Basically this, but I think thereās also a length cap for hard mode, because michiakig extracted only a certain number of moves from OJE.
Maybe it makes sense to make the hard mode solutions (going forward) disjoint from the normal mode solutions?
I wonder if there was some issue with a lack of full randomization in the puzzles. I could always reshuffle the ordering of future puzzles.
I think I imagined people would either play hard or normal not both, but it seems like people are playing both. Personally I am not sure there is that big a difference, there are enough uncommon joseki labeled ābasicā that normal is challenging for me, and hard would not be that different?
I looked at this briefly yesterday, and although there were a few in a row that start off the same, I didnāt actually see evidence the solutions are ordered somehow. The script did shuffle before producing the list, so although there is of course a possibility of a bug, I attribute it to the typical bias about ārandomā sequences not being ārandom enoughā (of course for the game, perhaps a non-random sequence is desirable, for instance alternating first moves, etc. but this would be more work)
I guess hard mode should just be viewed as āallā mode with hard and easy puzzles, where as normal cuts out any too hard puzzles.
The āhardā puzzle had only 4 moves today, but it was still a bit tricky to work out
Josekle #22 (4 moves, hard)
Fusekle #4 (5 moves)
Josekle #22 (9 moves)
Josekle #23 (6 moves)
Fusekle #5 (7 moves)
Josekle #23 (9 moves, hard)
Josekle #23 (6 moves)
Josekle #23 (9 moves, hard)
Josekle #25 (8 moves)
Josekle #25 (14 moves, hard)
Fusekle #7 (8 moves)
I havenāt been sharing my scores here since Iāve mainly been posting them in Discord, but I had to share today because the hard one looks like pixel art and fusekle has a hilarious dearth of green
Josekle #25 (8 moves)
I finally used OJE to help find the last move, then I still misplaced it once more.
I had never seen the joseki before but I really like the resulting shape
Josekle #25 (14 moves, hard)
Regular Josekle
Josekle #25 (8 moves)
Hard
Josekle #25 (14 moves, hard)
Seems a bit much to call this a joseki when the following is the entire corpus of its play in Waltheri.
According to that database, itās only been played twice, both times by Hashimoto Utaro and both times in the 1920s.
I think the current method of selecting josekis is good, but of course there are ways to improve it. Iād be willing to spend some time doing that, but at the moment I donāt quite know where to start.
I guess the question is, should it be in OJE at all? Or maybe it should still be there but not tagged with āJosekiā, or tagged with something indicating its uncommonness?
I think the options are:
- Lobby OGS devs to open up access to OJE to add better tagging
- Manually curate the collection we have scraped
When putting the new collection together, I wasnāt sure either of these had a clear pathā¦
Iād like to check the josekis from OJE against how common they are (e.g. according to Waltheri). Tags could also help.
I think rarely played sequences maybe shouldnāt be part of the daily solutions, but should still be part of the joseki dictionary.
Is there a reason to check against Waltheri specifically? I was under the impression it didnāt actually search the whole database for particular positions. TBH I donāt really understand how it works at all. Is it a better option than GoGoD?
All Iād want is a curation procedure, that I could automate to some extent and that takes into consideration how complex sequences and their branches are and how common they are.
I dunno if josekis have to be popular.
Popularity is not a proof of quality in itself.
Unpopularity is good for inspiration (Playing go) and difficulty (Playing josekle)
The unpopular ones might be especially good for hard mode
Fusekle #7 (8 moves)
Fusekle #8 (8 moves)
By the way, a slight change to Fusekle is now live. The puzzles have now been randomly flipped and rotated. Maybe most people would not have even noticed, unless you were using the polite triangle as a hint for the first move.
Josekle #26 (7 moves)
I like that. First corner is too much of a hint.