Hacker News new | past | comments | ask | show | jobs | submit login

SC1 pathing is mainly plagued by how it handles collisions/blocking.. and the poor choices they make to resolve in the face of temporary blockages (eg units) — their resolution strategy is basically wait a little, and route again. If it’s still blocked (which could be due to a different unit taking the blocking position after the original moved) it basically assumes it’s been blocked by a building or a unit in hold position, and finds any arbitrary new route to reach. And they won’t ever try to replan the path unless blocked again, so in a moving army the shortest path could be utterly wild in the exact turn they actually tried, and that wild path will simply be the plan until blocked/vcompleted.

Dragoons are particularly absurd, because their collision box changes as they walk, and they’re huge, so they block one turn and unblock the next, and block again, triggering the “full reroute” with even a trivial number of them despite congestion appearing pretty light.

I think this dumb collision resolution strategy would have been mostly fine if they had handled the ABA problem and waited longer in that scenario (or planned to replan for dramatic differences in route lengths after collision-rerouting)




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: