be “AI prefers…” or “Preferred by AI.” This way, a weak player like me may conclude, “Okay, I’m not AI, and I may not be capable of the necessary follow-ups.”
stable (it’s not important to continue locally) alerting (it’s important to continue locally) critical (should be answered almost always)
settled (all groups have a base) unsettled (some groups lack a base)
equal (“fair” result for both sides) advantage_black / advantage_white (one side got a better result)
common (often played) rare (not often played)
complicated (requires precise play) simple (relatively forgiving to mistakes)
Looking at existing tags:
I don’t know what’s the use-case for some of these. If a person looks at a sequence, they’re probably able to understand on their own, whether “Black gets the top” or not. For the filter, would one ever search for positions where “Black gets the corner, top and right”?
I proposed having one tag for antiquated joseki along with our current tag. Maybe rename or define the ‘current’ tag to be post AlphaGo. The old ones would be the ones from Edo or maybe Shin Fuseki period that really didn’t pick up traction. The ones that are in between wouldn’t be labeled with either tag.
I agree and re: the rest of this post. I think we have settled/no tag/fighting as kind of a spectrum itself. I think complicated could be a good addition to go along with Basic and a Common tag might be a pretty good addition.
Filters are for asking OJE to show you only sequences that lead to outcomes that you are interested in.
So you very much might want to filter for “Black gets the top” or “Black gets the centre”.
(Admittedly, "Black gets the corner, top and right” is kind of corner case, I think it only got in there because some Joseki do declare this as the outcome.)
The important thing to realise is that the filters, and the tags that they use, are for finding sequences that you want before you are looking at a specific sequence.
You Go to OJE and see a ton of potential joseki. But what you want is a 3-4 joseki that gives white the right side, so you click the first move (3-4) and select the filter that says “White gets the right”.
Similarly, the original purpose of “Joseki: Done” was so that you can filter out sequences that don’t end with “proper joseki”.
Unfortunately, one thing I can see is that if we add “Joseki - outdated” or similar, then we’ll need a filter that says “don’t show me these”. Right now the filters are all “show me only these” type.
Of course "Joseki - current’ is one way to filter out sequences that lead to “Joseki outdated” outcomes.
I believe I’ve said this before but all the positional tags that are ‘and’ should be removed. Then it would work like you want. If someone really wants a joseki that ends with the corner, they could filter for that and then joseki that give the corner and a side would still appear. Rather than now, they would only see joseki that only get the corner.
I’ve never really used filters and don’t know much about joseki, but anyway I’m confused about what you are saying. If I want an outcome, don’t I need to be aware of possibilities along the way that would let my opponent prevent it? Why would I want to hide those?
For example, if I filter for “White gets the corner” I can see this move:
But then at the next step, A is filtered out:
So I don’t get to see how Black could stop White from getting the corner. (And the only reason there is an “A” there is because the text happened to mention it.)
