There could be a fun collaboration with Collective development of a server for variants - #12 by benjito
(Sorry to keep throwing in comments in the middle of the game)
4 Likes
Agree, but if like to point out that one of the perks of Vsotvep’s multicolour go too’ its that it does NOT implement any rules, so it is easy to use it for many different variants, including variants that were testing and that don’t have definitive rules yet
6 Likes
Yeah I think there was discussion of a sandbox board/tool which I’ve seen some other sites do which is quite nice.
Allows manually testing things, new games, rules etc.
The crossover I was thinking of is that someone probably needs to make a board anyway if that server thing picks up steam - a method of drawing boards on different graphs/geometry etc. So there might be crossover there - not necessarily anything to do with game logic.
2 Likes
10, 6
Mirror go looks like it would have been very powerful on this board, with the big tengen hexagon. And ladders don’t seem to work here, either? Maybe it would be better to center on a small triangle.
2 Likes
10, 10
Perhaps a good reason to give Black komi!
Although I don’t see any long ladders working, I wouldn’t automatically say that there isn’t a counter to mirror play based on something similar to a ladder.
4 Likes
Against yourself or with an opponent?
1 Like
My kid beat me 
8 Likes
Pretty wild how the white group is completely separated in the middle but it still counts as an eye. I guess it’s like playing on a width 2 board there
1 Like
The possibility of mirror Go being an option does not necessarily imply that Black is at a disadvantage, right? By strategy stealing argument, it seems that the game is still at least a tie for Black, and I would guess that Black likely still has a slight first move advantage, except that White could potentially force a draw, if Black cannot find an effective counter to mirror. Black could potentially counter mirror Go by creating a capturing race.
6 Likes
The only way to find out is by starting a LZRhombitrihexagonalGo training run…
But I’m willing to bet that white still needs komi
1 Like
4 Likes
Any insight on what the convolutional kernels would look like?
9, noon
1 Like
I think it’s equivalent to this slightly weird shape?
4 Likes
Probably very similar to the board 
1 Like
Maybe apply graph convolutional neural networks.
Regular CNNs work on rectangular grids, while GCNs generalize the concept to arbitrary graphs.
3 Likes