Print

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

[Zanzetkuken The Great]

I like the recent gameplay additions, but the AI is still horrendous. Even the EU4 + CK2 AIs are better, and I consider those extremely weak as well. Paradox should hire a decent AI coder.

This problem has nothing to do with the competence of the AI coder. These games run in real time with tens of AI operating at once. Even hundreds of AI at times. A good AI would require tons of calculations and decisions which would make the game play at a crawl. Paradox has made the trade-off of having a mediocre but efficient AI in its games and it's probably the best choice they could make in these constraints. Better to have some people bad-mouthing the AI instead of nobody playing because a tick takes forever to process.

Lol.

If only there was a game with decent AI that could operate thousands if not hundreds of thousands of simultaneous AI calculations at the same time per tick without compromising performance too much.

Yeah. I wonder what game that is. It would be really cool if it was programmed by basically one guy too...

Hmm... I wonder if there's any game like that. That could show people that somehow the god-tier programmers at Paradox have managed to strike a beautiful balance between game performance and AI calculations.

Yes, if only there was some living proof out there that there is literally no alternative to improve their AI (especially not by some half-assed modder by all means) without sacrificing performance a single iota.

Hell, if only there was some other Paradox game that ran AI calculations on the individual person level with literally tens of thousands of AI characters working in real time and still had a game that was very playable. Too bad no other such games exist in their catalogue. Hell, maybe not even that. Maybe some Paradox games that even ran a few hundred countries at the same time could prove a good example too.

Sigh, we just have to accept games have to suck ass on the AI side of things because we run caveman computers from 60,000 BC, doing their calculations based off the evaporation rate of amber from the sap of trees.

It should be noted that just making an AI is, while difficult, comparitively easy to making an AI that weighs every decision properly and further is fun to play against that is not made too weak that a player can beat with ease, or way too good and breaks literally everyone.

And seriously man, while I have only a few hundred in this game, the AI can still kick my ass if I'm not paying attention and I tend towards the middle of the AI difficulty levels, without mods to boost it.

And are you seriously comparing this game that has undergone notable major gameplay shifts to ones that have been in generally the same overall gameplay just with some tweaks for a half-decade or longer more production time?

[umiman]

I like the recent gameplay additions, but the AI is still horrendous. Even the EU4 + CK2 AIs are better, and I consider those extremely weak as well. Paradox should hire a decent AI coder.

This problem has nothing to do with the competence of the AI coder. These games run in real time with tens of AI operating at once. Even hundreds of AI at times. A good AI would require tons of calculations and decisions which would make the game play at a crawl. Paradox has made the trade-off of having a mediocre but efficient AI in its games and it's probably the best choice they could make in these constraints. Better to have some people bad-mouthing the AI instead of nobody playing because a tick takes forever to process.

Lol.

If only there was a game with decent AI that could operate thousands if not hundreds of thousands of simultaneous AI calculations at the same time per tick without compromising performance too much.

Yeah. I wonder what game that is. It would be really cool if it was programmed by basically one guy too...

Hmm... I wonder if there's any game like that. That could show people that somehow the god-tier programmers at Paradox have managed to strike a beautiful balance between game performance and AI calculations.

Yes, if only there was some living proof out there that there is literally no alternative to improve their AI (especially not by some half-assed modder by all means) without sacrificing performance a single iota.

Hell, if only there was some other Paradox game that ran AI calculations on the individual person level with literally tens of thousands of AI characters working in real time and still had a game that was very playable. Too bad no other such games exist in their catalogue. Hell, maybe not even that. Maybe some Paradox games that even ran a few hundred countries at the same time could prove a good example too.

Sigh, we just have to accept games have to suck ass on the AI side of things because we run caveman computers from 60,000 BC, doing their calculations based off the evaporation rate of amber from the sap of trees.

It should be noted that just making an AI is, while difficult, comparitively easy to making an AI that weighs every decision properly and further is fun to play against that is not made too weak that a player can beat with ease, or way too good and breaks literally everyone.

And seriously man, while I have only a few hundred in this game, the AI can still kick my ass if I'm not paying attention and I tend towards the middle of the AI difficulty levels, without mods to boost it.

And are you seriously comparing this game that has undergone notable major gameplay shifts to ones that have been in generally the same overall gameplay just with some tweaks for a half-decade or longer more production time?

You and EnigmaticHat missing the point of the post.

USEC said that the reason the AI sucks is because the devs gimped it on purpose to save fps. Lol.

I said all that to demonstrate that there are plenty of Paradox games and other games that do way more calculations per tick and it doesn't affect fps. Or at least so significantly it affects playability.

I didn't say anything about whether the AI is good or not, though everyone here knows my thoughts on that.

[USEC_OFFICER]

I'm not saying that Stellaris's AI is the best it could possibly be. I'm saying that good AI requires trade-offs. Trade-offs that Stellaris's AI coder would love to make but can't because the game would be an unbearable slog if they did. The dude isn't incompetent or lazy. Just under difficult constraints and trying to handle multiple mechanic changes on top of that.

Stellaris is a different game than EU4, CK2, Victoria 2 and all of the other Paradox games. You can't import the same AI and get the same results, especially with the much larger emphasis on expansion and resource management that Stellaris has. The AI in CK2 just has to wake up from time to time, check its options, and pick which one it likes best. The AI in EU4 just has to keep itself from imploding while gobbling up provinces and the mechanics ensure that it'll be a reasonable opponent. The AI in Stellaris has to build an Empire from scratch and keep pace with a human player doing the same thing. If CK2 and EU4 have more and better AI, it's probably because less is demanded of the AI in those games. Not to mention their longer and more stable development.

_{Also, being pendantic, if those games ran more calculations per tick then they'd be running slower too. Assuming that the bottlenecks are the same or similar, a CPU can only handle so many calculations per second. No amount of programming voodoo will squeeze more calculations out of it. To run faster all you can do is reduce the number of calculations done per tick, which requires tons of optimization or a dumber AI.}

[LoSboccacc]

4x AI has been around since 4x were a thing, they're not breaking any new ground.

besides, the strategic level only need to make high level decision every couple minutes or so, and it doesn't need to be run in real time. i.e. allocate these resource to reserve, these to attack, these to patrol, decide a long term build list for each planet and for each fleet - all that could be done every now and then or on demand (i.e. post diplomacy action)

[Kanil]

I'm not saying that Stellaris's AI is the best it could possibly be.

I'm not asking for an AI that can compete without cheating, I get that that's impossible -- I'm asking for an AI that doesn't spam precinct houses, which doesn't require any processing power to fix, just some changes to a text file. There's some seriously low hanging fruit that Paradox hasn't plucked yet, and it's super fair to call them out for it.

The AI's basically been broken for the past 3.5 months, and "the AI doesn't fight wars, because it constantly patrols for pirates" is an infinitely less playable game than some hypothetical version where the game runs slower so the AI doesn't actually do that... (and of course, if the fix is just adjusting weightings, then that probably comes at no additional processing cost.)

[USEC_OFFICER]

EDIT: I cannot read.

[gimli]

I'm not saying that Stellaris's AI is the best it could possibly be. I'm saying that good AI requires trade-offs. Trade-offs that Stellaris's AI coder would love to make but can't because the game would be an unbearable slog if they did. The dude isn't incompetent or lazy. Just under difficult constraints and trying to handle multiple mechanic changes on top of that.

Stellaris is a different game than EU4, CK2, Victoria 2 and all of the other Paradox games. You can't import the same AI and get the same results, especially with the much larger emphasis on expansion and resource management that Stellaris has. The AI in CK2 just has to wake up from time to time, check its options, and pick which one it likes best. The AI in EU4 just has to keep itself from imploding while gobbling up provinces and the mechanics ensure that it'll be a reasonable opponent. The AI in Stellaris has to build an Empire from scratch and keep pace with a human player doing the same thing. If CK2 and EU4 have more and better AI, it's probably because less is demanded of the AI in those games. Not to mention their longer and more stable development.

_{Also, being pendantic, if those games ran more calculations per tick then they'd be running slower too. Assuming that the bottlenecks are the same or similar, a CPU can only handle so many calculations per second. No amount of programming voodoo will squeeze more calculations out of it. To run faster all you can do is reduce the number of calculations done per tick, which requires tons of optimization or a dumber AI.}

So you're saying that the coders are forced to [totally] dumb down the AI, because otherwise the game would be totally unplayable? [Extremely slow & laggy] So this is why the Stellaris AI is not even mediocre, but simply abysmal? Well then, perhaps they should disable singleplayer until strong enough processors will hit the market in ~2045. [Sorry for the sarcasm, but I am pretty sure that better coders could create a much better AI for the game. ]

[Criptfeind]

The dude isn't incompetent or lazy.

This seems like a pretty big assumption about the things we don't know about the stellaris AI given that the stuff we do know about most of stellaris game design and AI performance is full of lazy and incompetent mistakes.

Edit: Although to go outside a single snarky sentence for a moment. It might not be the fault of whoever is actually making the AI, it could just be that they get inadequate resources and feedback, but wherever the issue ultimately lies it seems clear that the stellaris dev teams approach as a whole to most of the game, including the AI, is lazy and incompetent.

[Telgin]

I do believe that their dev team is put under unreasonable time pressure with the DLC release schedules (the extended support period for 2.2 was only 6 patches...), but it's absolutely clear either way that the game is not tested properly.  The 2.2.6 bug with the colony unrest events should have been caught within 20 minutes of playtesting, which tells me that when releasing bugfix patches they must not actually play through the game one last time before packaging the patch.

Anyway, on the subject of Glavius's AI mod: I'm sure there's at least some that Paradox could take from it, but from what I understand the mod has tons of hacky fixes in it for the AI.  I've heard it makes some buildings indestructible to prevent the AI from ruining them by resettling pops, for example.  If I were in Paradox's position, I wouldn't want to use fixes like that either and would prefer to fix it at the fundamental level, even if that's going to be a lot harder.  I guess the devs are never going to have time to do that though, since I'm sure they're already working on the next DLC.

[Mephisto]

I now find myself in possession of the base Stellaris game. Would you all suggest diving in or are any of the DLC absolute must-haves for a newbie?

Also, I stumbled across The Spiffing Brit last night. Enjoy!

[ZeroGravitas]

I now find myself in possession of the base Stellaris game. Would you all suggest diving in or are any of the DLC absolute must-haves for a newbie?

this question has been asked a few times before; scroll up

[Mephisto]

scroll up

Thanks for that. No references on this page so "scroll up" is particularly useless advice. And nothing is on the last ten pages of this 552-page thread so I thought I'd ask.

[Telgin]

Utopia is probably the most important one since it enables most mega structures some ascension perks and I believe hive minds.  After that I'd say Apocalypse for titans, colossi, unity ambitions and probably a few things I'm forgetting, then MegaCorp for the new mega structures, ecumenopoleis and new ascension perks.  Synthetic Dawn and Distant Stars would be low on the list for me.  Leviathans I'm not sure about.  It probably adds more than space monsters, but that's all I can remember it adding, and they're honestly not that important.

[Il Palazzo]

I'd say you can leave the story packs for later, as they can freshen up the game a bit once you get bored with the core content, but there's enough of it even in the vanilla to last you at least a few dozen hours.
I think I'd get only Utopia for starters.

[Persus13]

Utopia is probably the most important one since it enables most mega structures some ascension perks and I believe hive minds.  After that I'd say Apocalypse for titans, colossi, unity ambitions and probably a few things I'm forgetting, then MegaCorp for the new mega structures, ecumenopoleis and new ascension perks.  Synthetic Dawn and Distant Stars would be low on the list for me.  Leviathans I'm not sure about.  It probably adds more than space monsters, but that's all I can remember it adding, and they're honestly not that important.

Leviathans adds the neutral space stations that trade you stuff, although I haven't played since the big 2.2 update so I have no idea how those work now. It mainly gives you more stuff to explore and makes the early game and exploring a bit more interesting from what I recall.