It was my first run of the full game, I've played the demo that's why I have a drawer. But you're right, not a completely blind first run, but I still discovered all the new stuff :).

I wasn't able to go much higher like I've seen elsewhere, I'm not good at this kind of game but I'm having fun!

IIRC A* is good to find the best path between two nodes, so if you want to have the whole graph and not solve the game maybe it's not necessary?

Oh, if so much are soft lock states yeah, the graph traversal is even less expansive!

And by "solving", do you mean find the solution, or exploring every state to make the graph? I'm pretty sure it's the second, but for the first, lots of stuff in the graph theory part of compsci :)

Oh yes, repealing force for every node is heavy... Even more with the graph being dynamic!

I think it's O(n+e) to traverse the graph, with the edges. Still linear at its core, but a greater cost compared to only the nodes. And I think puzzle games have more edges than nodes?

Oh, okay! I'd thought exploring the game's states would be the bottleneck... But maybe the design of sokoban is really constrained at its core.
For the viz, I suppose it's the spring-like tension that makes it compute intensive?

Really cool! It isn't too long to explore all the possibilities of a level with only this kind of walk?

Yes, their videos are incredible! It's like nerd catnip for me, always a pleasure to see one pop up.

Anyway, I'll follow this thread with great attention ;)

I love this kind of meta-visualization! Is it inspired by 2swap's videos about Klotski? It's a similar work with graphs about a game. I can't recommend it enough if you haven't seen it yet www.youtube.com/watch?v=YGLN...

You can also do `grep -C {N}` (C=context), it's the equivalent of what you wrote :)

First run 💪