Like many players, I love the deep positional reading and tactical fighting in Go, but opening memorization (joseki and heavy fuseki study) can sometimes make early games feel a bit repetitive.
Taking inspiration from Chess960 (Fischer Random Chess), I’ve been thinking about a concept called Go 960.
The Core Concept
Exact Point Count: Every board retains exactly 361 intersections (the same total volume of play as a standard 19 \\times 19 board).
Standard Rules Preserved: Plays follow regular Go rules—orthogonal line grid, 4 liberties internally, standard capture, eye space, and territory enclosure.
Dynamic Shapes: Before a game, a board shape is selected from a curated pool of 960 valid non-square geometries.
Because the corners, edges, and outer borders shift, memorized joseki patterns become obsolete on move one. Players have to rely entirely on pure shape intuition, local reading, and adaptive positional judgment.
Potential Board Shape Categories
To keep games functional and ensure group survival/two-eye shapes are naturally possible, shapes avoid 1-intersection-wide corridors and instead focus on interesting spatial dynamics:
The Hourglass / Bottleneck Series: Two large basins connected by a narrower central chokepoint (e.g., fighting for control of the middle to sever or connect top/bottom groups).
The Cross / Plus Series: A central square hub extending into four distinct outer wings (localized corner battles connected by a high-value central hub).
The Hollow Ring (Donut): A larger outer boundary with a hollow center void (no Tengen; introduces internal concave corners and looping dragons).
Asymmetric L-Shaped & Terraced Boards: Jagged or offset borders that drastically change diagonal shicho (ladder) paths and edge efficiency.
Modified Rectangles / Aspect Variations: Non-square 361-point layouts where the balance between side influence and central territory shifts.
Questions for the Community / Feedback Wanted:
What custom 361-point shape features do you think would lead to the most interesting tactical fighting? (e.g., internal concave corners, chokepoints, hollow centers?)
What layout elements should be strictly avoided to keep games balanced and fair for both Black and White?
Would you play a “Go 960” variant mode on OGS if it were available?
Looking forward to hearing your thoughts and ideas for specific board layouts!
Full disclosure : I got an ai to help me with writing this but the idea is mine. Also I have no actual coding skills so I am hoping someone will take this idea and make it happen!
You can just play weird but not that bad moves to avoid joseki on the normal board. When I played Matthew Macfadyen 6 dan, I was always impressed how quickly and easily he took us off the beaten track so you have to think for yourself rather than rely on joseki knowledge.
I’ve played on some strange small board shapes (like the Milton Keynes go board) but found them too different: I’d rather just not play joseki and standard fusekis on the normal 19x19 board to avoid the repetitiveness of that, than go to totally novel board shapes. At the very highest levels (pro) that desire to avoid joseki can make you play suboptimal moves the other player can benefit from, but for most of us we can just play weird moves to add variety and the game’s still even. And there’s even pros who play weird moves like Gan Siyang.
Also not sure why it needs 960 specifically, as that’s the number of permutations in Fischer random and nothing to do with Go, unless it’s just for the number recognition in the name.
I kinda like the idea, but it can be improved on in my opinion. Especially the katago part is somewhat inelegant. It increases the cost of generating random positions (there is a reason the website is limited to 100), it makes it impossible to implement the algorithm “offline” and there is good reason to also doubt the accuracy of katago in very novel positions like these ones, because they are very out of distribution.
I think just randomly generating a position by chance (for example with a D19 “offline”) - and then doing a komi bidding process - is superior. I quote: “[…] the players bid against each other in a standard auction, the player who bids the most points taking Black and giving that number of komi”.
Btw the linked website https://random-go.antontobi.com/ also has a truely random postion generator if you use the “custom” setting.
One more thought is that i also really like the idea of custom boards, but again, i think it needs the komi bidding process.
ou’re spot on that 960 is borrowed from Fischer Random for recognition rather than math—it just cleanly signals the concept of a standardized pool of variants.
To clarify, though, the goal of Go 960 isn’t just to dodge opening theory—it’s about creating a completely fresh spatial experience every game.
Playing a weird move on a standard 19 \\times 19 board still happens on the exact same square topology we all know. But on a board with an Hourglass waist, internal corners, or a hollow center, whole-board concepts like ladder paths (shicho), group connectivity, and direction of play change structurally.
Sticking to exactly 361 points ensures you still get the macro-scale depth, fighting volume, and game length of standard Go, but on a map where neither player has geometric familiarity.
If someone wants a unique board shape, then I do not have any suggestions.
But for those who are looking for unique board positions and to break out of joseki, I also recommend https://random-go.antontobi.com/
I have played several times at my local go club. It is a lot of fun!
We found you can even approximate handicap games by generating a random position and then removing a certain number of white stones
I don’t know if I would play ranked games this way, but if there was an option on OGS I would definitely use it. I suppose you could always create several positions and the “fork” a game with a friend
AI is just a tool you used. You can’t just get the praise when the result is good, but blame AI when it’s bad. Either way it was your decision to use the tool, and to use the result (rather than throwing it away).
