|
GunnerJ posted:The AI rework update: Shifting to a scoring based scheme is nice and sane but its still kind of baffling that they thought a weighted list script thing was ever a sane idea, since goal based scoring evaluations have been the conventional thing for, well, I think before there were even video games so lol. What I'm curious about now is if its capable of doing things like minimax ie "realizing" that specializing its planets can produce better gains in the long run as opposed to just building things haphazardly as needed, or looking ahead to realize it'll want x amount of resources in y turns, or even being able to account for risk/uncertainty (should I try to spread more or will that make me vulnerable or will I get boxed in if I go to slow etc). I'm guessing that's just going to be handled on the plan scripting side since evaluations for those kinds of problems can get kind of big and complex if you leave them too open ended, but theres enough cases where an empire has to account for short term / long term, especially when it crosses over into the military side of things (when is my best time to attack vs when is my enemies best time to attack and is it worth reducing my economic output to defend against a potential invasion (and what is the enemies' optimal course of action and is it feasible to counter it)). You can actually get the correct answer to a lot of those problems if you script things out well enough but, its slow, but I could see a potential between switching between various plans based on what some higher level evaluation decides prudent targets should be. Its just that, well, these questions are also ones that have to be answered to get out of the early game in a good position and if the AI can't think about them at all, even with a scoring system, its going to probably be still in a weak position after early game vs a player. Tiler Kiwi fucked around with this message at 10:21 on Mar 6, 2020 |
#
?
Mar 6, 2020 10:18
|
|
|
|
| # ? Apr 26, 2024 16:55 |
|
|
To be clear, the AI is still going to cheat like hell on everything but easiest difficulty
|
|
#
?
Mar 6, 2020 10:35
|
|
|
ShadowHawk posted:In the dev diary linked above the AI very much cares. ShadowHawk posted:To be clear, the AI is still going to cheat like hell on everything but easiest difficulty ...And now I realize you're the same person, so uh, I'm glad we agree?!
|
|
#
?
Mar 6, 2020 10:51
|
|
|
It'd be cool if the AI specializes planets, so that if I'm looking at their planets and find an 18 mining district planet with a mineral refinery then that's an actual prize I can aim for. Instead of now where all the AI's planets are weirdly thrown together messes that somehow have a deficit in everything.
|
|
#
?
Mar 6, 2020 11:09
|
|
|
dev diary posted:Moreover the build plan also tries to take into account on which planet a building would produce the most and build that building there, making the AI more likely to create specialized resource-production planets like “foundry worlds”. The dev diary says they will start specialising worlds better. But I will believe it when I see it.
|
|
#
?
Mar 6, 2020 11:22
|
|
|
ShadowHawk posted:To be clear, the AI is still going to cheat like hell on everything but easiest difficulty e: Demiurge4 posted:It'd be cool if the AI specializes planets, so that if I'm looking at their planets and find an 18 mining district planet with a mineral refinery then that's an actual prize I can aim for. Instead of now where all the AI's planets are weirdly thrown together messes that somehow have a deficit in everything. Splicer fucked around with this message at 11:33 on Mar 6, 2020 |
|
#
?
Mar 6, 2020 11:27
|
|
|
Aethernet posted:The dev diary says they will start specialising worlds better. But I will believe it when I see it. I think its possible; it shouldn't even really require forward thinking, since the scoring is going to naturally keep picking a mineral world for building mines on, and then decide that a mineral boosting building would achieve the highest score on that planet, which then would make it more likely to put mineral producing things on it, etc. If it can track opportunity cost, like, spending a building slot on a 200% mineral world on a not-mineral thing being evaluated worse than putting it on some boring rear end vanilla world, then that'd work. It gets complicated if you try to work in poo poo like population, military preparedness, and tech priorities, but that's what the plan scripts are for I guess. Hopefully the scoring stuff is flexible enough to allow people to create good heuristics for the AI since it should allow that kind of emergent flexibility and responsiveness.
|
|
#
?
Mar 6, 2020 12:42
|
|
|
Yeah from that design it doesn't look like the AI will be capable of: 1) Purposefully moving pops around to planets where they're more efficient (only natural growth), except maybe a heuristic to migrate off overfull worlds 2) Considering or even altering what pop type will use a building after it's built I'm also somewhat worried it's going to staff all the clerk jobs it can rather than ignore them given that he said "free jobs" was an indicator not to build more
|
|
#
?
Mar 6, 2020 13:17
|
|
|
Planet specialisation is weirdly balanced. Some of the specialisations a planet can have are way better than others. Like, take "Generator World" and "Urban World". Generator: +25% Generator District Build Speed +20% Technicians output One bonus gets your generators up and running quicker, the other makes them better. A good specialisation that you'll set a planet to fairly on in its life, and then leave like that forever. Urban: +25% City District Build Speed -10% City District Cost Two bonuses while you're building city districts, but literally nothing once they're built. The only use for this specialisation is temporary, during the building-up phase. ----- It'd be good if the specialisations were rebalanced so there weren't any that have only really temporary uses - Urban World and Fortress World are the main offenders, but stuff like factory, forge and research worlds are also a bit weird since they give bonuses to upkeep rather than output.
|
|
#
?
Mar 6, 2020 13:22
|
|
|
You'll have to pry my Urban specialization from my cold, dead Ecumenopoli.
|
|
#
?
Mar 6, 2020 13:29
|
|
|
Would be nice if it gave like 10% extra housing as well or something though. Would be in keeping with the theme of, "If you specialise a planet for something you get more of that thing", which seems logical to me.
|
|
#
?
Mar 6, 2020 13:32
|
|
|
ShadowHawk posted:Yeah from that design it doesn't look like the AI will be capable of: e: oh you were specifically referring to specialisation of planets. Yeah I get your concern but I still think that's a step beyond what's needed. 2) this should be handled by the pop shuffling mechanic, and they described how they reworked that in the last diary. Gort posted:Would be nice if it gave like 10% extra housing as well or something though. Would be in keeping with the theme of, "If you specialise a planet for something you get more of that thing", which seems logical to me. Splicer fucked around with this message at 13:42 on Mar 6, 2020 |
|
#
?
Mar 6, 2020 13:38
|
|
|
Ogdred Weary posted:Preview embargo has been lifted! So soon! That guy sounds so awful I couldn't keep watching it after 2 minutes.
|
|
#
?
Mar 6, 2020 14:15
|
|
|
Splicer posted:2) this should be handled by the pop shuffling mechanic, and they described how they reworked that in the last diary. This is probably not a terribly significant effect since they'll be counted once they're already working though
|
|
#
?
Mar 6, 2020 14:20
|
|
|
ShadowHawk posted:What I meant is that the AI will consider building 2 mining districts because it needs 1.5 mining districts worth of ore, but probably won't think ahead to the fact that it has +mining pops and really only needs one.
|
|
#
?
Mar 6, 2020 14:22
|
|
|
Gort posted:Urban: The same micromanagement goes for governors with blocker clearing cost reduction.
|
|
#
?
Mar 6, 2020 14:23
|
|
|
My big hope right now is that the economic AI is decent enough that you can automate planet development without worrying too much about it.
|
|
#
?
Mar 6, 2020 14:26
|
|
|
Entorwellian posted:That guy sounds so awful I couldn't keep watching it after 2 minutes. You missed nothing; he spends most of the video setting up a lovely RP.
|
|
#
?
Mar 6, 2020 14:39
|
|
|
I bought a pile of DLC rather cheap recently (utopia/apoc/megacorp) just in order to use all the really big megastructures and ecumpoli stuff. I'm dragging the game out as much as I can instead of blobbing, so the verdict is still out, but one thing I have noticed is it's VERY much just slapping more cement onto a very poor foundation. Like, GunnerJ posted:My big hope right now is that the economic AI is decent enough that you can automate planet development without worrying too much about it. These things here are the real big problems with the game, and not from a grognard powergamer point of view, but rather a "I just want to sit down and play a fun game" type of issue. So when I get all these new toys to play with, it really feels like a shiny distraction that's not really doing its job well enough, since I'm still stuck here with a bajillion planets I can't automate, while the game chugs to a halt, and half the implemented core mechanics are either byzantine and/or nonfunctional (ground combat, slavery, etc). It's like yea, sure, sell DLC to fund continual development whatever, but fix that broken poo poo. I'm not asking for new assets here! Or total rebuilds or whatever. Just make poo poo work please?
|
|
#
?
Mar 6, 2020 14:51
|
|
|
Aethernet posted:You missed nothing; he spends most of the video setting up a lovely RP. I tried to watch thinking people were being overblown and oh my god his voice what the hell. Anyway the AI update looks great. Hopefully this and slowdown fixes will breathe some life back into the game.
|
|
#
?
Mar 6, 2020 15:02
|
|
|
I like Quill
|
|
#
?
Mar 6, 2020 15:09
|
|
|
cheesetriangles posted:I'm gonna buy the DLC no matter what because I'm a broken person that owns every DLC for every modern Paradox game even the cosmetic and music ones. I need help. This, except for the rocky bois one (which I'm told is actually p.-to-v.deece and I'll probably get at the next sale). Will Federations have anything for the modern, on-the-go genocidal space fascist?
|
|
#
?
Mar 6, 2020 15:10
|
|
|
Schadenboner posted:This, except for the rocky bois one (which I'm told is actually p.-to-v.deece and I'll probably get at the next sale). An even bigger super-mega-hyper-battleship because the merely super-mega-battleships apparently aren't enough. I'll admit it kinda feels like they're running out of ideas.
|
|
#
?
Mar 6, 2020 15:19
|
|
|
Serephina posted:These things here are the real big problems with the game, and not from a grognard powergamer point of view, but rather a "I just want to sit down and play a fun game" type of issue. So when I get all these new toys to play with, it really feels like a shiny distraction that's not really doing its job well enough, since I'm still stuck here with a bajillion planets I can't automate, while the game chugs to a halt, and half the implemented core mechanics are either byzantine and/or nonfunctional (ground combat, slavery, etc). Seams posted:I like Quill Schadenboner posted:This, except for the rocky bois one (which I'm told is actually p.-to-v.deece and I'll probably get at the next sale). I think the Diplomatic Stances and a few minor things will play into how space assholes will play a bit, via some numbers going up.
|
|
#
?
Mar 6, 2020 15:21
|
|
|
AAAAA! Real Muenster posted:Yeah this is the hill I will die on. I think it is *great* to see that they invested this much time in the AI and improvements, but this is just the tip of the iceberg for me. Fix the AI. Fix the jank that is still pretty much the same as it was at launch. WE SHALL CLEANSE THE GALAXY VIA MARGINALLY-HIGHER EFFICIENCY RATIOS.
|
|
#
?
Mar 6, 2020 15:25
|
|
|
ShadowHawk posted:What I meant is that the AI will consider building 2 mining districts because it needs 1.5 mining districts worth of ore, but probably won't think ahead to the fact that it has +mining pops and really only needs one. The Dev Diary says the AI factors in bonuses. Manually shuffling pops is stupid and expecting the AI to do it is bad. The game shouldn't be balanced around a lovely mechanic and luckily it's not.
|
|
#
?
Mar 6, 2020 15:55
|
|
|
You can absolutely win at Grand Admiral without ever shuffling pops at all, but that's partly because the AI sucks. It might become mandatory for high level play in 2.6. He says, dreaming of competent AI.
|
|
#
?
Mar 6, 2020 16:32
|
|
|
If the AI improvements actually pan out we'll have to turn the game difficulty down a notch or two, double income is a huge band-aid over not fundamentally understanding how the hell the game works.
|
|
#
?
Mar 6, 2020 18:21
|
|
|
Crazycryodude posted:An even bigger super-mega-hyper-battleship because the merely super-mega-battleships apparently aren't enough. I'll admit it kinda feels like they're running out of ideas. There is always the possibility of a bigger ship. ALWAYS.
|
|
#
?
Mar 6, 2020 18:29
|
|
|
Shadowlyger posted:There is always the possibility of a bigger ship. There's a point where you hit diminishing returns, like what's next they've gotta one-up themselves yet again and add a Leviathan-class warship that can solo an entire empire? It stops being interesting and cheapens the effect of the already existing fuckoff huge ships. Like the Juggernaut's special thing that differentiates it from being just Battleship But Better is that it has a shipyard and can act as a mobile FOB so you can repair your fleets without needing to go home. Ok, cool, great idea - but why's it also an invincible mega-battleship on top of that, now it's just stealing the Colossus niche. I'd prefer something like a relatively flimsy mobile shipyard that still gives you the strategic benefit of being able to bring your repairs with you, but you can't just stick it in the battle line it has to be a jump or two behind the front. Crazycryodude fucked around with this message at 18:39 on Mar 6, 2020 |
|
#
?
Mar 6, 2020 18:36
|
|
|
Crazycryodude posted:There's a point where you hit diminishing returns, like what's next they've gotta one-up themselves yet again and add a Leviathan-class warship that can solo an entire empire? It stops being interesting and cheapens the effect of the already existing fuckoff huge ships. Like the Juggernaut's special thing that differentiates it from being just Battleship But Better is that it has a shipyard and can act as a mobile FOB so you can repair your fleets without needing to go home. Ok, cool, great idea - but why's it also an invincible mega-battleship on top of that, now it's just stealing the Colossus niche. I'd prefer something like a relatively flimsy mobile shipyard that still gives you the strategic benefit of being able to bring your repairs with you, but you can't just stick it in the battle line it has to be a jump or two behind the front.
|
|
#
?
Mar 6, 2020 18:43
|
|
|
I don't mind the game adding more and more cool new bigtoys, I just wish more of them could be relevant in the mid-game (or even late game) rather than being victory lap poo poo. ETA: Just thinking about megastructures, I really don't understand why they all have a very expensive "site" step that is useless and some even have a more expensive and equally useless "frame" stage?? What does this add to the game. GunnerJ fucked around with this message at 18:48 on Mar 6, 2020 |
|
#
?
Mar 6, 2020 18:44
|
|
|
There's also the issue of the UI. With ships with lots and lots of slots, some of them aren't reachable, because things like the required items(Power, Computers, etc) get in the way.
|
|
#
?
Mar 6, 2020 18:46
|
|
|
More mega ships, but also more unique corvettes and cruisers and destroyers and battleships and give me targeting preferences in combat
|
|
#
?
Mar 6, 2020 18:46
|
|
|
AAAAA! Real Muenster posted:Right but why couldnt they add this poo poo to Titans? Titans are already artificially limited to an extreme degree so now we just have a different flavor of Titan that is even more limited? Its very bad design. Oh right duh I meant Titans. See, there's already too many fuckoff huge ships that I can't keep them straight. The Colossi I kinda understand because you have to have a Death Star in your space game, but there is no reason for the Juggernaut to exist as its own separate thing instead of like a Mobile Shipyard module for a Titan that strips most of the guns and defenses out but lets it act like a shipyard. That's actually a meaningful choice, do I spend my Titan cap on a gunned up Titan for killing things, or do I want the strategic benefit of repairs at the cost of firepower?
|
|
#
?
Mar 6, 2020 18:47
|
|
|
Splicer posted:More mega ships, but also more unique corvettes and cruisers and destroyers and battleships and give me targeting preferences in combat It's probably too much to ask but I'd like it if combat actually required mixed ships, and not just as many of the largest ships you can field. Keep corvettes useful past early game.
|
|
#
?
Mar 6, 2020 18:48
|
|
|
Splicer posted:More mega ships, but also more unique corvettes and cruisers and destroyers and battleships and give me targeting preferences in combat There's a mod for that, but, uh, yeah.
|
|
#
?
Mar 6, 2020 18:49
|
|
|
How about a missile component for Destroyers?
|
|
#
?
Mar 6, 2020 18:58
|
|
|
A destroyer with torpedos? How preposterous! (I am actually unsure if it's possible in Stellaris or not...)
|
|
#
?
Mar 6, 2020 19:04
|
|
|
|
| # ? Apr 26, 2024 16:55 |
|
|
Fhqwhgads posted:It's probably too much to ask but I'd like it if combat actually required mixed ships, and not just as many of the largest ships you can field. Keep corvettes useful past early game. I mean, Corvettes and Battleships are the only two ship types worth massing in the end game, so your wish worked. Battleships for murdering, corvettes for area control with a minor in murdering.
|
|
#
?
Mar 6, 2020 19:06
|
|