In game design terms it's interesting how old games pathfinding is so much better than modern games.
If pathfinding is too efficient, and the same for all units, then the micro aspect of the game is killed. Micro is the "real time" part of real time strategy.
If you play something like Starcraft 1, Warcraft 2, or AoE 2 you'll see the pathfinding has immense peculiarities that as you get better at the game you learn to manage. Units get stuck on each other and so on.
Starcraft 2 has a very good pathfinding algorithm but it still intentionally adds a lot of variety to how units move to recreate that cumbersome peculiar pathfinding that rewards micro. Yet if you look at Starcraft 1 you see the units still move in much more unique ways.
I think it's too easy to fall into a technical rabbit hole and try to do the best pathfinding when it actively harms the game.
Interesting. Could you give a bit more of a concrete, specific example of "good pathfinding hurting the experience"? It's an odd sounding idea, to me.
Also, you seem to be deep in this world, I'll also ask you - are there RTS games out there where the specific terrain is crucially important for various unit types - like we've seen in the real world recently, with weather updates waiting for deep mud to dry out being the main factor in when tanks and other sorts of vehicles can be useful? My experience of RTS games is limited, but I remember maps being very flat overall, with very few types of steepness modelled.
It's a well known phenomenon in StarCraft 2 that the super efficient pathfinding increased lethality and reduced defender's advantage compared to the first game. And the game is overall considered to be VERY high lethality with a very weak defender's advantage.
Small fights don't make the difference as noticeable, but larger armies are so much more efficient in SC2 compared to SC1, that it's harder to hold off a larger force with a smaller but better-controlled force. The bigger "deathball" tends to just win, it's harder for someone to come back from an army disadvantage with skillful play.
Another small example there is that "ling runbys" in SC2 are vastly more punishing for even small mistakes in leaving a gap open in a building wall, because a huge number of lings can run through a small gap extremely quickly.
If pathing efficiency is the goal, why not make every unit in an RTS extremely tiny? That would make it more efficient for sure. Or, hell, just turn off unit collision entirely. Or make units all move ultra fast, or get rid of all map choke points? All of these things would improve how efficient pathing is.
Pathing efficiency isn't the goal itself, it's part of the game designer's toolbox. Plenty of things are intentionally pathing-inefficient -- like big, slow units -- as part of the game's design and balance.
The lead designer of Stormgate, which is the closest thing we're gonna have to StarCraft 3 probably, has talked about SC2's pathing efficiency problem himself. Granted, it doesn't sound like he wants StarCraft 1-style pathing, he just wants to compensate for the efficiency in other ways, like maybe making unit hitboxes bigger.
Great insights. The gaming experience and pathfinding relationship can also be expanded to game AI in general. I found Lars Liden's slides specifically mentioning "intelligence! = fun". So, for a better gaming experience, it is better to dumb down the AI.
Well, it really depends on the game and user. Personally I'd love smarter AI in some of my games, for some enemies (I realize that I'd probably hate smart AI in other games).
Like, StarCraft skirmish AI is pretty basic, having a smarter opponent to practice against would be nice. As it stands, I can practice a build vs the AI just to get the timings right, but otherwise the AI is mostly useless because it sucks too much compared to a regular player.
If you look at basic zergling vs zealot combat in SC1 vs SC2 you'll see what I'm saying. In SC2 they almost move like liquid and they don't form accidental choke points. The fight is over much faster and simply rewards the beefier army.
For example in Warcraft 3, which is much smaller scale with 10 - 20 units per player at any time, players constantly look for ways to block each other. While the fight is happening you might sneak a unit around just to block their path when they escape. This wouldn't be possible if the pathfinding was better at steering.
In terms of terrain maybe land/sea/air in Red Alert 2?
The more tactical RTS like Company of Heroes and Dawn of War 2 use the terrain a lot, but in small scale ways like moving from cover to cover and building to building. The terrain is very complex there.
The new Dune Spice Wars only has desert but it has some interesting ideas about movement. It's a much slower RTS and has been described as real time 4X.
Spring-Recoil engine games like Zero-K and BAR have not only various unit types affected differently by slopes and water depth, some maps even has terrain like ice that makes some units faster on it ! (Also Zero-K has terraforming.)
You are only talking here about a very specific sub-genre of RTS. BAR and Zero-K have quite good pathfinding, and (especially Zero-K) advanced unit AI, and also much more powerful user commands available, yet they still manage to have unit micromanagement matter a lot.
If pathfinding is too efficient, and the same for all units, then the micro aspect of the game is killed. Micro is the "real time" part of real time strategy.
If you play something like Starcraft 1, Warcraft 2, or AoE 2 you'll see the pathfinding has immense peculiarities that as you get better at the game you learn to manage. Units get stuck on each other and so on.
Starcraft 2 has a very good pathfinding algorithm but it still intentionally adds a lot of variety to how units move to recreate that cumbersome peculiar pathfinding that rewards micro. Yet if you look at Starcraft 1 you see the units still move in much more unique ways.
I think it's too easy to fall into a technical rabbit hole and try to do the best pathfinding when it actively harms the game.