.png){kind=avatar}
PI, your path finding needs some work...
- Thread starter TehJumpingJawa
- Start date
-
We have updated our Community Code of Conduct. Please read through the new rules for the forum that are an integral part of Paradox Interactive’s User Agreement.
Caused by enemy forts, presumably? It'd be shorter to take a "Northern" path, but then detects blocked provinces, so it detours again ... perhaps.
Wish that mechanic would gain some refinement. Some of the effects it has are rather annoying/make no sense.
Wish that mechanic would gain some refinement. Some of the effects it has are rather annoying/make no sense.
- 3
Oh, wow, that's pretty nice actually (the GIF, that is, not the behaviour)
Please tell me you took the time to make a save as well as all those screenshots?
I'd be very interested in it
Could you upload it somewhere and link here?
Please tell me you took the time to make a save as well as all those screenshots?
I'd be very interested in it
Could you upload it somewhere and link here?- 7
- 3
- 1
It's not forts.
If he can walk to Valencia unimpeded, there is no reason forts should suddenly factor into further movements.
- 2
Hmm, not sure I did make a save actually
Though this 2nd gif gives a strong indication as to what's happening behind the scenes.
As Pyoro correctly deduced; the path seems to be chosen, and then modified according to fort ZoC logic.
Solution would be to check for sub-optimal paths once these ZoC modifications have been applied; could be quite computationally expensive for long paths. (caching of sub-paths would be a sensible optimization in such cases)
Though this 2nd gif gives a strong indication as to what's happening behind the scenes.
As Pyoro correctly deduced; the path seems to be chosen, and then modified according to fort ZoC logic.
Solution would be to check for sub-optimal paths once these ZoC modifications have been applied; could be quite computationally expensive for long paths. (caching of sub-paths would be a sensible optimization in such cases)
- 3
The game is punishing you for allowing those arragonese borders to become so ugly.
- 13
- 1
Catalonian, please!
It's only temporary; now that Castile* has been wiped off the map, Iberia will soon be repainted a glorious shade of woodland green.
*(The only way I've found of stopping the virulent yellow plague spreading across new & old world alike)
- 1
Yeah, there seems to be some problems with pathfinding. In my current Ottoman game one army was able to move to adjacent enemy province straight away, while another army in the same province had to take a detour.
That sounds like a fort lock. The armies probably arrived from different places and so were locked in different ways.
I assumed that the units avoid mountains and try to cross grassland / farmland. I think that is what is happening.
1st screenshot shows the last phase pathing going clean over a mountain province, through multiple forts under siege, instead of simply taking a path where stack stays in-borders and suffers less attrition. I'm not sure on the march time between the two, but pathing through enemy forts there rather than simply going Valencia --> target movement is questionable.
If you look at the second screenshot the path is good. In the first screenshot it tried that way but then steered clear of the forts, as Pyoro said. Only solution atm for players is to micro more I guess
There are old, old problems with pathing that far predate the invention of forts. The automatic pathing algorithm simply cannot be relied upon to track the fastest path between two tiles.
One easy example of this is the two open ocean tiles to the west of France, "Western Approaches" and "Bay of Biscay". Going through "Western Approaches" (the outer tile) is one day faster than through Bay of Biscay. But if you set a path around that part of Western Europe, the algorithm will pick one of the two seemingly at random.
That's just one example, but this type of suboptimal pathing happens in every region of the map, on land and at sea, I'd say at least 50% of the time. And it's not just a waste of a single day; many of these bad paths (especially in places like South America, Persia, the steppes) can cost weeks of wasted travel time.
After thousands of hours observing this behavior, I have a hunch that the algorithm does something like orient itself towards its destination like a compass each tile, in order to "guess" which tile should be next on the path, rather than exhaustively determining which sequence of tiles produces the shortest travel time. Even if that's not the exact mechanism, there's clearly some flawed heuristic at work.
Maybe there's performance reasons that the calculations can't be more exhaustive, but even if that's the case, I would really really like to see this improved upon. It's been a constant irritation pretty much as long as I can remember.
One easy example of this is the two open ocean tiles to the west of France, "Western Approaches" and "Bay of Biscay". Going through "Western Approaches" (the outer tile) is one day faster than through Bay of Biscay. But if you set a path around that part of Western Europe, the algorithm will pick one of the two seemingly at random.
That's just one example, but this type of suboptimal pathing happens in every region of the map, on land and at sea, I'd say at least 50% of the time. And it's not just a waste of a single day; many of these bad paths (especially in places like South America, Persia, the steppes) can cost weeks of wasted travel time.
After thousands of hours observing this behavior, I have a hunch that the algorithm does something like orient itself towards its destination like a compass each tile, in order to "guess" which tile should be next on the path, rather than exhaustively determining which sequence of tiles produces the shortest travel time. Even if that's not the exact mechanism, there's clearly some flawed heuristic at work.
Maybe there's performance reasons that the calculations can't be more exhaustive, but even if that's the case, I would really really like to see this improved upon. It's been a constant irritation pretty much as long as I can remember.
Last edited:
- 2
- 1
I do think pathing doesn't take into account whether you've occupied the forts or not, until maybe the next month? Something fishy is going on there.
I always assumed it isn't intended to always track the fastest path. The general impression I get is that it tries to find a fast path which is also a low attrition path; I guess it probably also has some kind of threshold for saying "this path is good enough, stop looking". One of the notable quirks it has is that if your origin and destination are settled provinces, then it won't route through an uncolonized province unless it absolutely has to - which is reasonable, even though it's annoying, because uncolonized provinces have very low supply limits.
- 1
Yeah, here's a perfect example: (Photoshopped to make the paths more obvious)
I'm at peace in this example, so fort ZoC should play no part.
Same start node, and the plotted paths even cross twice, yet the 2 routes are quite different, and likely have very different journey costs.
My guess would be the heuristic function of the A* search must be overestimating, thus permitting sub-optimal results.
Last edited:
- 1
Yes, it definitely does avoid uncolonized provinces—but I'm not convinced supply limit has anything to do with it. I've seen cases where it seemed like that was what was happening, but I've also seen cases under similar conditions where it pretty clearly wasn't. It's a bit of a mystery.