No-Go-Zones and overriding the Pathfinding Algorithm

Better safe than sorry I say...

I disagree, I can think of enough cases where you would curse your units heartily, when they all line up before a bottleneck and none of them find the way around that little hill in the middle of nowhere.
Sure the fine line between situations where splits are useful and where they are harmful is somewhat arbitrary (or at least complex), but I wouldn't say never.

For example, in known friendly territory I would go for the fastest route, splitting and all. In known enemy territory or a generic unscouted region I would go for a tight formation without split and take possible delays as price. But if and how such more global information can be used in an updated fashion is another question. Besides, this distinction reflects just my view and is more a matter of personal taste than universal game mechanic rule.

I agree, if you really never allowed units to split up by themselves this would mostly defeat the whole flowfield idea. And that is why the player needs to have both options. First option is letting the units find the best route on their own, free to split up. Second one is telling them that they must not split up. This could be achieved by said specific move command. Or a certain movement behavior that can be toggled on and off like stances would do too.

Two points:

1) The flow field stuff demoed is just for movement of units, getting from A to B has nothing to do with with how many units you have and weather or not you want them to stick together. That is the job of the formations, which have yet to be implemented. Having a huge debate about this is moot until we see formations, which at that point, maybe, the discussion will be relevant. But asking for a lever, when we don't know all our levers is premature.

2) As someone who has implemented a variation on flow fields, people seem to be missing 2 key points about the tech. First is that this is not actually supposed time optimal, it's supposed to be cost near optimal, the difference being, time can be a component of cost, but not vice versa. And secondly, the behavior heavily depends on how exactly you set up the cost, we didn't see it, but I bet a lot of tuning is going into the flow field to make it work right. In fact, what we saw was a debug tool, probably to help with exactly that.

You don't really understand how pathfinders work.

1: A pathfinder does not *know* if a path (a potential route for units which your *eyes* can identify on the map) is completely blocked. It can *only* look at local context starting from where your units are, square by square.

2: *If* a valid (in terms of where units can freely move) path exists between where your units are and where you told them to go, a (non-broken) pathfinder will find that path even if it does not match your *idea* of what's ideal / valid.

3: Furthermore, no pathfinder (no matter how advanced) can ever know what *your* idea of the ideal / valid path is if all the information you give it is one single waypoint. The *only* way to explain your idea to the system is by handing out more intermediate waypoints.

4: A pathfinder does not and can not know what the most direct (non-blocked valid) path *is* before computing it from local context, only your eyes have that information. Your general rule cannot be applied.