Define "cheating"
I'm interested in what an AI can get away with before players call it bullshit, but still be able to compete with a player savescumming, powergaming, etc. If a game has a random chance of an outcome from an action, is letting the AI re-roll it a couple of times and take the "best" choice cheating? Is just giving it an item it would be best with cheating? If there's some sort of decisions to be made and they give various rewards but you have to do something (e.g. kill some bandits) would it be cheating to just give the AI the rewards without it making the decisions?(especially think Endless Space 2, would it be cheating to give the AI all faction quest rewards on, say, turn 20 instead of any decision or task completion) Would it be cheating to let an AI see the area around any space its unit could end up at if it moved so it could simulate a player ending up next to or ambushed by a powerful enemy and reloading, or so the player wouldn't think the AI happened to move their weak army next to their strong one stupidly when the AI actually couldn't see the player's army was there to begin with. Would it be cheating if, given a game with a random generated map, on harder modes more resource nodes were placed nearby the AI faction's starting area instead of giving it boosted income or spawning units?
 It's All Fucked Shirt $22.14 |
 DMT Has Friends For Me Shirt $21.68 |
It's All Fucked Shirt $22.14 |
A games underlying systems should be complex enough, and interact with one another seamlessly enough, while allowing the player/s and the AI access too and influence over the emergent data logically enough to render cheating entirely unnecessary.
Cheating AI is the reserve of the lazy or incompetent coder.
But is it necessary that the AI actually interact with the complex systems instead of just receiving the benefits it would anyway without wasting any effort on having it choose solutions the player wouldn't even see it choose?
This is why the gamemaster AI of Alien was so interesting at the time and still is. You could effectively block/unlock logic pathways based on how the player interacted with the AI throughout the game but also give small hints to the AI about player location or generate how often certain weapons are used. IF you for example took a strategy game and had an AI profile built off your playstyle you could effectively do the same thing .
Not sure how viable or expensive it'd be but having a perfect AI to fight isn't fun so negotiating how often it gets these abilities or if the AI has to follow the same ruleset would be interesting to see.
?t=106
Also going off what I said RTS side of AI IMO you could gate keep actions per minute, I know AOE4 does something similar hence why it sometimes seem un-reactive but this has its own problems such as an AI's inability to act to player movements to a degree.
Writing an AI which successfully fakes interacting with a complex system is at least as difficult as writing an AI which actually interacts with the system.
If you simplify down the interactions to save yourself time, you leave holes where something can happen which doesn't impact the AI as expected. Then the player notices your AI isn't actually following the same rules they are.
A better solution for adding a simple AI that is still challenging is to explicitly have it running off completely different rules and goals from the player. AI War, for example.
>Define "cheating"
No
That being said, can anyone explain me why using macros is cheating?
Some units are balanced around APM limits. Macros disrupt this.
In a turn based game, or something that does not actually account for reaction times, macros are not cheating. In a game where your reaction time and the speed of your actions does actually matter, like an RTS, macros give you an unfair advantage because you can do way more things than a normal person could or they let you do things you personally cannot. Think of a fighting game where a macro lets you do a frame perfect combo you cannot normally do because you suck. It's cheating because it lets you something that even pros drop from time to time.
I guess people would call it cheating when the AI has knowledge of the entire map, its resources, including potential future ones, and knowledge of the disposition of all my forces, despite fog of war
I call that fun
look at that hairline jesus christ
I don't want to cheat, and I don't want them to cheat. I want the exact same rules, advantages, and restrictions on both sides.
If I can't have that, I'm not playing with your garbage AI.
What about AI that can savescum, look up strategy guides, and use the nastiest cheese against you? Same rules, advantages and restrictions on both sides.
>ai that can savescum
How would that work? It crashes your game if you beat it?
>Be me.
>Playing C&C Tiberian Alert 9
>About to destroy the Allied base with my Nod Kirovs.
>Half of base is gone, got this game in the bag.
>Loadscreen.jpg
>WTF, I didn't do anything.
>Suddenly back at the half way point of the match, I didn't save and the game don't have auto saves.
>AI use a different strategy and take out my warfactory that was 98% on making the priest that would convert their tesla tanks and win me the battle.
>mfw
It would be interesting to have a "non-cheating" AI that kept its saves differently from players whenever the game was saved so that when players reloaded it'd still have all its prior knowledge. It'd probably get players a few times until they figured out how to break it.
Actually now that I think about it, MTG2012's AI has been accused by a couple of my friends of willfully crashing the game or refusing to finish turns via endless loops if it can't win, but can get the right conditions to not let you win.
>>ai that can savescum
>How would that work?
Not him, but I gave examples in the op
>Would it be cheating to let an AI see the area around any space its unit could end up at if it moved so it could simulate a player ending up next to or ambushed by a powerful enemy and reloading
>If a game has a random chance of an outcome from an action, is letting the AI re-roll it a couple of times and take the "best" choice cheating? Is just giving it an item it would be best with cheating?
Additionally, assuming it's a game that has some sort of auto battle or battle with random factors (such as damage rolls), if it could simulate a battle with a nearby army of yours and then decide whether to attack or avoid it, would that be the same cheating as if you attacked it, lost or pyrrhically won, and then reloaded? Although you wouldn't be able to tell it fought you before in that case instead of just calculating your power.
Another type of example in the op would be equivalent to generating maps over and over until you got a favorable start
>if, given a game with a random generated map, on harder modes more resource nodes were placed nearby the AI faction's starting area
In that case it's not cheating resources, it just has access to more of them to start, same as if a player had an amazing start.
Also regarding
>look up strategy guides, and use the nastiest cheese against you
Is it cheating for AI to know all unit stats even if they aren't shown to the player? That would be more of an RTS thing than a turn based one probably, since turn based games tend to let you view stat info. With regards to RTS and games with fixed map and fog of war, I don't think it would be cheating for an AI to know the map/be able to see the map in the sense of having the map revealed but not know where enemy units were, and even have certain coded waypoints to investigate based on points of interest and starting points since the player would eventually memorize the map as well.
I don't approve of savescumming on either side, and the AI should be aware of possible strategies, yes.
The player should always have the choice wherever he wants to play fair against the AI or if he wants to cheat/cheese it, the opposite however is not true. If you give AI the ability to cheat you are already alienating most of the playerbase without adding anything in return (the cheater player would have cheated anyway with or without justification). Its baffling this whole concept is alien to you and nu-gamedevs
That's pretty much the issue - humans cheat all the time. Anons deliberately abuse their intimate knowledge of all the AI's shortcomings, as well as beneficial exploits and glitches, and then they wonder why the AI is so easy to beat. And when there's an option to give the AI a bit of handicap for added difficulty, they cry how unfair that is. If the AI had any actual intelligence, it wouldn't put up with any of this bullshit and simply refuse to play against cheaters.
>AI is only easy to beat because of exploits, glitches, and people using intimate knowledge about how the AI functions
Can you name even one game that this is true for?
Civ 3 on the hardest difficulty
>complains about players cheating
>gives an example of an AI that's cheating 100x harder than any player ever has
>Anons deliberately abuse their intimate knowledge of all the AI's shortcomings
Except I don't. The AI just sometimes is beyond moronic for no reason.
Let's take my recent V2 game with GFM as an example. I declared war on Austria for Lombardy, I know I can cut Austria off from their western Italian holdings due to the mountain range, I set up a frontline and am willing to sit on the wargoal and ticker them out for a minor victory because it would be a hard war to win a large victory in, however the mountains are hell and my units are dying of attrition. So I cut my mountain stacks in half to 10k, the Austrian AI sees this and starts moving in, I instantly move in my armies and I quickly reach 80k in the battle, Austria responds by moving in 140k. I call over more to match them, Austria sits in that battle for a month, then I decide if they're going to suicide like that I'm going to stack wipe them, and I casually stroll one cavalry around the alps across 6 unoccupied provinces to complete the encirclement and stackwipe Austria's entire army.
Now, that is a common mountain baiting strategy. However I didn't mountain bait. I entered the war with a plan, cut off the enemy and secure a victory they can't contest. Attrition forced me to drop the units I was using to enforce this frontline and the AI decided to try its luck. Here's where things diverge, in mountain baiting a player keeps their armies in reserve to avoid the AI retreating until they complete the encirclement. I gave the AI plenty of time to retreat by showing my entire hand and sitting in that battle that they were blatantly losing. By the time I completed the encirclement Austria had 60k to my 120k, a battle that even the AI should be capable of seeing is a clear loss even on even terrain. But they didn't retreat and the grace period I gave them to retreat ended. It is just the AI being completely moronic despite me trying to help it not be.
Does the AI in vicky or eu4 even actually retreat? I've seen it take the actual worst fights even to the bitter end even if it could just run away.
It does not know how to play defensively, period.
Wrong, most AI is fricking dogshit and instead of doing a better job at it the devs just give the computer cheating bonuses and call it a day. They honestly can't do better.
CivV AI is literally easier to beat on highest level of difficulty, because players can use its insane cheating gains against them using meta knowledge.
When I can see an whole army just spawn in their base instead of leaving their bases, or shit out buildings with no build time, or when their units take noticably more damage to destroy than mine when we have the exact same units.
I am fine to an extent with them getting more resources and having no fog of war, unless it is wargame, wargame AI is just mad gay.
I guess I'm in the minority when I say that no fog of war for the AI is the most annoying cheat. I can live with decreased production costs/ internal unrest or increased tech, because that all manifests as a nation with more and better troops. That can be difficult, but it just requires strategizing differently. The AI having those buffs is not to dissimilar from playing Belgium and being sat next to France and Germany, you just have to fight as the underdog. But when the enemy is overextended and I move some troops towards their position only to have the AI pull back well before they should have been able to see my reinforcements, its really fricking annoying. No FOW is the most salient form of cheating to me. Sure, give them more space marines, I can deal with that. But when it feels like they have spies at all levels of my administration that I can do nothing about, that infuriates me
>they have spies at all levels of my administration that I can do nothing about
that's their bonus. yours is that you are the human player
Maybe, but like I said its the salience of it that annoys me. Them having more men is fine, but when I start moving my army and they start moving theirs in a way to respond to it well before they should be able to see, it's annoying. Give them material advantages and I can change my material conditions to handle it, but when they can see everything it just punishes me in a way I can't really make up for. The only solution is just accepting that I don't get to attack the enemy in their vulnerable position and that the enemy simply gets to pull back to consolidate around better terrain
>I can deal with the enemy having more units
>I can't deal with the enemy having fewer units but gathering them in one area
I can deal with it, its just annoying. Its not a problem that strategy can overcome, I just have to accept that the computer gets to see everything. If anything, it takes away from strategy and planning
But AIs are often predictable in their movements. For example, you can bait them to commit forces to attack one of your weaker units that you can afford to lose, then counterattack with your good units and come out ahead. You can create temporary holes in your defense to bait their commitment to a specific route or attack.
I understand what you are saying though. Perhaps you are even having more fun than I am by imaging that the AI is a real opponent and are more immersed in the game that way. I can see how an omniscient AI would diminish this immersion and wonder if I too can have this type of fun
True, though I think that baiting into mountain battles is dishonest, if that makes sense. I don't do it since it feels closer to opening the console and typing in a win command
>but still be able to compete with a player savescumming, powergaming,
AI shouldn't be able to deal with powergaming... well at normal difficulties, if the player can powergame so should the AI without "cheating".
>savescumming
The AI shouldn't be able to deal with that, if the player chooses to savescum they should have an advantage, just like everyone IRL would if they had the ability to see the future or rewind time.
a very simple example is just captured by 'rules for me but not for thee'
in fact a good heuristic is to suppose that if the AI were replaced by a human, given what the AI has at its disposal or given its ruleset, would that human have been accused of cheating?
You will NEVER be able to compete with savescumming and it's foolish to try. It exists to give real players a leg up against an AI because players like to win.
Powergaming is solved by just making a good AI. It doesn't need hacks, it needs strategy.
Savescumming is essentially cheating in any strategy game. You only do it when learning the mechanics. Imagine if you are playing multiplayer, ur losing but then can just load up a save and restart. it's not different to console commands where you give yourself 10000 resources.
In nuTotal Wars there is an ability to confederate AI faction so you take over their armies and cities. And it always pisses me off because it reveals that AI factions are completely nonfunctional. When you confederate them suddenly your economy tanks, because AI built cities are inefficient and they couldn't actually afford all those armies.
You can't even reorganize those armies normally because lol general only armies. If you want for that one more elite unit to join your other army, you need to move them alongside all other units and general to your other army.
In a symmetrical game, AI shouldn't cheat and should interact with the games' system the same way a player would.
In an asymmetrical game (like AI War) the AI can cheat within limits, where the limit is the players' fun.
So, for example in AI War, the AI starts controlling all of the map, has nigh infinite resources and units and his units are better than yours, and he still knows where you are if you piss him off hard enough and then try to go dark. But that is fine because it's fun to fight this bullshit in a game where the main thing is this bullshit in the first place.
Meanwhile in (insert most popular RTS here) the AI should, in theory, play like a human player, limited by fog of war, resources, and the skill level setting. In theory, the best AI for these games will be just like fighting a human, where it sends scouts to see what you're doing, you can mindgame it and it can mindgame you, and all that good stuff. But I'm pretty sure there's a Nobel Prize in artificial intelligence research and development if you actually make this ideal AI because at that point you can claim it passes the Turing Test and after that all hell will break loose
Agree. I’d much rather play an entertaining, honestly unsymmetrical game where you have your own advantages and the computer its own, tailored to make the challenge interesting, than a game where AI *pretends* to play the same game as you.
The very worst thing are AI’s that pretend to play the same game as you, but are so inept that they need to be given insane bonuses, ultimately turning the game into something else than what it is supposed to be. Good example being Wargame series: its AI is utterly braindead, doesn’t do basically anything that the game is supposed to be about (scouting, planned pushes to weak spots), only spams endless hordes of tanks fueled by cheat-only unit availability. So in practice, playing against it feels like tower defense, because if you try to play against it ”properly”, it will just overrun you with mass, but by creating obvious defense points on roadside you can just annihilate their rushes without them doing anything to respond to it. That can be brainless fun for a match or two, but god damn it is disappointing if you wanted to play real Wargame with it, or even worse, teach new players the game by using AI as training opponent.
>AI shouldn't cheat and should interact with the games' system the same way a player would.
The premise of the thread is if it would be considered cheating if they interacted with the games' system exactly like a cheating player (reloading, avoiding things it shouldn't have known about, having abnormally good starts and luck) but didn't have any sort of bonus like extra income or free units. What does cheating mean to you?
>Meanwhile in (insert most popular RTS here) the AI should, in theory, play like a human player, limited by fog of war... where it sends scouts to see what you're doing
Is it okay if it knows the map layout and resource placement of a non-random map like someone who played it several times would? Is it okay if it knows the starting areas as well, if they're limited, and sends scouts immediately to those areas? Is it okay if it knows all the unit stats, even if they aren't shown like someone who's trained and read the wikis and uses that in its tactics? Is it okay to do that thing where it can calculate the range of your units and stand outside your defenses/squeeze through a tiny gap a player probably wouldn't notice unless they really knew the ranges of units well? How much micro is cheating?
>Is it okay if it knows the map layout and resource placement of a non-random map like someone who played it several times would? Is it okay if it knows the starting areas as well, if they're limited, and sends scouts immediately to those areas? Is it okay if it knows all the unit stats, even if they aren't shown like someone who's trained and read the wikis and uses that in its tactics? Is it okay to do that thing where it can calculate the range of your units and stand outside your defenses/squeeze through a tiny gap a player probably wouldn't notice unless they really knew the ranges of units well? How much micro is cheating?
None of that is cheating. It's annoying, but those are good behaviours to implement into the higher-level AIs. But a good human player can do all that. They might not be able to do it consistently, and the AI probably shouldn't be able to either or at all when you are playing against the easier difficulties.
If I were to take what you said, and turn it into actual cheating, it would be something like this: It knows the map layout and resource placement on a non-random map, even through the fow. It know exactly what starting area you are in, without scouting you out. It gets some bonus to unit stats to make up for its inability to use them properly. Its units can attack from further than a player normally could. And it can micro units to a degree that it is obviously avoiding shots it couldn't see.
I am fine with an cheating AI that is a bit suicidal and just smashes it's forces against my defences.
AI with Aztec monks
when someone is better than me
The AI is "cheating" any time it doesn't obey the rules of the game as they are enforced on the player. Every strategy game AI must cheat in this sense, the cost/benefit matrix for developers to build the computer player to be proficient using only the input/output systems of the human on the other end is outrageous. Consider that for the AI not to "cheat" it would need to have a limited field of view, some form of decaying and saturate-able memory (eg it could "forget" certain details of things it had "seen"), and a hard limit of between 60 and 150 apm for commands that include moving that field of view around, as well as occasionally putting in false commands as misclicks. None of that is trivial and it's also largely a pointless exercise.
What you're looking for is the game not to rub it in your face. It should not make a big production out of granting the computer player huge quantities of extra resources and allow it to make transparently omniscient decisions based on information it could not ever possibly have obtained any other way than reading the raw game state, if the game relies on fog of war/information asymmetry.