AI vs Players and Cheating: The Irony

[Lestrad]

Has anyone else ever noticed that the very conduct that would generally earn a player a VAC/Perma Ban is generally lauded when conducted by an AI opponent, simply because the developer prefaces it by saying "Brutal".

Even more to the point, many RTS or 4x games have modders go out of their way to create so-called "improved" AI. But is it really? Or does it simply cheat harder and more often?

When you consider what AI actually entails (Watson, Deep Blue, etc) most automated opponents are not actually intelligent, they receive bonuses to damage reduction, or hit points, or spawn free units if they ever have less than the player, etc. But they've never been intelligent.

So why is a false sense of difficulty lauded, as "brilliant" and "challenging", when a player committing those same actions gets banned? When are we going to start expecting more than inflating difficulty via cheats and bonuses, instead of having the AI operating on all the same limitations the player is, but have them actually play and adapt smarter?

[qb]

ai can cheat but it will not selectively cheat in most cases even aimboted ai will probably find its demise being stuck in a wall or smth, cheating players however will find continously ways to cheat even if it makes "life on the server miserable of anyone else on the server", ai in that sense causes somewhat of a problem but its probably not as bad, and most of the time its totally under control of its creator, so we always hope its going to be fixed sooner or later, people usually are not that easy to bargain with, even vac and other things not really fully fix the problem, in short ai is not good enough even with cheats to pose any significant problem compared cheating players.

[Morkonan]

Първоначално публикувано от Lestrad:

...
So why is a false sense of difficulty lauded, as "brilliant" and "challenging", when a player committing those same actions gets banned? When are we going to start expecting more than inflating difficulty via cheats and bonuses, instead of having the AI operating on all the same limitations the player is, but have them actually play and adapt smarter?

When a bot gets bonuses or abilities a competing player does not get, it's because it isn't possible to produce good competitive gameplay against a bot without it. That's likely due to a lack of true or even just competent "intelligence" in the bot's "A.I."

Players, however, are balanced around commonly expected intellect and ability to play the game. To make their gameplay "fun," bots need an edge... even against bad players.

"A.I." isn't "intelligent" in this context. It's not easily possible to have an A.I. "adapt" to all the novel situations a player is capable of orchestrating in a few minutes of competitive match-play like you're describing. AI is just not capable of the powers of reasoning you're assuming, here.

One day it might be, but it may also exceed human capability in gaming and then we're the ones that will need bonuses. :)

[Hobbit XIII]

Първоначално публикувано от Lestrad:

So why is a false sense of difficulty lauded, as "brilliant" and "challenging", when a player committing those same actions gets banned? When are we going to start expecting more than inflating difficulty via cheats and bonuses, instead of having the AI operating on all the same limitations the player is, but have them actually play and adapt smarter?

On improving a player vs bot game, I'll use close combat 5 and Unreal Tournament as examples .

Close combat 5 I would add into the code to predict what the player did based on past peformances on that map.
I wont write any real code but a concept :

CC5 Beach Landing map after the battle the game records positioning of the players assault team lets say it's goal is a hedge beyond the sea wall.
So next time I would make a roll for the machine gunner 42 to be a couple of hedges away from that location. The roll would be which of these hedges for difficulty.

If the player did same move next time it played that map then put a stronger probability to put the MG42 directly where the assault team would be trying to get to.

Third time I would have mortars firing in that area as well as having the MG42 waiting in ambush.

Without any real use of AI.

Unreal Tournament I would start off by making the error of shots from the bots quite high.
Then I would either add predictive shooting based with high error allowed then gradually narrow that error range as game went on.
If I really wanted to be a git made famous by MGS is let the bot know what the player was going to do and react to it almost instantly - the famous work around was for the player to switch ports for the controller.

Lots of examples I am sure this topic will be filled with. That is just a couple I would do as a starter.

[Stranger]

Първоначално публикувано от Hobbit XIII:

Първоначално публикувано от Lestrad:

So why is a false sense of difficulty lauded, as "brilliant" and "challenging", when a player committing those same actions gets banned? When are we going to start expecting more than inflating difficulty via cheats and bonuses, instead of having the AI operating on all the same limitations the player is, but have them actually play and adapt smarter?

On improving a player vs bot game, I'll use close combat 5 and Unreal Tournament as examples .

Close combat 5 I would add into the code to predict what the player did based on past peformances on that map.
I wont write any real code but a concept :

CC5 Beach Landing map after the battle the game records positioning of the players assault team lets say it's goal is a hedge beyond the sea wall.
So next time I would make a roll for the machine gunner 42 to be a couple of hedges away from that location. The roll would be which of these hedges for difficulty.

If the player did same move next time it played that map then put a stronger probability to put the MG42 directly where the assault team would be trying to get to.

Third time I would have mortars firing in that area as well as having the MG42 waiting in ambush.

Without any real use of AI.

so you'd rig the RNG with an omniscient AI, and frustrate players out of playing the way they want to by creating adaptive pressure against everything they do.

That's literally torture.

[Hobbit XIII]

Първоначално публикувано от Stranger:

Първоначално публикувано от Hobbit XIII:

On improving a player vs bot game, I'll use close combat 5 and Unreal Tournament as examples .

Close combat 5 I would add into the code to predict what the player did based on past peformances on that map.
I wont write any real code but a concept :

CC5 Beach Landing map after the battle the game records positioning of the players assault team lets say it's goal is a hedge beyond the sea wall.
So next time I would make a roll for the machine gunner 42 to be a couple of hedges away from that location. The roll would be which of these hedges for difficulty.

If the player did same move next time it played that map then put a stronger probability to put the MG42 directly where the assault team would be trying to get to.

Third time I would have mortars firing in that area as well as having the MG42 waiting in ambush.

Without any real use of AI.

so you'd rig the RNG with an omniscient AI, and frustrate players out of playing the way they want to by creating adaptive pressure against everything they do.

That's literally torture.

Oh that is so judgemental!
What you call torture I call family entertainment.

No, meaning if the player chose a progressively harder game they could and that is what I would do as one of the basic and easily coded moves.

If I play a game where I want to win a lot more than lost then I would make simple restrictions like limit ammunition for primary weapons for the bots and make them a little more likely to panic or surrender for arguments sake lets say my guys have a number 50 tolerance and the bots have a number 10 tolerance.
Things like that which is already widespread in games.

Example :
I got 100 soldiers
Enemy has 1000 soldiers
I want to make lots of enemy casualties, so reduce there capability but then about mission 5 there last 50 soldiers are going to be die hard elites and the remains of my guys are going to be levelled up tough guys and now the real battle begins.

Add into the predictive play you could get some pretty interesting results which could also be exploited by the player themselves and that is where the hidden reserve enemy force somes into play. hee hee hee.

same with the FPS if a player really wants to improve their reflex they can make it improve and adapt to the player's skill gently nudging them so after a couple hours play they have improved without being massacred.
Or simply increase the speed of the bots improvement pending requirement.

[Stranger]

it might work for you, but this sounds like stuff people would hate.

have you made a game with these systems? you might find out you hated it yourself.

[Uncle Sam]

Eventually AI will adapt on the fly to what an human player does, atm it can go and win against DOTA 2 champions.
https://www.youtube.com/watch?v=tfb6aEUMC04

[Lestrad]

Първоначално публикувано от Morkonan:

When a bot gets bonuses or abilities a competing player does not get, it's because it isn't possible to produce good competitive gameplay against a bot without it. That's likely due to a lack of true or even just competent "intelligence" in the bot's "A.I."

But that just goes back to my point. When are we going to start saying "do better"? Sure, true AI requires literal super computers to function (Deep Blue for instance), but when extremely cheaty AI begins to negatively effect gameplay, it's never called out.

Take for example Total War. You take out an enemy army and suddenly, they have a entirely new army full to capacity, immune to attrition and upkeep. An army you can look at and know EXACTLY how many turns it should have taken for them to field it. This was a huge problem in Shogun 2, and that was playing on "Easy".

I definitely believe there are steps developers could make to provide a challenge yet still make the game fair. A good example would probably be AI in the much maligned (and generally unheard of) RTS War of the Ring. The AI would adapt to your defenses. Build a guard tower they'd attack from a different direction. If that wasn't possible, they'd attempt to flank around it just outside your visibility. If that wasn't possible, then only then would they try to just overwhelm it. You don't see that kind of adaptation in most RTS games.

Overall, I would argue that if a player can identify exactly HOW the AI is cheating (without having to go into the lines of code), it's not done well. If a player can say "they have to be cheating...but how" THEN you've done a better job.

[Devsman]

Well, actually complex AI is not always a good thing. It depends on how the game is designed. You have two broad categories when it comes to challenges in video games: obstacles and opponents.

If the game is designed as player vs opponent(s), then more complex AI might work. But on the other hand, many things that look like opponents are actually obstacles. Like a goomba, for instance. Kind of obvious example. It's a living thing and it wants to eff you up, but that's just polish. Mechanically, it's just a damage zone that moves across the ground. An obstacle. Mario would suck if the goombas were smart.

And most of your single player genres have enemies that are all goombas. Shooters have goombas with guns that randomly decide whether to run for cover. Racers have goombas that follow the track. Fighters and strategy? Maybe not so much. Because those are designed as player vs opponent(s).

And yeah... I've played a game with smart goombas before. It was cool for about ten minutes, then it was frustrating as hell. That game was Metal Gear Solid 2. They would find out if you killed somebody. They would injure you and follow your trail of blood if you tried to hide. They would split up and surround you. They would search (almost) every hiding spot there was before resuming patrol. Not even its own sequels, some of which released more than a decade later, attempted such complex AI again. Because it was infuriating.

[Capt Spack Jarrow]

AI is limited by the ability of it's creator, AI cannot ever be intelligent, it is programmed.

[Lestrad]

Първоначално публикувано от Devsman:

Well, actually complex AI is not always a good thing.

True, but I'm not arguing for "release the kraken", completely unshackled. What I'm calling out is when it's blatantly obvious that "this is how they're cheating, and here's the video evidence".

Until the day comes when Cray super computers are in every home (metaphorically speaking), we're never going to have actual artificial intelligence.

Especially with all the loony's holding CEO positions or public office who watched Terminator once and decided that it's science fact.

[Stranger]

Първоначално публикувано от Lestrad:

Първоначално публикувано от Devsman:

Well, actually complex AI is not always a good thing.

True, but I'm not arguing for "release the kraken", completely unshackled. What I'm calling out is when it's blatantly obvious that "this is how they're cheating, and here's the video evidence".

Well the point of games is to frustrate players; little known fact.

[Hammer Of Evil]

nice thread attempting to minimize cheating in video games.

[Hobbit XIII]

Първоначално публикувано от Stranger:

it might work for you, but this sounds like stuff people would hate.

have you made a game with these systems? you might find out you hated it yourself.

You ignore that it is all optional when I was writing I am sure I stated this.

And yes I make many programs its what I do testing and theorizing and building. Forums are a way to work on concepts without the over focus.

[Stranger]

Първоначално публикувано от Hobbit XIII:

Първоначално публикувано от Stranger:

it might work for you, but this sounds like stuff people would hate.

have you made a game with these systems? you might find out you hated it yourself.

You ignore that it is all optional when I was writing I am sure I stated this.

And yes I make many programs its what I do testing and theorizing and building. Forums are a way to work on concepts without the over focus.

Ok.