|
It was deep! Remember the ablative shield of artillery? I ragequit at least once after sitting through 60 seconds of the "strangled trumpeter" death sound effect, after a stack wipe.
|
#
?
Aug 5, 2021 02:25
|
|
|
|
| # ? Jun 1, 2024 16:22 |
|
|
FrancisFukyomama posted:Didn’t civ 4 have pretty competent AI but mostly bc combat was extremely simple? Ehh... After the final BTS version the AI got halfway decent. Most serious players played with the various Better AI mods though. And even then there was only so much they could do. edit: You had to play with the aggressive AI option turned on in BTS to get decent AI performance. I remember a goon from Firaxis talking about that here. They made a pretty big AI overhaul that made it much more competitive with the player, and as a result it was much less passive. Firaxis was worried that a lot of the more casual players wouldn't like the less passive AI, so they shoved the revamped AI into the aggressive AI option and never told anyone lol. (that option previously didn't do much other than make the AI declare war on the player more often) Dr. Video Games 0031 fucked around with this message at 03:50 on Aug 5, 2021 |
|
#
?
Aug 5, 2021 03:25
|
|
|
The main problem is the combat AI, which was much better in Civ 4 than Civ 5 or 6 ever got. People whinge about doomstacks, but at least the AI could actually get its troops to the front and inflict significant losses, instead of being choked up in a traffic jam getting shot with infinite arrows, and what is an army but a doomstack anyway? Also, limiting units to one per tile makes moving units even in peacetime a pain in the neck. The reason why bad combat AI is so killer to a game like Civ is that warfare is the answer to every problem in the game. Another civ's going to launch a spaceship? Kill them. Someone going for a culture victory? Kill them. You want to win a peaceful victory? Better make sure you have enough military to fight anyone who cares to stop you. If the AI can't effectively do combat, it can't effectively win the game.
|
|
#
?
Aug 5, 2021 10:48
|
|
|
Tree Bucket posted:It was deep! Remember the ablative shield of artillery? If you enabled stack vs. stack combat it played those 60 trumpets simultaneously and rattled the windows of the entire neighbourhood.
|
|
#
?
Aug 6, 2021 21:49
|
|
|
https://twitter.com/XboxWire/status/1424717100908785664 Neat.
|
|
#
?
Aug 9, 2021 14:10
|
|
|
victrix posted:I'm of the opinion that it's impossible, once system complexity passes a certain point In my view, it's sort of the opposite. With how they can now make AI learn by iteration, in the not-too-distant future it won't be difficult to make AI that just crushes all human players without any need for bonuses to resources and all that jazz in any 4x or strategy game. To me, the genuinely interesting question is: from a starting point of perfect or near-perfect play, how can they dumb down AI in games to match the 'feel' of human players? And not just have some variable like 'every 10th action make an intentional error'. All kinds of interesting and complex things to think about, beyond slowing down action speeds. Like can you narrow an AI player's effective 'FOV' so that it only notices certain a narrower band of things, as a human player would, rather than being aware of every single piece of minutiae on the map at any given nanosecond.
|
|
#
?
Aug 9, 2021 14:37
|
|
|
I'm not too worried, until I see evidence of a single game designer ever understanding the balance and value of systems in their own games.
|
|
#
?
Aug 9, 2021 15:13
|
|
|
Jeza posted:In my view, it's sort of the opposite. With how they can now make AI learn by iteration, in the not-too-distant future it won't be difficult to make AI that just crushes all human players without any need for bonuses to resources and all that jazz in any 4x or strategy game. To me, the genuinely interesting question is: from a starting point of perfect or near-perfect play, how can they dumb down AI in games to match the 'feel' of human players? And not just have some variable like 'every 10th action make an intentional error'. Supposedly the technology for that already exists. So why we are not seeing that? (genuine question, because I also expected that soon all new wonderful developments from the AI field would leak into videogames, but so far there's seem to be no sign of it) edit: I love Civ 6 but they managed to make an AI even worst than it already was in Civ 5. 16 years later
|
|
#
?
Aug 9, 2021 15:25
|
|
|
Poil posted:I'll have you know the combat was extremely deep and engaging with putting all your units into a doomstack and being the first to slap it into the enemy doomstack to win. honestly, this is the by far the best and maybe the only correct way to do combat in strategy games. especially in games that try to be cute and do "smart" combat. oh, i should use the correct units in correct ratios to win wars efficiently? no, gently caress you, i'm gonna become the production superpower and have more tanks than fit on entire goddamn map lmao. i love endless legend, and despite liking the tactical combat i avoid it like the plague, because really it's just busywork i can ignore and since i don't have infinite free time i instead spend more effort on number going up. early on when you only have 1-2 armies and a single city it's a nice distraction, but as soon as poo poo starts snowballing it's just too much micro for not nearly enough gain
|
|
#
?
Aug 9, 2021 15:31
|
|
|
Elias_Maluco posted:Supposedly the technology for that already exists. So why we are not seeing that? I don't know enough to answer that question with any real certainty, but I assume the technologies behind things like OpenAI and DeepMind are proprietary to the companies that develop them and are not yet in a state where they can be licensed as an actual product to game studios. Also no clue as to the infrastructure, compatibility and timescale requirements to actually run these things on a new game and get useful outcomes, but I assume they would be significant. Still, as proof of concept, it seems pretty clear that making bots play each other for millions of games can create AI that far surpasses human ability, even if you factor in putting in mechanical impediments so they can't act at like 10,000 APM. The trickledown of this into game development seems like a question of time rather than feasibility to me, but it might be 10 years away, it might be more. Or less? Who can say?
|
|
#
?
Aug 9, 2021 15:48
|
|
|
I don’t think it’ll happen any time soon because there’s no real incentive to do it. Civ V was played more than IV and VI more than V despite worse AIs - the majority of people just don’t care that much.
|
|
#
?
Aug 9, 2021 16:01
|
|
|
Or because they dont have any choice, like people said here: is bad everywhere, the whole genre is cursed by it I play Civ 6 because it is a fun game, a good 4x, despite the fact the AI sucks. But surely it would be better it the AI was better. People (at least a good number of them) do want a challenge, or they would not be playing in the highest difficult levels. And in the case of Civ, higher difficulties just mean such a absurd initial advantage for the AI players that it becomes kinda of puzzle game to catch up to it, which primarly affects early game. Once you catch up, the AI becomes 100% harmless, which is why late game is so boring no matter the difficult settings
|
|
#
?
Aug 9, 2021 16:09
|
|
|
AI programmer on PC strategy games has to be the most thankless job in game development. Most of the loudest people in the room (myself included) talk about it endlessly, but you can’t just make the AI good at the game, they have to be good at the game while still allowing the player to have fun and feel like they’re interacting with something of an organic opponent. And in doing so you just open up different kinds of exploits, so even when the AI is “good” when playing them straight up, you can break them in ways that look silly, and not doing so makes the AI a “cheater” instead for using information a player in the same situation wouldn’t have. It sounds impossible.
|
|
#
?
Aug 9, 2021 16:37
|
|
|
I think people also underestimate how much of a bubble they are in with regard to 4x games. The types of people complaining about the AI being braindead are also usually the 1% of gamers who tear things apart and figure out how to exploit things. I see it all the time in management games, for example. Someone will have thousands of hours in Oxygen Not Included, and come up with whacky poo poo (like using pooled liquids for air locks, etc.). This stuff then gets shared through internet videos, so it becomes collective knowledge of anyone who is hardcore into the game, and becomes essentially a baseline for their play. These are the people who are then often calling AIs or systems stupid. Meanwhile, Todd McCasual gamer loads up a game of Civ VI, doesn't watch YouTube videos on it, does sit there staring at every unit stat to figure out the synergy between Ottoman cannons and the Eiffel Tower, and just plays the game.
|
|
#
?
Aug 9, 2021 16:48
|
|
|
SlyFrog posted:I think people also underestimate how much of a bubble they are in with regard to 4x games. The types of people complaining about the AI being braindead are also usually the 1% of gamers who tear things apart and figure out how to exploit things. I dont know, Im not that kind of player myself. I never watched a youtube video to learn how to play a videogame, I dont minmax, I dont exploit the cheat out of games in general, and Im not a specially great player on any 4X. In Civ 6, Im a mild effective player who can win most of the time in Emperor difficulty, but usually with some restarts. So the challenge per se is pretty much enough for me. Still, it bothers me how dumb the AI is. How dumb it is at figthing a war, or even choosing units to build; and how completely clueless it is, and even more unjustifiable, on developing their lands, building their cities and so on Or how everytime I played Endless Legend I fell I was alone in the world because the AI is even worst than Civ and behaves like a dead corpse laying there in the map. Even though I suck at the game, it still bothers me
|
|
#
?
Aug 9, 2021 19:22
|
|
|
A big part of the problem is that 4X games tend to have an objectively superior strategy. You could make a much better Civ V AI by programming it to always choose Tradition and always research Civil Service as soon as possible, but then the different civs would have no personality. Making a good AI with personality is probably much harder than making a good AI, and the Civ developers seem much more interested in personality.
|
|
#
?
Aug 9, 2021 23:19
|
|
|
Really wish they announced this sooner so I could've saved a pre-order. Let this be a lesson to never pre-order games, even for beta access.
|
|
#
?
Aug 10, 2021 00:03
|
|
|
Dr. Video Games 0031 posted:Really wish they announced this sooner so I could've saved a pre-order. Let this be a lesson to never pre-order games, even for beta access. get you a refund! just mentioned the gamepass to steam and they did it in an hour
|
|
#
?
Aug 10, 2021 01:06
|
|
|
This is through Epic, and a refund isn't available through normal means since it's not eligible by their terms, but I may as well send in a ticket asking for a refund anyway. edit: Got it. 
Dr. Video Games 0031 fucked around with this message at 04:01 on Aug 10, 2021 |
|
#
?
Aug 10, 2021 02:30
|
|
|
Anno posted:AI programmer on PC strategy games has to be the most thankless job in game development. Most of the loudest people in the room (myself included) talk about it endlessly, but you can’t just make the AI good at the game, they have to be good at the game while still allowing the player to have fun and feel like they’re interacting with something of an organic opponent. And in doing so you just open up different kinds of exploits, so even when the AI is “good” when playing them straight up, you can break them in ways that look silly, and not doing so makes the AI a “cheater” instead for using information a player in the same situation wouldn’t have. It sounds impossible. I have a weird thing with games that are too hard for me to beat where I like to play on the hardest difficulties, but with some minor cheats to make it easier on myself. Kind of the reverse of what scaling the difficulty up usually does - the AI gets "better" and I'm the one who gets a few exploity cheats that aren't completely game-breaking to level the playing field. For example when I was learning how to play Total Warhammer I would slam the difficulty up to Legendary but start myself off with 100k gold, then make it a goal to keep a positive cash flow while having the safety net of that 100k gold to let me stay in the running if I made some poor choices. Personally I find it a good way to improve and learn how to play a game better - I can't regularly beat a Total Warhammer campaign on Legendary without cheating, but learning how to manage my economy against a much tougher opponent helped me regularly do it without the cheats on Hard and Very Hard. Anyway I mention it because I think an official developer-sanctioned version of this as an optional difficulty mode in games would be a cool way to ease up some of the stress you're talking about here - you can make an opponent that's strong enough to be a challenge for min/maxers without needing to worry about making it organically beatable for players who aren't stats nerds, while still letting more casual players have the experience of playing against that opponent and even being able to win. Maybe it's all a pipe dream and it would be far harder to design than I imagine, but I've always thought it was weird that most games handle difficulty by adjusting the enemy's power level, and very few handle it by adjusting the player's power level. e: As an example scenario for a 4x - playing against an AI that makes all of the right choices and doesn't predictably leave themselves open to dumb attacks to simulate human mistake-making - with my hope of winning being the fact that I'm the one showing up with an army worth more than my empire could feasibly produce and taking that AI by "surprise" after they calculated my army's value based on the unadjusted economy numbers and sent the rest of their forces elsewhere. I totally get that for some players that takes away from the sense of achievement and those more hardcore types are likely to concentrate around gaming forums like this, but as an optional mode I think it would be rad for the majority of the playerbase who might even feel achievement from conquering opponents that are too hard for them to have usually conquered - and in return, the hardcore types get their hardcore AI with no punches pulled to play against. Edit 2: To put it in other terms I think the new contemporary power fantasy for cool people is taking down established unbeatable opponents as an underdog with a bit of supernatural assistance (e.g. just look at the news, or superhero movies where a powerful superhero faces off against unbeatable odds) and the way difficulty is handled is a prime opportunity for experimenting with that in games. deep dish peat moss fucked around with this message at 06:10 on Aug 10, 2021 |
|
#
?
Aug 10, 2021 05:46
|
|
|
deep dish peat moss posted:I have a weird thing with games that are too hard for me to beat where I like to play on the hardest difficulties, but with some minor cheats to make it easier on myself. Kind of the reverse of what scaling the difficulty up usually does - the AI gets "better" and I'm the one who gets a few exploity cheats that aren't completely game-breaking to level the playing field. I've never thought of setting up an "easy" game for myself this way before, but it's a great idea and makes sense why you'd want the ability to set up starting conditions that way.
|
|
#
?
Aug 10, 2021 07:28
|
|
|
I don't expect the AI to meet me toe-to-toe in the battlefield but I do have some very basic expectations, like "build a plane" and "farm the land right next to the capital sometime before 1800" I often see civ 5 AI failing at one of these things. Guess which one? Both of them
|
|
#
?
Aug 10, 2021 07:37
|
|
|
Elias_Maluco posted:Supposedly the technology for that already exists. So why we are not seeing that? Well you just need to rent a supercomputer to train the neural network AI for a couple of months, playing millions of games to establish the optimal strategy. Then copy over the result and deal with the fact AI turns take several minutes each on home PC, other than that, yes the technology exists. Also if you patch the game balance then you have to repeat the whole process. Pyromancer fucked around with this message at 08:23 on Aug 10, 2021 |
|
#
?
Aug 10, 2021 08:20
|
|
|
drat, I like AI chat but I thought all these posts mean the game was out
|
|
#
?
Aug 10, 2021 10:16
|
|
|
Pyromancer posted:Well you just need to rent a supercomputer to train the neural network AI for a couple of months, playing millions of games to establish the optimal strategy. Then copy over the result and deal with the fact AI turns take several minutes each on home PC, other than that, yes the technology exists. Also if you patch the game balance then you have to repeat the whole process. The solution for that is probably online AI JeremoudCorbynejad posted:I don't expect the AI to meet me toe-to-toe in the battlefield but I do have some very basic expectations, like "build a plane" and "farm the land right next to the capital sometime before 1800" Thats whats saddens me on Civ 6 AI: ok, 1UPT combat is too hard for an AI, I get it. But goddam, programming it to actually use its builders dont seems that hard. Every AI civ late game will have a ton of undeveloped land, but strategic and luxury resources included, and will usually have some builders wandering around doing nothing too;
|
|
#
?
Aug 10, 2021 11:58
|
|
|
Is it going to be good???
|
|
#
?
Aug 11, 2021 18:42
|
|
|
The big red flag for me was the pacing and a lot of balance in the last beta, so that'll be the kicker for me if it's good or not. A lot of times I had enough stars to enter a new era before spending almost any time in the current one, and it was also way too easy to completely break the game with overpowered civs. Taking only two production cultures to have every city build everything in 1 turn or wealth and buyout cost reduction to trivialize everything was a big problem, not even including some of the larger offenders like Korea or the horde cultures with growing units
|
|
#
?
Aug 11, 2021 18:55
|
|
|
The thing that worries me with this game is it’s like they came up with a bunch of really cool and innovative ideas but don’t really seem willing/able to either balance them or pare them down in a way that actually works mechanically. Might have fallen a little too in love with their own creation.
|
|
#
?
Aug 11, 2021 20:45
|
|
|
It was also so easy to just vassalize the AI and have infinite money to just buyout everything. I've been sitting on pre-ordering the game but at this point I'm scared off of it and will just wait for reviews first.
|
|
#
?
Aug 11, 2021 20:54
|
|
|
Jeza posted:In my view, it's sort of the opposite. With how they can now make AI learn by iteration, in the not-too-distant future it won't be difficult to make AI that just crushes all human players without any need for bonuses to resources and all that jazz in any 4x or strategy game. To me, the genuinely interesting question is: from a starting point of perfect or near-perfect play, how can they dumb down AI in games to match the 'feel' of human players? And not just have some variable like 'every 10th action make an intentional error'. Check out https://maiachess.com/ if you're interested in this topic.
|
|
#
?
Aug 11, 2021 21:16
|
|
|
Are there well-written early reviews yet or is this still embargoed? (If it's embargoed... that's not a great sign)
|
|
#
?
Aug 12, 2021 17:13
|
|
|
Pyromancer posted:Well you just need to rent a supercomputer to train the neural network AI for a couple of months, playing millions of games to establish the optimal strategy. Then copy over the result and deal with the fact AI turns take several minutes each on home PC, other than that, yes the technology exists. Also if you patch the game balance then you have to repeat the whole process. The other huge issue is that you cant really tune it. If the AI develops a behavior you dont like, you can just rewrite that part of like you can with a normally written AI. Its a black box and while you can just retrain it, its gonna be a crapshoot everytime what kind of nuances it decides to glom onto to.
|
|
#
?
Aug 12, 2021 17:21
|
|
|
alcaras posted:Are there well-written early reviews yet or is this still embargoed? (If it's embargoed... that's not a great sign) Embargo seems to drop on the 16th. Anything pre-release is plenty fine imo
|
|
#
?
Aug 12, 2021 17:42
|
|
|
alcaras posted:Are there well-written early reviews yet or is this still embargoed? (If it's embargoed... that's not a great sign) Seems like review embargo until release from what some redditors said that steamers said. They’re also not allowed to show the contemporary era or any of those random new mechanics like pollution. Lotta warning signs
|
|
#
?
Aug 12, 2021 18:04
|
|
|
Twigand Berries posted:get you a refund! just mentioned the gamepass to steam and they did it in an hour same, pretty hard to justify buying it for nearly $50 when I can get it for 10
|
|
#
?
Aug 12, 2021 22:53
|
|
|
dy. posted:The thing that worries me with this game is it’s like they came up with a bunch of really cool and innovative ideas but don’t really seem willing/able to either balance them or pare them down in a way that actually works mechanically. Might have fallen a little too in love with their own creation. This seems to be a common theme in Amplitude's 4x games. Endless Legend was a fluke in that it just clicked with itself and worked for the most part. That's why I've been so bummed out about this game since they first announced it - the thing that always made their 4x series worth playing for me was the really weird, bizarre and imaginative universes that set the backdrop for the kind-of-okay gameplay. I'll give them a lot of credit for always trying to do things a bit differently from everyone else but it doesn't seem like they know how to make "different" and "excellent" go hand in hand  edit: lmao I made almost this exact same post in here two years ago this month. I have completely forgotten about all things pre-covid I guess. deep dish peat moss posted:This is exactly the problem with Amplitude making a historical-inspired game. They've been all about world building above all else which has cultivated a certain fanbase for their games and now they're making a game with none of the core hook that drew that fanbase in. So they're going to lose out on some of their core fanbase like me because it doesn't have the things I buy their games for in it, and it's probably going to disappoint the wider audience because their games always end up a little bit on the shallow side but propped up by fantastic lore and creative, unique settings edit: For what it's worth I refund games on Steam a lot and in my experience you literally never need to enter a reason (no reason not to for the gamepass thing though). If you meet the "within 2 weeks and under 2 hours of play" criteria it's auto-approved, and preorders or anything else that's not playable are approved even beyond the 2 weeks. I would guess they have a 60 or 120 day cutoff period for refunds but afaik you can always get steam wallet funds if nothing else. I like using Paypal for Steam because there's no processing time on their refunds so I usually have the money back a few minutes after requesting a refund 
deep dish peat moss fucked around with this message at 03:40 on Aug 13, 2021 |
|
#
?
Aug 13, 2021 03:29
|
|
|
Veryslightlymad posted:I'm not too worried, until I see evidence of a single game designer ever understanding the balance and value of systems in their own games. Designers don't have to ubderstand that. The AI will figure it out
|
|
#
?
Aug 13, 2021 03:58
|
|
|
I'm sure it's been posted before but regarding AI in strategy games I'll always recommend Soren Johnson's video: https://www.youtube.com/watch?v=IJcuQQ1eWWI
|
|
#
?
Aug 13, 2021 07:35
|
|
|
Charlz Guybon posted:Designers don't have to ubderstand that. The AI will figure it out You do actually have to have an understanding of what good outcomes are to make machine learning approaches work- also, pretty much all of our current machine learning success stories are much, much less complex games than any 4x title.
|
|
#
?
Aug 13, 2021 10:48
|
|
|
|
| # ? Jun 1, 2024 16:22 |
|
|
Panzeh posted:You do actually have to have an understanding of what good outcomes are to make machine learning approaches work- also, pretty much all of our current machine learning success stories are much, much less complex games than any 4x title.
|
|
#
?
Aug 13, 2021 13:27
|
|