I noticed that with the map in the top right, I am able to get a much clearer snapshot, to the point, that I thought, hey, I bet at a certain point, due to procedural generation of each "section of the map" that the exit or WP is in a general direction, with each section explored further increasing the confidence.

For people who forget "rules" the overlay would easily show the rules, but again, to be clear this is just confidence, so if after 3 or 4 sections counter clockwise, the overlay indicating the best general direction to go, may change course to clockwise instead.





I figure it would be very subtle overlay, that just finds the blue + and generically points in the direction it thinks the WP and next priority may be, with the target usually being the next zone, however if possible from just the minimap alone it would also point towards tristram, nihl, ancient tunnels, etc.
(Creating multiple layers of overlay so that 2+ priorities can be indicated, for maps where there are multiple encounters/zones was my big concerns in terms of CPU/GPU resources, as I would likely use vscode and free AI tools to create the overlays )

I have never trained a model, but given there is ample data for d2 maps, and known rules to start with to help speed the process up, countless, durance, etc I thought it may be worth it, again, it won't be much better than human intuition, but figure maybe for some maps and scenarios it may actually be helpful, and more reliable than my old failing memory.



However, before I start something that would be too resource intensive to make it worth it, as I only have old i7 and 6700xt , and nearly 10 year old ddr4 ram, I figured I would poke around and seen if anyone with experience with overlays/prediction models had any insight to offer me.