Josekle #30 (9 moves, hard)
Josekle #30 (7 moves)
Fusekle #12 (6 moves)
1 Like
Josekle #30 (7 moves)
Josekle #30 (9 moves, hard)
Fusekle #12 (6 moves)
Fusekle #12 (6 moves)
My first guess completely inverted black/white 
1 Like
To get an idea for how good we are at Fusekle (and how much the go aspect helps), it could be interesting to analyse the game where we’re just trying to guess a random permutation of a given length.
For length 2, there is a 50% chance to get it on the first guess, otherwise you will get it on the second guess.
For length 3, you will get either 0, 1, or 3 green from your first guess. If you get 1, you’re guaranteed to get it on the next guess. If you get 0, you could again get 0, 1 or 3 on your second guess, but you’re guaranteed to get it in at most 3 guesses. (after two 0’s there is only one possible permutation remaining)
It’s easy to see that N guesses should suffice for a sequence of length N: you could spend your first N-1 guesses guessing cyclic permutations of the same sequence, and then you have perfect information for your last guess (every move has either been green once, or purple all the time, in which case you can place it by elimination).
It seems plausible that N is also the best we can guarantee, but I’m not 100% sure. Let’s do some simulations to not have to think too hard. This is the outcome (number of guesses required) of 10 000 simulations for each sequence length:
Data
Length 3:
1: 16.81%
2: 66.19%
3: 17.00%
Length 4:
1: 4.67%
2: 45.12%
3: 43.56%
4: 6.65%
Length 5:
1: 0.73%
2: 22.01%
3: 51.34%
4: 24.70%
5: 1.22%
Length 6:
1: 0.13%
2: 7.77%
3: 39.71%
4: 42.33%
5: 9.60%
6: 0.46%
Length 7:
1: 0.01%
2: 2.37%
3: 22.07%
4: 46.04%
5: 25.67%
6: 3.76%
7: 0.08%
Length 8:
1: 0.01%
2: 0.58%
3: 10.50%
4: 35.33%
5: 39.27%
6: 12.98%
7: 1.31%
8: 0.02%
Unless I’m missing some way the guessing can be made smarter (I’m just randomly picking a sequence from the ones consistent with earlier hints), this data supports the theory that N guesses is the best you can guarantee. Probably there’s some easy proof for this.
Expected number of guesses for sequences of different lengths:
(also calculated from 10 000 simulations each)
3: 2.0
4: 2.5
5: 3.0
6: 3.5
7: 4.1
8: 4.6
Looks to be about (N+1)/2.
So, for today’s Fusekle of length 6, doing it in 3 guesses or below is better than you could get on average for a random sequence (meaning that you successfully used some go knowledge to your advantage).
2 Likes
Well, one could still get lucky while applying zero Go knowledge. We can only argue, given enough data, that one might have statistically significant evidence of applying Go knowledge.
On the other hand, when I take more than N guesses to get the Fusekle, what might that say about me 
1 Like
Of course 
It says that it’s not easy to see all information at a glance with the current hint system 
1 Like
Josekle #31 (8 moves, hard)
Josekle #31 (4 moves)
Fusekle #13 (8 moves)
Further evidence of lack of Go knowledge
I am planning to do something about this, by the way. The purple cross has not been forgotten about. Just trying to figure out what the appropriate logic should be. I’ve also recently noticed a weird issue with how I’m propagating persistent hints that I need to address.
2 Likes
Fusekle #13 (8 moves)
(just feels like I’m cheating 
)
3 Likes
Josekle #32 (6 moves)
Fusekle #14 (5 moves)
Josekle #32 (15 moves, hard)
Yoooo @Jon_Ko the squares are lit 
3 Likes
Does anybody know whether there’s a way to prevent the lines of 3/fewer emojis from getting sized up? Or do I just need to remember to only try variations 4 moves or longer? 
2 Likes
Putting something else on the line makes the emojis display in small mode. One way to do that, without inserting any visible text, is to put a dummy tag like <blank>, which will not be shown.
:green_square::green_square::green_square:
:green_square::green_square::green_square: <blank>
Further example, in my post above:
2 Likes
Fusekle #14 (5 moves)
Fusekle mild spoiler
Guessed wrong about the symmetry, otherwise would have got it in one.
I hope you try Fusekle #14
Josekle #32 (6 moves)
Josekle #32 (15 moves, hard)
2 Likes
Mild fusekle spoilers
I did it earlier today, was also unlucky with the symmetry
Fusekle #14 (5 moves)
Fusekle
There are also some solutions with identical end positions but different orders (and possibly some which are completely different but give the same purple pattern), so even if I somehow remembered the entire solution-set I would need 2 or 3 guesses sometimes.
2 Likes
Josekle #33 (14 moves)
Fusekle #15 (8 moves)
Josekle #33 (7 moves, hard)
Tried hunting around a bit for the last move, then used the OJE
Fusekle #16 (8 moves)
1 Like
Not this variant specifically but a similar “recognise a joseki” game was played on tsururin channel recently and in case you were wondering how much modern pros study joseki, watch this bit with ichiriki guessing the joseki (that the hosts got wrong funnily, he shows the correct variation at the end of this segment)
https://youtu.be/hzYJ6se1qHM?t=291
And by recently I meant last year, because it’s common knowledge that it’s not 2021 anymore.
1 Like
Josekle #34 (8 moves)
Josekle #34 (12 moves, hard)
Fusekle #16 (8 moves)
Inspired by a slightly earlier video, there was even a forum thread about this game:
As for Josekle, it is possible to play it in “one-color mode” where the positions of all of the stones are first revealed with purple hints. This can be toggled on the help page. Also, the Fusekle variant is set up to use one-color mode by default (but this can be toggled off for a “blind start”).
1 Like
This feature is now live on both Josekle and Fusekle. Basically, the only change is that a purple hint (on an empty point) will now also have a purple cross, if the next click should not be on that point (as implied by the timing of earlier purple hints).
2 Likes