Certainly open to exploring that more, though it raises the question, “Where’s move 1?”. If we extract them elsewhere, the table is now incomplete from that perspective.
But yeah, if we have some better ideas on how to present this that’d be great to explore.
Well, yes, though in the current implementation that comes down to whether or not the AI figured out if it was on the other end of a ladder or not, if it did it’d detect it as a blunder and wouldn’t be flagged as joseki here.
I’d argue that is true for all excellent moves. The interesting ones are the bad moves, that’s where you can learn. (There may be a difference between reviewing your own games and reviewing games from other, stronger players though. But for me reviewing my own is much more important.)
Indeed a concern. E.g I’m better than Shin Jinseo 9p and Shin Minjun 9p here My road to OGS #1 rank :) (that name was breakfast’s game name, not mine, he failed 
). You can belive me I didn’t cheat with AI in that game because it was played before AlphaGo, but if I played such a great game now there would be many doubters.
I doubt someone with an established reputation or history on the site has to worry about that. It’s mainly fresh accounts that would be under heavy suspicion of cheating, before they have an established rank and 100s of games in their match history.
Right - but that’s the exact point: “flagged as joseki” - is this flag right?
I think I’m arguing that from my understanding of what “joseki” means: a move is joseki whether or not the AI think’s it’s excellent.
Ah - actually, there are two three points emerging here:
Or does something called “Joseki” need to meet some commonly understood definition of “Joseki”?
Is “the AI likes this in the corner” now a commonly understood definition of Joseki?
Isn’t it excellent to find the excellent joseki out of all of them?
and 
Darn nit-picking definition-obsessing computer-programming users. 
Another related question: is there even such a thing as “Joseki” in 9x9?
Would we say “black played joseki here, it’s not really excellent”?
Corner square gave a good example of a bad move that’s certainly not joseki that was classified as joseki, just because it’s early and in the corner and not terrible (because no 3rd line moves on move 3 are terrible, unless you nobi from your first one perhaps!).
Joseki means standard sequences from human go knowledge so it has to be by reference to a database of joseki, a good move according to AI in a corner early doesn’t make it part of that corpus of human knowledge.
Let’s skip over middle game joseki on the side, invasion and reduction joseki etc…
Wikipedia defines Joseki as “studied sequences of moves for which the result is considered balanced for both black and white sides.”, and senseis defines it as “standard sequences of moves played out in a corner that result in a locally even exchange.”
I feel the crux of the issue is we’re going beyond “studied/standard sequences” here at times, so perhaps the term “Joseki” isn’t accurate but I don’t have a better word for it.
I’ll have to invite you to an OJE policy meeting discussion, you can be on my team
Maybe not joseki, but can’t be that bad if an 8p played it: Fang Tianfeng (8p) vs. ??? (5p) | Waltheri's go pattern search (well at least after the other corners had a stone in them)
So the question is “why is this ‘type’ in the list of ‘types’”?
Why do we have “Joseki” as a “move type” in the same set as “Excellent” or “Blunder”.
It appears that the answer is
“because I thought that we all would find it not-interesting or helpful to categorize the move-quality of moves that are 'just joseki, not really good or bad”.
Does this statement (which is not a direct quote, it’s a putative paraphrase) ring true?
Maybe we should rename it to “Strong Opening”, or something of the sort? Alternatively maybe we could mark them differently under the “Excellent” / “Great” category so that the opening corner positions can be visually differentiated still, but still counted as “Excellent” or “Great” ?
Pros are human too and like to have fun / troll / experiment / surprise opponent / use a novelty they’ve studied. I’m sure Fang wouldn’t claim that was joseki (or the Chinese equivalent).
Yes, I think so given their current logic for detection. If everyone gets a few free excellent moves for the first corner moves is that such a bad thing? The issue a separate joseki category really helps with IMO is differentiating a game with lots of long joseki sequences from one with very little: both players getting 50 excellent moves whilst playing out the large avalanche or taisha or flying knife isn’t as indicative of excellent skill as 50 excellent moves in unknown positions (but it’s still pretty darn good!). But then difficulty of position and ease of getting excellent moves also applies outside the opening: a calm territorial game with little fighting is much easier to play excellent or good moves versus a chaotic fighting game, in the former 2 amateur dans could get better scores on these AI comparison stats than 2 top pros in a swingy fighting game.
But to not get bogged down in criticism of the joseki category, overall this is very cool to integrate into OGS. I particularly like the little red dots showing you where the instances of that category of moves were on the graph with their pretty little animation when you pop open the move list.
(And that also half solves the excellent but joseki issue, if they’re all bunched up at the start)
Yeah me too!
It turns out that this behaviour just fell out of the components previously put in place: totally awesome component design in the first instance!
And so … on user interface:
I like the suggested change to the popup itself, but I don’t understand what’s weird
It pops down right under the arrow that you clicked on to open it, which is pointing at it.
Maybe we should highlight the active arrow caret thingy.
(Aside from layout, I just noticed that the move numbering in the pop up is one-off from the move numbering on the move-tree, oops)
I agree that “joseki” seems out of place in this chart. Whether a move is interesting or not shouldn’t affect whether it’s excellent or not. Besides, interesting is a subjective concept isn’t it.
Ok the Joseki category has been removed and those moves are now categorized based on score loss like all other moves. They are still differentiated as being grayed out a bit to call out that they are opening corner moves.
Well chess com for instance has a move categorisation that’s not too different and they too have a “book move” part of their table. Because a lot of openings are very well studied and you’ll find them in theory books.
Maybe honte or “expected” better as joseki?
I saw that too. I think chess openings are so well studied that many opening moves are considered book moves. Even I play book moves for my first few moves. Eventually all moves may be considered book moves. Still that shouldn’t decrease its “interestingness”
