mobas aren't even strategy games, they also peaked in popularity a decade ago and everyone else has moved on since
when they were still popular the genre was dominated by koreans
They are literally the current most advanced form of the RTS. Even moreso, since you need to coordinate 5 people, rather than doing the work entirely yourself.
8 months ago
Anonymous
at best they're a watered-down team-based action-rpg spinoff of RTS
8 months ago
Anonymous
>They are literally the current most advanced form of the RTS.
The frick did I just read?
dota is 80% strategy and koreans never were good at dota because of that
8 months ago
Anonymous
>dota is 80% strategy
kek no it isn't, you might as well start including team-based shooters under the strategy umbrella if you're going to stretch the definition this far since there is some element of team planning involved
Literally the first time I've ever heard anyone say anything about Europeans being good at strategy games, especially since they fricking suck, or at least the bongs and germans do. Russians seem to be pretty decent.
But quit trying to steal the thunder from Asians. Everyone knows the Chinese, Koreans, and Flips are all the best strategy players.
>best dota teams are all europeans
wdhmbt?
DOTA isn't a strategy game, moron. It's an isometric arena fighter.
for me its the issue of pausing and switching between monitors to do something else for a while. I do that all the time in singleplayer while multiplayer requires 100% commitment at all times otherwise you lose, too stressful and tiring for something that isnt a 40m session.
I'm tired of casuals pretending that bumping up health/damage values for higher difficulties is some sort of sin. It can change your strategies up as much as doubling the size of an encounter depending on the game.
Its not that we're too "casual", its that cheating buff suck all the fun and are a sign of low effort from the dev's part.
Take civilization V for example. The AI is bottom of the barrel dogshit and every difficulty above Prince gives them buffs. Thing is, its still the same dumb AI so although these massive buffs make them overpowered early game, you can quickly catch on.
As long as I survive to the medieval era, I'm very likely if not certain to win my Diety game. And usually I can tell within the first 5 turns whether I'm going to survive to the medieval era or not based solely on who my neighbours are and what the map is like.
So although they increased difficulty with stupid buffs, the AI plays so poorly that you can catch up to them by optimizing if you just bide your time. Wow! How fun! (-_-)
Its sad that we will never get good AI in most strategy games because morons like you defend the low effort slop we get
>Play a newb >It's worse than AI >Play a no-lifer >Basically an AI following a flowchart >Play with a friend >Co-op comp-stomp
Seriously though, 4x AI at least is shit because
A: The player can savescum
B: The game usually progresses like >I can tell within the first 5 turns whether I'm going to survive to the medieval era or not based solely on who my neighbours are and what the map is like.
rather than any actual strategy. To many things matter other than just winning battles. An AI that plays safe will lose to a player that goes all out knowing they only have to be stable enough to win, and an AI that goes all out will be beat by a player that waits for it to crumble. If an AI does win, it's typically because it snowballed harder than the player. Even when you have real life strategy games, they get "solved" in such a way that you can prove one player is at a disadvantage, so getting ahead and staying ahead is an inherent problem of strategy games.
A semi-C problem is AI isn't typically programmed to give feedback, so you don't know if it's being stupid or just got a bad spawn point and is being overrun by npc factions, and every action it does seems random even if your own are no less so. Players also seem convinced AIs should be suicidal and attack their armies even when it knows it cannot win just because a player might use several waves to kill a superior army by attrition.
>Its sad that we will never get good AI in most strategy games because morons like you defend the low effort slop we get
Nah, morons like you think you can good AI but in reality good AI would rape you all day long. In the end you'd complain about the game being too hard
>the developers should design a shitty AI, a mediocre AI, and a good AI, and then reserve the good AI for only the top difficulties
statements dreamed up by the utterly deranged. developers should create the best AI they possibly can and then use it at all levels. creating a great AI and then neutering it is fricking moronic.
no?
higher HP pools just make you do the same shit longer
unless you've made a bad game that lets people ignore mechanics by dpsing hard, then that you're fault and you have way bigger problems than just the hp pool being too low
You got a point anon even if it is not systematic.
Boring games get more boring when you extend encounters/give more health or stuff to ai. It is a simple as that.
If you repeat 3 same braindead moves againt an enemy, of course it will be frustrating if he was stronger.
If game give many strategy/tactics and different ways to approach things, it would be fun to take on someone who is cheating.
People like you disgust me.
You're the same type of people that would defend bots in games having infinite [resources], be it fuel, ammo or whatever else.
Or worse, AIs that just magically know the whole map inside out and know exactly where to go.
You have unfair advantages of your own. Might as well cut the machine some slack.
There is no reason a very rudimentary learning system can't be made.
It's extremely fricking easy, and more so today than ever, systems have plenty of memory to play with for even more advanced and finegrained learning.
No, don't come out with "duuurrr u expect GPT" bullshit, things like this were done in games before where AI didn't have perfect understanding of their (dynamically changing) environment and had to adapt to it.
Lazy ass AI devs aren't an excuse.
It's not hard to do something as simple as giving them a few virtual eyes and a basic tile-based memory of where things are in the map, each having a decay rate so they forget areas not travelled regularly, and a desire to stay away from the most active areas unless it has certain resources for example. (this can apply to all types of games, be it RTS, racing or FPS)
People make shit like that for fun to try speed-run games automatically, but people seem to get it in their heads that it needs a top-tier GPU and thousands of hours to do that shit. No, that's for fricking thousands of iterations running thousands of times per second usually, not mere dozens of units running in realtime!
Having an AI that can learn alongside you can make vastly superior bots and a more enjoyable experience. (and having the ability to delete their memory to start over, likewise, if you want a fresh start / haven't played in a long time)
Bulletspongeitis is no excuse to be lazy with something that doesn't even take all that much effort to do, especially for a typical gamedev with their skillsets.
Simple AI like this isn't that hard, you're literally tile-counting in an array and assigning an emotion to the tile based on how active it was / how often their assets died in the area / what utilities or resources are there, etc.
I'm not making an RTS game.
Although I would like to, if I could come up with an engaging enough idea.
Ideally I'd love to make a galaxy-scale grand strat in the Stargate universe, but that's an IP nightmare, and now Amazon own it.
It's criminal how little good games are set in that universe. Fricking MGM.
7 months ago
Anonymous
You don't have to make an RTS, just make a proof of concept or an AI that can play an existing RTS. Doesn't have to be an RTS. You can use Endless Space 2's AI editor and see if you can make something decent out of that game's AI.
7 months ago
Anonymous
Why? Why repeat things that have been done dozens of times by others?
There are videos all over Youtube of people making bots capable of learning entirely new levels from scratch.
Some of them are, admittedly, horribly brute force shitty genetic algorithm crap where they die all the fricking time, others are specifically geared with knowledge on how to play the damn game they are playing instead of wasting the time re-inventing the wheel.
Generally speaking these bots are designed to be run in the thousands in parallel to brute-force speedruns, but I know there are bots that learn level layouts but I couldn't find you them even if I tried because Youtube search has been so fundamentally broken for the past 2 years it is impossible to find anything even if you know most of the fricking title. (and sometimes it literally won't even show you if you DO know the full title)
You're acting like this is some sort of super duper turbo hard AI. It's literally adding weights to tiles and weights to the relationships between those tiles, and that information passed on top of its usual pathfinding algo to influence exactly where it will go instead of just giving it exact knowledge of a map and where everything on it is.
The reason devs don't do it is lazy bastards.
Only a handful of devs care to give their AI dynamic verbs and limited memories.
Hell man, loads of roguelikes give YOU memory - as in, memory of where you have explored on the world.
7 months ago
Anonymous
If it's so easy why aren't there any strategy games that have it and why won't you be the first?
7 months ago
Anonymous
Wasn't Google showing off it's AI that could play Starcraft at the level of a pro years ago? Whatever happened to that?
They managed to make AI that could play chess better than the best humans and now the cheapest mobile phones can beat the best humans. Why can't the same be done for other games?
7 months ago
Anonymous
Because chess, like many other tabletop games with fixed moves, is ultimately a scam. There are only so many positions and all the best ones are now known. In even a simple 4x you don't only need to consider the game board and pieces but each piece can move different numbers of spaces as it feels rather than being locked in and can have different stats other than movement speed. You also have to worry about an economy and maybe even stupid shit like morale. If it's not single units, you have to worry about overall army composition as well. Also each side moves all its units at once instead of one per turn.
Simple games are simple to make good AI for even though if it's too simple it runs the risk of being solvable. I think this game is simple and has decent AI https://www.falcom.co.jp/vantage/index_e.html although most of it comes from maps being weighted against you.
Going on a tangent, I think the pursuit of a good battle AI is pointless. AI is made by a human, so you're just fighting a human that doesn't adapt. If you're good you'll win if the enemy is a human or an AI, which is why handicaps exist even in real life. An AI that can make good and responsive or planned armies (a lot of AI can do this but rarely get the chance), AI that can handle economies, research, building stuff properly and managing its non battle aspects, an AI that is good at diplomacy and working together on coordinated actions. That's what should be being pursued. Ultimately
https://i.imgur.com/AoSAABe.png
>higher difficulty >its just the AI cheating with insane buffs
doesn't come from a bad battle AI but from the AI being unable to manage its resources or unable to go all out, so they get boosts to simulate what they could do if they were able to manage themselves properly.
7 months ago
Anonymous
>Whatever happened to that
Just like with openai in dota it was a PR move to get investments/get acquired by a big company which they successfully did. Videogames were just a means to an end
You don't get it
there's very little fun in beating an opponent that you know is cheating
it can be fun for the first or maybe second time
after that you figure out all the gamey mechanics, all the blind spots AI has that you can exploit and so on and the game is exactly the same every time
Sounds like a 'you' problem. Anons use the cheesiest strats against AI over and over and then wonder why they're not having fun. Then they use the exact same stuff on a human opponent - who can actually get buttmad over it, ragequit and cry about it being OP on Di*cord - and they enjoy it because it was a fellow fleshbag who smashes his keyboard in anger and not a machine that can't get mad no matter how much of a dick you are.
Forget boosting AI's proficiency, just hook it up to ChatGPT or something to have the computer talk shit about your mom whenever it sees you wall off its resources 2 minutes into the game. That way players will get a kick out of it regardless of the artificial opponent's ineptitude.
This is why I play SP games on mostly normal-hard. Above that the AI is obviously cheating and that pushes you into turtling slogs or weird gamey strategies until you optimise fun out of your game. I'd just use whatever I find fun even if it's not optimal and roleplay. For real fair contest of skill you play MP games and it can be fun too, but it's a different thing.
Good AI takes time. You don't form plans in 5 seconds and neither will the AI. If they made a good AI, then you would complain it takes 15 minutes per turn.
The kinds of computations that are necessary for good decision-making in strategy games does not entail "deepmind-style AI", and is not necessarily anything that would be expensive to compute. Computers are much faster at a variety of tasks compared to humans, and it is really not appropriate to assume that outplaying even a skilled player at a 4X game would require a lot of processing time for a consumer PC.
Eagerly awaiting the low-effort insult that I already know you're programmed to respond with.
Dota and Civilization are incredibly difficult for an AI to solve. Civilization beyond its reliance on hidden information requires insanely long-term thinking, even more so than go, while also being build on a significantly more expansive game than both Starcraft/Go. There's just so much stuff, so many units, trees, policies, etc.
Dota is complex in the same way Starcraft is (hidden info, continuous space, real time), but also has a much larger amount of stuff in the game than starcraft does. The endless varieties in heroes, abilities and items make generalization really really hard. Especially since these factors also massively influence long-term decision making. But what may make Dota the hardest game, is that you're supposed to work with teammates, and communication with teammates (verbal/nonverbal) is absolutely essential to winning. The AI would need to be able to not just handle the enemy team, but would also need to be able to read its teammembers, which may require general intelligence.
Every time someone posts about moronicly huge ai cheats I'm reminded of the stainless steel mod where the ai gets so much cheats the best strategy is to not attack anyone for the first 20 or so turns so they can cheat their way to a maxxed settlement because it would take you 80+ turns to grow it instead so the faster you attack the ai the weaker you will be later when you have to fight multiple stacks of endgame units with unbreakable morale and other huge buffs if they're led by the king or his heir.
and this is on normal.
i dont get how this is still practice, its outdated as frick approach, we are in the age of ai, we have genius chess ai since 30 years, u cant tell me its that hard to make a competent ai in a videogame
the genius NN ai or chess ai's are made by turbo autist computer scientists, they are VERY difficult and expensive to make. Game devs are just average coders, none of them have anywhere near the expertise required to make an ai like that.
It's extremely easy to make an AI that can consistently beat humans without "cheating" (stuff like extra resources, stronger units, perfect vision through fog of war, input reading). The problem is doing it in a way that's interesting rather than the AI simply have flawless and unrelenting execution.
i wan an actual intelligent ai that can learn from playing against me and maybe end up reading and predicting me. that would be cool. no cheating shit, same stats, resources, fog of war etc as me.
>Wasting countless game dev hours trying to get a best-in-class AI to perform around the level of a human because you're too scared to hit "Multiplayer".
Lol. This is a complete lie peddled by Google´s and Elon Musk´s ai teams that couldn´t admit they wasted millions of dollars to create ai´s in dota and starcraft that ended up getting beaten by casual players. There is simply no truth to it and that´s why I also know how far they must be from creating anything actually resembling general intelligence.
The fact is that the dota ai could only play 17 out of the 124 heroes and when they went online with it, it still got beat 10 times in a row by the same casual players. In a row. With only 17 heroes to choose for both the ai and humans. That´s pathetic.
The same happened with Google in Starcraft. They made the ai´s, they had several, available to play at some starcraft convention. It got beaten constantly by causual players. Since starcraft requires more mechanics then dota, that ai had an easier time brute forcing it´s way to some victories. Online it became grandmaster, playing against humans. That´s pretty good. On the other hand you can go on youtube and find players not good enough to be pro´s that do moron runs like.
"getting to grandmaster on only one base"
"getting to grandmaster using only one unit"
So while they did get to grandmaster, it´s not a bigger deal than really good players can do it for fun.
Its not that we're too "casual", its that cheating buff suck all the fun and are a sign of low effort from the dev's part.
Take civilization V for example. The AI is bottom of the barrel dogshit and every difficulty above Prince gives them buffs. Thing is, its still the same dumb AI so although these massive buffs make them overpowered early game, you can quickly catch on.
As long as I survive to the medieval era, I'm very likely if not certain to win my Diety game. And usually I can tell within the first 5 turns whether I'm going to survive to the medieval era or not based solely on who my neighbours are and what the map is like.
So although they increased difficulty with stupid buffs, the AI plays so poorly that you can catch up to them by optimizing if you just bide your time. Wow! How fun! (-_-)
Its sad that we will never get good AI in most strategy games because morons like you defend the low effort slop we get
>Its not that we're too "casual", its that cheating buff suck all the fun and are a sign of low effort from the dev's part.
It´s not low effort. It´s just not possible to do it. If they could, Google and Elon Musk´s ai teams would have had better results, they were aiming to get ai´s that were just as superior as the one in chess and they failed. They could have tried in a Civ game, that way you take away the ai´s superhuman reflexes. It would only have been a bigger disaster and Civ casuals would laugh when they found out, that the super duper ai can´t play random maps and it still loses.
>It´s just not possible to do it
i'm going to interject here.
the AI doesn't have to be better than decent players to be considered good.
it can be given a bunch of strategies to do.
like cavalry in prophecy of pendor mod for warband, they always go for a flanking attack or hammer and anvil and aims for archers instead of just charging head on like in the base game.
It's not hard to beat the player as long as options are limited.
Lets take an rts like starcraft. You could extremely easily make an ai that could stomp all but pro players because there is an optimized build order for most strategies, you simply feed this to the AI and it excutes it flawlessly. There is also unit micro to give you the advantage in fights, even pro players use groups which are inefficient because it's physically impossible for anyone to control every single unit individually. AI could do it though.
But it wouldn't be fun.
An 'AI' opponent can't actually think, you know. It only follows the logic the devs gave it and they can't hope to anticipate the crazy tactics players can come up with. So don't treat the computer as an opponent because it can't be one.
But even though it's not smart, it can be entertaining. It has it's advantages: it's always ready for a game, it never gets tired or pissed, you can customize it for a relaxing compstomp or a brutal slog against multiple foes, it relentlessly executes multiple actions at the same time, it won't tell you to uninstall the game and kys for being a noob.
But because it's literally moronic, it needs a handicap. In fact, some of the most fun SP games are designed to be asymmetric in the first place, so that you could use your brains against the enemy's superior brawn.
I feel like we've actually seen AI stagnate or even decline across the entire gaming industry. Half-Life and F.E.A.R did their best and nobody has even tried since.
>maybe
What a horse shit.
Haven't met a single person, ever, who'd go out of his way and claim - "wow the A.I. is so fricking good in this game....".
Hearing the opposite quite often though.
What purpose do these rat articles even serve? Are they for investors?
Good enemy AI is easy in fps and other action games. Enemies live for few seconds, are preset in locations and don't have to keep up with player and develop alongside him, using these same game mechanics as him. There is no need for them to plan long term strategy, make interesting maneuvers during battle and develop their cities. Actually you can easily make them annoyingly superior to player with perfect aim
>enemies don't have to use the same game mechanics as him
This is what SP-focused strategy games already do, actually. You have enemies who play a completely different game than the player and don't follow the same rules and it's okay as long as it's challenging and fun to fight against them.
Everyone knows, for instance, that campaign scenarios in RTSes often have the player start with a rudimentary base while the enemy has several bases, multiple layers of defence, and tons of starting units that will be let loose after giving you some head start, so you can't exactly employ MP strategies like rushing.
How did you get that from my post you moronic Black? I want to play against an AI that simulates human players, not one that cheats and receives insane buffs to compensate for beign dumb as bricks. Thats never fun.
Not a strategy game but another example that comes to mind is Mount and Blade: Warband. The AI lords dont have to play by the same recruitment rules as the player, they can just magically spawn new fully-trained units after a while. Whereas human players have to go manually from town to town and village to village to collect recruits and slowly train the into competent units. This doesnt make the game particularly hard but it sucks the fun out of victories to know that they dont have to jump through the same hoops as you after they get defeated
>but it sucks the fun out of victories to know that they dont have to jump through the same hoops as you after they get defeated
Aren't you able to automate recruitment once you get a castle?
This is possibly the most low IQ post I've ever seen on /vst/.
If you have done 5 minutes of googling on AIs and difficulties and game design, you would know that (you) as a player don't know what you fricking want. "Good AI" is just smoke and mirrors on the part of the developer and it's not even actually good, it just pretends to be good, but actually is just bad enough so that the player can still win. For example, F.E.A.R. AI was praised for its tactical genius when in fact it was mostly a bunch of radio chatter screaming "I'm flanking them" or whatever, so it seemed smart. It's like when everyone complained that music shuffle wasn't random when it was in fact completely random, so now it's trying to predict how to make you feel like it's random.
The problem with many /vst/ genres is that they try to make AI do the same exact stuff as the player, so there's not really a good way to hide the tricks. AI war is a good example of how asymmetry can be used to make AI seem much better. The dev also posted a bunch of dev diaries on the topic, I highly recommend it.
Is a good AI one that plays like a metagay or one that plays like it's actually acting out its role?(e.g. playing like it's actually a horde of orcs or Britain during WW2 and making decisions appropriate for that even if they aren't the best ones.)
We don't really want *good* AI, we just want something that challenges us enough that we feel smart when we beat it. But, as others have pointed out, even that is much easier said than done.
The issue with RTS AI isn't that they are dumb or incapable of complexity, the issue is that they are given the most basic set of instructions to follow so the devs don't spend a frickload of resources testing, debugging and refactoring the code for it so neither your computer nor your internet shit the bed because of its sheer size
A RTS AI will never be programmed to take into account what faction it is playing, what strategy and goal it can go toward, how to manage their economy to achieve that goal efficiently, how to position and attack the enemies, how to micro different fronts in a smart way, how to adapt to the enemies producing counter units, so on and so forth
Only God know how much i want to witness an AI spitroasting me with a frontal assault, a flanking attack and a hit'n'run on the worker units at the same time, it would be a painful AI to deal with if you have tunnel vision or only swear to the almighty singular unit blob of doom, but it would leave one hell of an impression in return
Did anyone ever try to make a reinforcement learning AI that got trained by playing against decent players or mimicking their actions? I imagine something like this would probably completely frick up 99.9999% of the user base but it's better than cheating.
I mean cheating makes things harder for sure but the AI is still dumb, I'm not way a good player but for civ for example the main thing you need to do vs diety is knowing how to please them all while rushing shit like cavalry or cuirassiers because even diety AI can't deal with a min-maxing human player
It wouldn't fix the fundamental problem of AI being unable to adapt. Even if they used top meta tactics, if it used the same ones over and over, anyone would eventually become able to beat it.
The only game I've heard of that did that was Quake which wasn't a strategy game, I don't know if a strategy game would be too complex to have a similar ai.
>I don't know if a strategy game would be too complex to have a similar ai.
If anything it would be easier. It's way more complex for AIs to learn movement and positioning, than juggling a few variables, which is what most strategy games boil down to. Not to mention, Quake is real time, while many strategy games are turn based so it has more time to process through everything.
>its just the AI cheating with insane buffs
this is what I always hated the most in all kinds of strategy games
in modern times it had only become worse since most developers don't care about making good AI opponents
Iron Harvest is probably the worst offender I still got in my recent memory >easy
is just AI playing with the same things available to the player >medium and hard
are just options of "how much do you want the AI opponent to cheat"
>multiplayer competitive games >it's just rock paper scissors with step by step tryhard guides >if you don't play this specific way you'll always lose
lol
The same applies to every competitive event, including irl sports where everything is solved for optimum output: techniques, equipment, training, even diet.
>multiplayer competitive games >it's just rock paper scissors with step by step tryhard guides >if you don't play this specific way you'll always lose
lol
The same applies to every competitive event, including irl sports where everything is solved for optimum output: techniques, equipment, training, even diet.
Yeah MP is just a confidence trick where you scam less capable players into matches you know they will lose.
This is why difficulty sliders shouldn't exist in games to begin with, because attempts to make the game "more challenging" just involve lazy design of artificially inflating hp and damage numbers, instead of building encounters specifically to be a rewarding obstacle. This is something that Dark Souls managed to actually do right, in that it told difficulty sliders to frick off and made the game an actual experience.
Age of Empires 2 DE AI had been re-written to follow some common online strategies and have 0 cheats, the difficulty is mostly hard-coded with the amount of economy they're allowed to use + some higher tier strategies locked behind the final difficulty.
It's not perfect and still folds to rushes but it offers a decent challenge in the mid and especially late game.
Much more fun than >The AI straight up just gets free shit
I'm starting to think this genre is dead cause of the difficulty curve and the game throws at you whatever it feels like if you try to play it at harder difficulty despite learning all the basics. There's a reason why MOBA exists which was conceived from Warcraft 3 cause the younger generations, the zoomers are too intimidated try out RTS.
Tell that for the 99% of people who letf the multiplayer of Starcraft 2 because it only consist of wich player is the fastest to produce basic units and destroy the enemy base in the first 3 minutes of the game.
AKA "fun gameplay".
You can't just wing it in an RTS unlike other online games. At the very least you need a general idea of what you're trying to do and know the precise steps to make it happen. You also need to hurry the hell up while keeping track of all the variables.
Play against other human beings if you want a real opponent.
No
>higher difficulty human opponent
>It's just a european with better education
Europeans always get mogged in competitive strategy games
>best dota teams are all europeans
wdhmbt?
mobas aren't even strategy games, they also peaked in popularity a decade ago and everyone else has moved on since
when they were still popular the genre was dominated by koreans
They are literally the current most advanced form of the RTS. Even moreso, since you need to coordinate 5 people, rather than doing the work entirely yourself.
at best they're a watered-down team-based action-rpg spinoff of RTS
>They are literally the current most advanced form of the RTS.
The frick did I just read?
dota is 80% strategy and koreans never were good at dota because of that
>dota is 80% strategy
kek no it isn't, you might as well start including team-based shooters under the strategy umbrella if you're going to stretch the definition this far since there is some element of team planning involved
TheViper disagrees
>Europeans always get mogged in competitive strategy games
Oh, you mean games like StarCraft 2?
Europeans always get mogged
lmao. european pro teams that consist of 100% ethnic asians maybe
Thats a statement i haven't seen before
Typically everyone is just whining about koreans
Literally the first time I've ever heard anyone say anything about Europeans being good at strategy games, especially since they fricking suck, or at least the bongs and germans do. Russians seem to be pretty decent.
But quit trying to steal the thunder from Asians. Everyone knows the Chinese, Koreans, and Flips are all the best strategy players.
DOTA isn't a strategy game, moron. It's an isometric arena fighter.
Also, Evil Geniuses aren't European.
>Warcraft 3
Yuropoors are pretty dominant in AoE2 to be fair
>play human opponent
>they're somehow worse than AI at the game
no
Introduce me to some humans who are willing to play 4x 10-14 hours a day on weekdays.
for me its the issue of pausing and switching between monitors to do something else for a while. I do that all the time in singleplayer while multiplayer requires 100% commitment at all times otherwise you lose, too stressful and tiring for something that isnt a 40m session.
you can't tell /vst/ this, these people have no friends and are terrified of playing against strangers.
>computer beats humans at chess and go and shit
>somehow we're still stuck with human opponents for challenge in strategy games
I'm tired of casuals pretending that bumping up health/damage values for higher difficulties is some sort of sin. It can change your strategies up as much as doubling the size of an encounter depending on the game.
Frick off boring low IQ lazy dev
Make the AI itself better at higher difficulties
Buffs are boring shit
Its not that we're too "casual", its that cheating buff suck all the fun and are a sign of low effort from the dev's part.
Take civilization V for example. The AI is bottom of the barrel dogshit and every difficulty above Prince gives them buffs. Thing is, its still the same dumb AI so although these massive buffs make them overpowered early game, you can quickly catch on.
As long as I survive to the medieval era, I'm very likely if not certain to win my Diety game. And usually I can tell within the first 5 turns whether I'm going to survive to the medieval era or not based solely on who my neighbours are and what the map is like.
So although they increased difficulty with stupid buffs, the AI plays so poorly that you can catch up to them by optimizing if you just bide your time. Wow! How fun! (-_-)
Its sad that we will never get good AI in most strategy games because morons like you defend the low effort slop we get
>Play a newb
>It's worse than AI
>Play a no-lifer
>Basically an AI following a flowchart
>Play with a friend
>Co-op comp-stomp
Seriously though, 4x AI at least is shit because
A: The player can savescum
B: The game usually progresses like
>I can tell within the first 5 turns whether I'm going to survive to the medieval era or not based solely on who my neighbours are and what the map is like.
rather than any actual strategy. To many things matter other than just winning battles. An AI that plays safe will lose to a player that goes all out knowing they only have to be stable enough to win, and an AI that goes all out will be beat by a player that waits for it to crumble. If an AI does win, it's typically because it snowballed harder than the player. Even when you have real life strategy games, they get "solved" in such a way that you can prove one player is at a disadvantage, so getting ahead and staying ahead is an inherent problem of strategy games.
A semi-C problem is AI isn't typically programmed to give feedback, so you don't know if it's being stupid or just got a bad spawn point and is being overrun by npc factions, and every action it does seems random even if your own are no less so. Players also seem convinced AIs should be suicidal and attack their armies even when it knows it cannot win just because a player might use several waves to kill a superior army by attrition.
>Its sad that we will never get good AI in most strategy games because morons like you defend the low effort slop we get
Nah, morons like you think you can good AI but in reality good AI would rape you all day long. In the end you'd complain about the game being too hard
>the developers should design a shitty AI, a mediocre AI, and a good AI, and then reserve the good AI for only the top difficulties
statements dreamed up by the utterly deranged. developers should create the best AI they possibly can and then use it at all levels. creating a great AI and then neutering it is fricking moronic.
no?
higher HP pools just make you do the same shit longer
unless you've made a bad game that lets people ignore mechanics by dpsing hard, then that you're fault and you have way bigger problems than just the hp pool being too low
it's just such a lazy fricking way to do it
Then describe an algorithm for Civ 5 or whatever game it is you're complaining about that would perform better.
just let chat GPT play your game for a few days so that it learns how to beat 99% of players
then put that into the game
This but unironically. Surely there is a publicly-available AI that you can use this way, right?
Or maybe have the Mongols actually fricking come at me? Maybe not have Zakarov or Lal just leave their bases unguarded next to my speeder units?
You got a point anon even if it is not systematic.
Boring games get more boring when you extend encounters/give more health or stuff to ai. It is a simple as that.
If you repeat 3 same braindead moves againt an enemy, of course it will be frustrating if he was stronger.
If game give many strategy/tactics and different ways to approach things, it would be fun to take on someone who is cheating.
People like you disgust me.
You're the same type of people that would defend bots in games having infinite [resources], be it fuel, ammo or whatever else.
Or worse, AIs that just magically know the whole map inside out and know exactly where to go.
>AIs that just magically know the whole map inside out and know exactly where to go
Why don't you? Not enough practice?
There is no reason a very rudimentary learning system can't be made.
It's extremely fricking easy, and more so today than ever, systems have plenty of memory to play with for even more advanced and finegrained learning.
No, don't come out with "duuurrr u expect GPT" bullshit, things like this were done in games before where AI didn't have perfect understanding of their (dynamically changing) environment and had to adapt to it.
Lazy ass AI devs aren't an excuse.
It's not hard to do something as simple as giving them a few virtual eyes and a basic tile-based memory of where things are in the map, each having a decay rate so they forget areas not travelled regularly, and a desire to stay away from the most active areas unless it has certain resources for example. (this can apply to all types of games, be it RTS, racing or FPS)
People make shit like that for fun to try speed-run games automatically, but people seem to get it in their heads that it needs a top-tier GPU and thousands of hours to do that shit. No, that's for fricking thousands of iterations running thousands of times per second usually, not mere dozens of units running in realtime!
Having an AI that can learn alongside you can make vastly superior bots and a more enjoyable experience. (and having the ability to delete their memory to start over, likewise, if you want a fresh start / haven't played in a long time)
Bulletspongeitis is no excuse to be lazy with something that doesn't even take all that much effort to do, especially for a typical gamedev with their skillsets.
Simple AI like this isn't that hard, you're literally tile-counting in an array and assigning an emotion to the tile based on how active it was / how often their assets died in the area / what utilities or resources are there, etc.
do it urself then if its so ez
I'm not making an RTS game.
Although I would like to, if I could come up with an engaging enough idea.
Ideally I'd love to make a galaxy-scale grand strat in the Stargate universe, but that's an IP nightmare, and now Amazon own it.
It's criminal how little good games are set in that universe. Fricking MGM.
You don't have to make an RTS, just make a proof of concept or an AI that can play an existing RTS. Doesn't have to be an RTS. You can use Endless Space 2's AI editor and see if you can make something decent out of that game's AI.
Why? Why repeat things that have been done dozens of times by others?
There are videos all over Youtube of people making bots capable of learning entirely new levels from scratch.
Some of them are, admittedly, horribly brute force shitty genetic algorithm crap where they die all the fricking time, others are specifically geared with knowledge on how to play the damn game they are playing instead of wasting the time re-inventing the wheel.
Generally speaking these bots are designed to be run in the thousands in parallel to brute-force speedruns, but I know there are bots that learn level layouts but I couldn't find you them even if I tried because Youtube search has been so fundamentally broken for the past 2 years it is impossible to find anything even if you know most of the fricking title. (and sometimes it literally won't even show you if you DO know the full title)
You're acting like this is some sort of super duper turbo hard AI. It's literally adding weights to tiles and weights to the relationships between those tiles, and that information passed on top of its usual pathfinding algo to influence exactly where it will go instead of just giving it exact knowledge of a map and where everything on it is.
The reason devs don't do it is lazy bastards.
Only a handful of devs care to give their AI dynamic verbs and limited memories.
Hell man, loads of roguelikes give YOU memory - as in, memory of where you have explored on the world.
If it's so easy why aren't there any strategy games that have it and why won't you be the first?
Wasn't Google showing off it's AI that could play Starcraft at the level of a pro years ago? Whatever happened to that?
They managed to make AI that could play chess better than the best humans and now the cheapest mobile phones can beat the best humans. Why can't the same be done for other games?
Because chess, like many other tabletop games with fixed moves, is ultimately a scam. There are only so many positions and all the best ones are now known. In even a simple 4x you don't only need to consider the game board and pieces but each piece can move different numbers of spaces as it feels rather than being locked in and can have different stats other than movement speed. You also have to worry about an economy and maybe even stupid shit like morale. If it's not single units, you have to worry about overall army composition as well. Also each side moves all its units at once instead of one per turn.
Simple games are simple to make good AI for even though if it's too simple it runs the risk of being solvable. I think this game is simple and has decent AI https://www.falcom.co.jp/vantage/index_e.html although most of it comes from maps being weighted against you.
Going on a tangent, I think the pursuit of a good battle AI is pointless. AI is made by a human, so you're just fighting a human that doesn't adapt. If you're good you'll win if the enemy is a human or an AI, which is why handicaps exist even in real life. An AI that can make good and responsive or planned armies (a lot of AI can do this but rarely get the chance), AI that can handle economies, research, building stuff properly and managing its non battle aspects, an AI that is good at diplomacy and working together on coordinated actions. That's what should be being pursued. Ultimately
doesn't come from a bad battle AI but from the AI being unable to manage its resources or unable to go all out, so they get boosts to simulate what they could do if they were able to manage themselves properly.
>Whatever happened to that
Just like with openai in dota it was a PR move to get investments/get acquired by a big company which they successfully did. Videogames were just a means to an end
You have unfair advantages of your own. Might as well cut the machine some slack.
You don't get it
there's very little fun in beating an opponent that you know is cheating
it can be fun for the first or maybe second time
after that you figure out all the gamey mechanics, all the blind spots AI has that you can exploit and so on and the game is exactly the same every time
Sounds like a 'you' problem. Anons use the cheesiest strats against AI over and over and then wonder why they're not having fun. Then they use the exact same stuff on a human opponent - who can actually get buttmad over it, ragequit and cry about it being OP on Di*cord - and they enjoy it because it was a fellow fleshbag who smashes his keyboard in anger and not a machine that can't get mad no matter how much of a dick you are.
Forget boosting AI's proficiency, just hook it up to ChatGPT or something to have the computer talk shit about your mom whenever it sees you wall off its resources 2 minutes into the game. That way players will get a kick out of it regardless of the artificial opponent's ineptitude.
This is why I play SP games on mostly normal-hard. Above that the AI is obviously cheating and that pushes you into turtling slogs or weird gamey strategies until you optimise fun out of your game. I'd just use whatever I find fun even if it's not optimal and roleplay. For real fair contest of skill you play MP games and it can be fun too, but it's a different thing.
Good AI takes time. You don't form plans in 5 seconds and neither will the AI. If they made a good AI, then you would complain it takes 15 minutes per turn.
>You don't form plans in 5 seconds
I don't?
computers aren't that fricking slow
Oh our resident computer scientist, please tell us more about the calculation times for Deepmind style AIs in games far more complex than Chess or Go
The kinds of computations that are necessary for good decision-making in strategy games does not entail "deepmind-style AI", and is not necessarily anything that would be expensive to compute. Computers are much faster at a variety of tasks compared to humans, and it is really not appropriate to assume that outplaying even a skilled player at a 4X game would require a lot of processing time for a consumer PC.
Eagerly awaiting the low-effort insult that I already know you're programmed to respond with.
Dota and Civilization are incredibly difficult for an AI to solve. Civilization beyond its reliance on hidden information requires insanely long-term thinking, even more so than go, while also being build on a significantly more expansive game than both Starcraft/Go. There's just so much stuff, so many units, trees, policies, etc.
Dota is complex in the same way Starcraft is (hidden info, continuous space, real time), but also has a much larger amount of stuff in the game than starcraft does. The endless varieties in heroes, abilities and items make generalization really really hard. Especially since these factors also massively influence long-term decision making. But what may make Dota the hardest game, is that you're supposed to work with teammates, and communication with teammates (verbal/nonverbal) is absolutely essential to winning. The AI would need to be able to not just handle the enemy team, but would also need to be able to read its teammembers, which may require general intelligence.
Every time someone posts about moronicly huge ai cheats I'm reminded of the stainless steel mod where the ai gets so much cheats the best strategy is to not attack anyone for the first 20 or so turns so they can cheat their way to a maxxed settlement because it would take you 80+ turns to grow it instead so the faster you attack the ai the weaker you will be later when you have to fight multiple stacks of endgame units with unbreakable morale and other huge buffs if they're led by the king or his heir.
and this is on normal.
i dont get how this is still practice, its outdated as frick approach, we are in the age of ai, we have genius chess ai since 30 years, u cant tell me its that hard to make a competent ai in a videogame
it is,
the genius NN ai or chess ai's are made by turbo autist computer scientists, they are VERY difficult and expensive to make. Game devs are just average coders, none of them have anywhere near the expertise required to make an ai like that.
but im fairly certain even old games had good ai, like there is this old story of homm3 having such a god tier ai they had to dumb it down
>this old story of homm3 having such a god tier ai they had to dumb it down
Myth
No that another game
If that was true, they could've just kept it on the highest difficulty instead of just giving it more resources and taking away all your resources
It is, but players won't be thankful for it.
It's extremely easy to make an AI that can consistently beat humans without "cheating" (stuff like extra resources, stronger units, perfect vision through fog of war, input reading). The problem is doing it in a way that's interesting rather than the AI simply have flawless and unrelenting execution.
i wan an actual intelligent ai that can learn from playing against me and maybe end up reading and predicting me. that would be cool. no cheating shit, same stats, resources, fog of war etc as me.
>Wasting countless game dev hours trying to get a best-in-class AI to perform around the level of a human because you're too scared to hit "Multiplayer".
this is a good argument for RTS but not for 4x games that take dozens of hours to complete
Lol. This is a complete lie peddled by Google´s and Elon Musk´s ai teams that couldn´t admit they wasted millions of dollars to create ai´s in dota and starcraft that ended up getting beaten by casual players. There is simply no truth to it and that´s why I also know how far they must be from creating anything actually resembling general intelligence.
The fact is that the dota ai could only play 17 out of the 124 heroes and when they went online with it, it still got beat 10 times in a row by the same casual players. In a row. With only 17 heroes to choose for both the ai and humans. That´s pathetic.
The same happened with Google in Starcraft. They made the ai´s, they had several, available to play at some starcraft convention. It got beaten constantly by causual players. Since starcraft requires more mechanics then dota, that ai had an easier time brute forcing it´s way to some victories. Online it became grandmaster, playing against humans. That´s pretty good. On the other hand you can go on youtube and find players not good enough to be pro´s that do moron runs like.
"getting to grandmaster on only one base"
"getting to grandmaster using only one unit"
So while they did get to grandmaster, it´s not a bigger deal than really good players can do it for fun.
>Its not that we're too "casual", its that cheating buff suck all the fun and are a sign of low effort from the dev's part.
It´s not low effort. It´s just not possible to do it. If they could, Google and Elon Musk´s ai teams would have had better results, they were aiming to get ai´s that were just as superior as the one in chess and they failed. They could have tried in a Civ game, that way you take away the ai´s superhuman reflexes. It would only have been a bigger disaster and Civ casuals would laugh when they found out, that the super duper ai can´t play random maps and it still loses.
People need to stop drinking the ai koolaide.
>It´s just not possible to do it
i'm going to interject here.
the AI doesn't have to be better than decent players to be considered good.
it can be given a bunch of strategies to do.
like cavalry in prophecy of pendor mod for warband, they always go for a flanking attack or hammer and anvil and aims for archers instead of just charging head on like in the base game.
It's not hard to beat the player as long as options are limited.
Lets take an rts like starcraft. You could extremely easily make an ai that could stomp all but pro players because there is an optimized build order for most strategies, you simply feed this to the AI and it excutes it flawlessly. There is also unit micro to give you the advantage in fights, even pro players use groups which are inefficient because it's physically impossible for anyone to control every single unit individually. AI could do it though.
But it wouldn't be fun.
numod moment
not like the white & based crimeamod (don't play if you're a snowflake)
shit mod
An 'AI' opponent can't actually think, you know. It only follows the logic the devs gave it and they can't hope to anticipate the crazy tactics players can come up with. So don't treat the computer as an opponent because it can't be one.
But even though it's not smart, it can be entertaining. It has it's advantages: it's always ready for a game, it never gets tired or pissed, you can customize it for a relaxing compstomp or a brutal slog against multiple foes, it relentlessly executes multiple actions at the same time, it won't tell you to uninstall the game and kys for being a noob.
But because it's literally moronic, it needs a handicap. In fact, some of the most fun SP games are designed to be asymmetric in the first place, so that you could use your brains against the enemy's superior brawn.
Don't forget:
>80% of the gameplay mechanics cease being viable at all, forcing the same few boring playstyles
I feel like we've actually seen AI stagnate or even decline across the entire gaming industry. Half-Life and F.E.A.R did their best and nobody has even tried since.
Only modders are trying
the bold quote is entirely correct, civ v/vi's target demographic just wants a punching bag
>the bold quote is entirely correct, civ v/vi's target demographic just wants a punching bag
thats what we fell in love with
>maybe
What a horse shit.
Haven't met a single person, ever, who'd go out of his way and claim - "wow the A.I. is so fricking good in this game....".
Hearing the opposite quite often though.
What purpose do these rat articles even serve? Are they for investors?
Not a lot of people enjoy losing to a computer, most will just make excuses how its the games fault
It's because all the effort is in the image reading and generation industry.
Good enemy AI is easy in fps and other action games. Enemies live for few seconds, are preset in locations and don't have to keep up with player and develop alongside him, using these same game mechanics as him. There is no need for them to plan long term strategy, make interesting maneuvers during battle and develop their cities. Actually you can easily make them annoyingly superior to player with perfect aim
>enemies don't have to use the same game mechanics as him
This is what SP-focused strategy games already do, actually. You have enemies who play a completely different game than the player and don't follow the same rules and it's okay as long as it's challenging and fun to fight against them.
Everyone knows, for instance, that campaign scenarios in RTSes often have the player start with a rudimentary base while the enemy has several bases, multiple layers of defence, and tons of starting units that will be let loose after giving you some head start, so you can't exactly employ MP strategies like rushing.
Go play checkers or some other tabletop game the AI can actually solve mathematically if you want your ass beaten every single time.
How did you get that from my post you moronic Black? I want to play against an AI that simulates human players, not one that cheats and receives insane buffs to compensate for beign dumb as bricks. Thats never fun.
Not a strategy game but another example that comes to mind is Mount and Blade: Warband. The AI lords dont have to play by the same recruitment rules as the player, they can just magically spawn new fully-trained units after a while. Whereas human players have to go manually from town to town and village to village to collect recruits and slowly train the into competent units. This doesnt make the game particularly hard but it sucks the fun out of victories to know that they dont have to jump through the same hoops as you after they get defeated
>but it sucks the fun out of victories to know that they dont have to jump through the same hoops as you after they get defeated
Aren't you able to automate recruitment once you get a castle?
are there any game's that DON'T do that?
>game's
frick me how did that end up there
This is possibly the most low IQ post I've ever seen on /vst/.
If you have done 5 minutes of googling on AIs and difficulties and game design, you would know that (you) as a player don't know what you fricking want. "Good AI" is just smoke and mirrors on the part of the developer and it's not even actually good, it just pretends to be good, but actually is just bad enough so that the player can still win. For example, F.E.A.R. AI was praised for its tactical genius when in fact it was mostly a bunch of radio chatter screaming "I'm flanking them" or whatever, so it seemed smart. It's like when everyone complained that music shuffle wasn't random when it was in fact completely random, so now it's trying to predict how to make you feel like it's random.
The problem with many /vst/ genres is that they try to make AI do the same exact stuff as the player, so there's not really a good way to hide the tricks. AI war is a good example of how asymmetry can be used to make AI seem much better. The dev also posted a bunch of dev diaries on the topic, I highly recommend it.
Is a good AI one that plays like a metagay or one that plays like it's actually acting out its role?(e.g. playing like it's actually a horde of orcs or Britain during WW2 and making decisions appropriate for that even if they aren't the best ones.)
I don't understand why we can't just have some Ayyyyyiiiiiii playing civ against itself and players, continually improving.
We don't really want *good* AI, we just want something that challenges us enough that we feel smart when we beat it. But, as others have pointed out, even that is much easier said than done.
The issue with RTS AI isn't that they are dumb or incapable of complexity, the issue is that they are given the most basic set of instructions to follow so the devs don't spend a frickload of resources testing, debugging and refactoring the code for it so neither your computer nor your internet shit the bed because of its sheer size
A RTS AI will never be programmed to take into account what faction it is playing, what strategy and goal it can go toward, how to manage their economy to achieve that goal efficiently, how to position and attack the enemies, how to micro different fronts in a smart way, how to adapt to the enemies producing counter units, so on and so forth
Only God know how much i want to witness an AI spitroasting me with a frontal assault, a flanking attack and a hit'n'run on the worker units at the same time, it would be a painful AI to deal with if you have tunnel vision or only swear to the almighty singular unit blob of doom, but it would leave one hell of an impression in return
Did anyone ever try to make a reinforcement learning AI that got trained by playing against decent players or mimicking their actions? I imagine something like this would probably completely frick up 99.9999% of the user base but it's better than cheating.
I mean cheating makes things harder for sure but the AI is still dumb, I'm not way a good player but for civ for example the main thing you need to do vs diety is knowing how to please them all while rushing shit like cavalry or cuirassiers because even diety AI can't deal with a min-maxing human player
It wouldn't fix the fundamental problem of AI being unable to adapt. Even if they used top meta tactics, if it used the same ones over and over, anyone would eventually become able to beat it.
>If
Anon, go watch a game. If your matches are turning into only that, you're uninventive.
The only game I've heard of that did that was Quake which wasn't a strategy game, I don't know if a strategy game would be too complex to have a similar ai.
>I don't know if a strategy game would be too complex to have a similar ai.
If anything it would be easier. It's way more complex for AIs to learn movement and positioning, than juggling a few variables, which is what most strategy games boil down to. Not to mention, Quake is real time, while many strategy games are turn based so it has more time to process through everything.
dota2, sc2, minecraft (if that one can count).
I think WARNO does that. I don't know for sure, but the AI unironically uses meme strats so that is my assumption.
I want to honk honk her
>her
Oh nonono!
as long as the makeup stays on
>Higher difficulty improves the AI but also lets it cheat with buffs
>Higher difficulty
>/vst/ gets filtered by it
non-issue if you gain proportionally from enemy killed.
stop playing against ai and play mp. problem solved.
i wanna frick that clown
>higher difficulty doesn't affect the AI at all
>it just slaps you with a bunch of penalties
These fellows are working on something that doesn't cheat.
>its just the AI cheating with insane buffs
this is what I always hated the most in all kinds of strategy games
in modern times it had only become worse since most developers don't care about making good AI opponents
Iron Harvest is probably the worst offender I still got in my recent memory
>easy
is just AI playing with the same things available to the player
>medium and hard
are just options of "how much do you want the AI opponent to cheat"
There's no good reason to play RTS on normal or higher you're better off playing it on easy
vst is full of shitters. They cry about not having an AI that can't challenge them but don't play multiplayer...
>multiplayer competitive games
>it's just rock paper scissors with step by step tryhard guides
>if you don't play this specific way you'll always lose
lol
The same applies to every competitive event, including irl sports where everything is solved for optimum output: techniques, equipment, training, even diet.
Yeah MP is just a confidence trick where you scam less capable players into matches you know they will lose.
truly the words of someone who's never been competitive at anything adversarial in his life
please don't speak on this topic again, you are clueless
>higher difficulty
>its has enemy difficulty multilied by 4/5
>now a guardsman easily raping you terminator squad with his bayonet.
This is why difficulty sliders shouldn't exist in games to begin with, because attempts to make the game "more challenging" just involve lazy design of artificially inflating hp and damage numbers, instead of building encounters specifically to be a rewarding obstacle. This is something that Dark Souls managed to actually do right, in that it told difficulty sliders to frick off and made the game an actual experience.
Age of Empires 2 DE AI had been re-written to follow some common online strategies and have 0 cheats, the difficulty is mostly hard-coded with the amount of economy they're allowed to use + some higher tier strategies locked behind the final difficulty.
It's not perfect and still folds to rushes but it offers a decent challenge in the mid and especially late game.
Much more fun than
>The AI straight up just gets free shit
I like the difficulty when they put more enemies on the screen rather than the same number of enemies with just more HP and more damage.
I'm starting to think this genre is dead cause of the difficulty curve and the game throws at you whatever it feels like if you try to play it at harder difficulty despite learning all the basics. There's a reason why MOBA exists which was conceived from Warcraft 3 cause the younger generations, the zoomers are too intimidated try out RTS.
Tell that for the 99% of people who letf the multiplayer of Starcraft 2 because it only consist of wich player is the fastest to produce basic units and destroy the enemy base in the first 3 minutes of the game.
AKA "fun gameplay".
>AKA why koreans should learn of why others sports like soccer, basketball, etc, got so global.
Not sorry for any shitty Starcraft 2 elibreasts player that got offended btw.
You can't just wing it in an RTS unlike other online games. At the very least you need a general idea of what you're trying to do and know the precise steps to make it happen. You also need to hurry the hell up while keeping track of all the variables.