Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Rolan7]

I think it's Gateway that trivializes one of strongest drawbacks of wide Empire - its size itself.

I always play with x0 gates because I appreciate how difficult it is to cover entire territory with fleets, especially with player's tendency to clump entire fleets together. It makes sense as well; galaxy-spanning nation would have difficulty managing its large territory with fleets taking years to travel, making them more vulnerable to multi-front attacks and rebellions from backwaters. Instead one can simply spam Gateway and eliminate distance aspect of gameplay entirely, which benefits wide empiree much more than tall compact ones.

Gateways certainly change the game fundamentally once you get them, and I tend to deploy them... generously.  That led to some mild friction (more like confusion) when I was playing multiplayer with some friends for a few weeks.  I was gating' up my entire megacorp, naturally, to *eliminate* piracy... naturally building one in my capital, and my megashipyard, and all sorts of strategic places.  Meaning that every war was mostly on just one front - the entire gateway network.  Any incursion into any gateway system was both extremely dangerous and extremely easy to react to.

That worked out pretty well, but it was a surprising doctrine to my fellow players who had to adjust a bit.

Maybe that's part of why I've always been confounded at the idea of the crisis lasting more than a decade or two.  It's always seemed so all-or-nothing:  They easily reach a gateway and then the matter is settled in a few years.  Either I destroy them or they destroy me, there's no detente.

Anyway, I think gateways are pretty good *because* they change the game so much.  They're one of the few meaningful ways that the lategame is different from the midgame.  The only "megastructure" that even comes close to their importance are habitats, the Coordination Center, and sometimes the Interstellar Assembly.

[Telgin]

That's generally how I build gateways too.  Not sure if it's a good or bad idea, but I usually have a single main shipyard system one jump from my home system, which has a mega shipyard and 6 shipyard spaceport.  I stick a gateway in that system, then one about a jump away from each border station, and if my empire is particularly massive I'll sprinkle a few gateways in pockets.  That largely eliminates piracy because of the overlapping trade protection and short distances back to my home system, while allowing rapid deployment and reinforcement of ships to frontlines.

I've always wondered if it was better or worse to build the gateways in the border systems with bastions.  If I lose that system then I lose the gateway, but if I lose a bastion system I'm probably losing whatever is behind it anyway I guess.  Doing it a jump away lets me share a gateway to several bastions sometimes though.

I do like that gateways change things up in the late game and generally make the game more convenient.  Jump drives help a lot too, but are more situational.  However, gateways do also mitigate the biggest drawback to big empires, as ventuswings pointed out.  I have a bit of a love-hate relationship with that.  I like that fleet positioning matters... until it inconveniences me, at which point I hate the fact that it takes so long to move ships around.

Regarding the crisis: this is one reason I almost prefer to not have L-gates in my empire.  If the crisis ever reaches the L-cluster you just cannot keep these systems.  They'll pop in, zap the station in 2 seconds, convert it to one of theirs, and pop out.  Well, this and the fact that the AI will unleash the Grey Tempest when nobody in the galaxy has even 10k fleet power.  I've had crises last 100+ years, but those were cases where I had a hard time reaching them due to idiot neighbors who wouldn't open up borders to me so I could fight the crisis.  I've never, ever, seen one of the mechanics trigger where the crisis has spiraled out of control and other factions show up to help / make it worse.

[Dunamisdeos]

I build gateways in:

My home system
My shipyard system(s)
My chokepoint systems, including vassals. If I intend to vassalize/tributary I always demand a convenient system to build a gateway in. Gets me around the galaxy faster and also good luck to those fringe provinces if you want to break free, I can get my fleet there whenever I please.

[Duuvian]

I do that plus for trade hubs. No more pirate patrols once you have gateways.

[Rolan7]

I do that plus for trade hubs. No more pirate patrols once you have gateways.

Very much the same.

I sometimes wonder whether I'm actually playing optimally, or whether I'm only playing to *feel* like I'm playing optimally.  I find strategy like playing tall Megacorp on small galaxy, avoiding piracy whenever possible and then doing whatever I want once I have gateway tech - and I feel like I'm beating the system.

Then I read about people facing up against 25X crises and realize... oh... I'm just participating in a power fantasy.
And it feels great!  I'll load up another game on Captain - which I thought was fair, but benefits the AI - and suffer a few scrapes before making the galaxy my federated, xenophilic oyster <3

And that is why I'm fine with the AI being intentionally unoptimized.  Particularly when it means the AI makes more "emotional" choices according to their empire type.

[Criptfeind]

I don't think that the AI is intentionally unoptimized. It's understandably quite difficult to make an AI that both good and feels good to play against. But there's a pretty vast gulf between where stellaris AI is now and where frankly it reasonably easily could be. You can forgive and understand a not perfect AI, but stellaris AI is I think the worst AI I've ever seen in a game where the AI is at least nominally working, and some patches you don't need that last caveat.

[Damiac]

I don't think that the AI is intentionally unoptimized. It's understandably quite difficult to make an AI that both good and feels good to play against. But there's a pretty vast gulf between where stellaris AI is now and where frankly it reasonably easily could be. You can forgive and understand a not perfect AI, but stellaris AI is I think the worst AI I've ever seen in a game where the AI is at least nominally working, and some patches you don't need that last caveat.

It's hard making an AI for a game where the rules are known and unchanging. It's impossible to make a good ai for a game that is still good after you change all the rules.

Stellaris has been at least 3 different games over the time I played it, original 3 FTL modes, that weird second version with the strange war exhaustion stuff, and then the version after that where outposts completely changed, not to mention all the DLC stuff.

I understand it's changed more since then, which only makes the problem worse. And why spend money making the AI better, which you have to give in a free update, when instead you can make a new DLC and sell it.

[Telgin]

I've complained about the AI as much as anyone else since there really is no excuse for Paradox as a company to release DLC or updates with the AI in the state it's in, considering it's plainly visible to the developers and testers if they play the game for 50+ years, but I feel bad for the developers for those reasons.  They get beat up about it when I know it's the company direction and pacing that prevents them from actually fixing the AI and keeping it working.  It's miserable enough to iterate on a thousand weights and conditions to make it work decently without having to throw it all away and start over on every new patch.

Then I read about people facing up against 25X crises and realize... oh... I'm just participating in a power fantasy.
And it feels great!  I'll load up another game on Captain - which I thought was fair, but benefits the AI - and suffer a few scrapes before making the galaxy my federated, xenophilic oyster <3

This is largely what Stellaris has become to me over the years, where I just like to see how big I can make my numbers go, but I think I more or less hit my peak in my last game by beating the 10x Contingency.  That required some very atypical game settings, and trying to push it to a 25x Contingency just doesn't sound fun.

I'm debating in my next game if I should dial it back down to a 1x crisis and see what happens if I don't fight it.  Two things I've considered:

1. See if the galaxy can deal with the crisis without me for a change.  An awakened empire might be able to, but I doubt any normal empire could in this patch.
2. Build up enough defensive stations that I'm immune to the crisis, and let the crisis have fun.  Maybe I'd get to see the extra factions that pop up when the crisis spirals out of control for a change.

[MorleyDev]

I mean, there's mods like StarNet AI and StarTech AI that make the AI work pretty well, so it's not like Paradox couldn't easily take those mods as a basis and make the AI immediately human-competitive.

I think part of the issue and reason they can't just take what the mod does or contract out the mod developers is that mods like Starnet AI balance the AI so that it's competitive with humans who have played the game before even on the lowest difficulty that doesn't give the player buffs (player buffs risk teaching players bad habits, hence restricting them to the very lowest difficulty).

From a game design balance pov, the perspective they're taken for AI in Stellaris is balancing it to keep the game appealing to new players, so they'd ideally want an AI that on lower settings a new player can survive against, whilst is competitive on higher difficulties with more experienced players. Otherwise you risk players getting steamrolled when they don't know what they're doing and giving up rather than learning (think most people who try Dwarf Fortress).

The 'best' way to do that is completely AI behaviours for different difficulties, like Starnet-style being used for higher ones and 'current vanilla' for lower ones. That's a testing and maintenance nightmare though, which is why most games go the buff/debuff route for AI difficulties.

So it's definitely more complex than just "make AI good", because you can't make AI too good to always beat new players whilst also making AI good enough against older players. Hence the hack of giving a crapper AI buffs games use.

----

As for Planet/Sector AI, something they've expressed interest in before to reduce micromanagement is to let you create a 'template' for the different planet types it can follow. So you tell it what a Agri-World should look like, and it'll build each one up to look like that. The issue is just getting the time to design and test that, since it'd to be a UI/System able to take into account all the different sizes and modifiers a planet could have in some way.

[Nelia Hawk]

all this AI talk reminded me that years ago i did read dev diaries of "galactic civ 2" and that they often talked about how advanced their ai is.

also reminds me of "AI war 2" with the the ai getting more and more difficult and has more things and tactics with higher difficulty nmbers.

[Telgin]

Balancing the AI against new players is certainly part of it, and I'm sure that's one reason they don't use something like Starnet as a template.

I don't think it's the whole picture though, or necessarily even a large part of it.  Someone found that the Ai is currently explicitly configured to not build research labs for the first 50 years, for example.  That was probably introduced as a bandaid to improve economic stability until they could figure out why the AI was tanking consumer goods or mineral production or something.  It somewhat fixes that issue and does seem to have helped with the stability spirals of 3.0.1, but at the cost of making the AI utterly non-competitive past the first 50 years.

I agree that it would be really nice if the AI was smart enough to use advanced tactics, but held those in reserve at lower difficulties.  Maybe on lower difficulties it does dumb things like suicide small fleets against bigger ones, but on higher difficulties it knows to stack them up, or to use special station builds in systems that nullify shields, or camp hyperlanes exits with corvette swarms to counter battleships, etc.

It would be great if they could make it so it could compete without cheating, though I know that's a lofty goal they probably will never achieve.  Funnily enough, I played Stellaris on baby mode the first few times because I was used to games like this cheating like crazy even on lower difficulties, and back in 2.0 it was able to mostly keep up with me on my first couple of games even then.  I wonder if new players have that experience in 3.0, or if the AI has gotten objectively worse since then.

[Damiac]

It would be great if they could make it so it could compete without cheating, though I know that's a lofty goal they probably will never achieve.

The only problem is they literally said the AI didn't cheat and that it was just a well programmed AI.  Well not the only problem, but that's a big one.

[MorleyDev]

Often when people talk about AI cheating they mean it can operate with information the player doesn't have. Take the classic RTS Warcraft, where the AI didn't have fog-of-war and always knew exactly where your units are. This is what is usually meant in game design when talking about a 'cheating AI', also called the "All-Seeing AI".

I think the Stellaris AI in that sense doesn't cheat. It is making decisions with the same kinds of information a player has about other empires etc (Not sure if that changed post-espionage, or if they also need intel to know relative fleet strength without Intel), and is pulling from a deck of science in the same way.

But it is 'cheating' in the sense that difficulty levels give it fixed percentage 'buffs' to production the player doesn't have on higher difficulties.

---

The talk about Admin Cap changes is interesting. They want to make it a resource to be managed, rather than a fixed penalty/limit you work around, so that it can:
a) Be technically simpler to work with (since it can use the same resource system as everything else).
b) Be usable by future features.

Admin cap as-is is currently a bit of a dead end, nowhere to go with it for future stuff, just adds extra technical burden to consider. Making it a resource empires consume as they grow would make it much easier for other systems to access.

They've previously spoken about wanting to add Institutions and Religion. I could see a system where Institutions use whatever resource replaces admin cap to produce resources and buffs, whilst unity gets used by religion. Materialists would get powerful institutions but not have access to or much weaker religion, whilst Spiritualists institutions are weaker but they get more from Religion.

[Telgin]

I'm fairly sure the AI does mostly play by the same rules as humans, including espionage limiting knowledge of fleet power.  I say that because since 3.0 I've had several empires declare war on me early despite me having superior fleet strength that normally would have scared them off in earlier patches, which was only discovered when they didn't ever send any ships to attack me and I probed their systems with my fleets.  Or maybe that's just bugged, since the AI declaring war and never attacking is a bug that's been around for a while.

Yeah, the AI isn't cheating like, say, Star Trek: Birth of the Federation, where I'm pretty sure the AI gets free ships and such.  Either that or it gets massive economic boosts even at low difficulties.

[Criptfeind]

I'm not totally sure, and afaik Paradox hasn't outright come and talked about it (and, given their past history with outright lying about the AI, even if they did make a statement on it, that wouldn't exactly be definitive ). But I'm pretty sure the Stellaris AI is cheating in the way you're talking about as well MorleyDev. I've not really been able to find any confirmation on that for stellaris, but in other paradox games the AI does see though the fog of war and I'd find it odd that stellaris AI has a particular apparently hard to make work properly feature that their other AI doesn't have when the stellaris AI is so outstandingly poorly made in other respects. There's certainly a question of how much it's programmed to take things into account which for that sort of cheating is more important then it's ability to cheat.