How weak is the AI?

[MkLh]

Anyone who says they are not disappointed in finding out AI doesn't scale past Chiefdom is lying to themselves. The scaling bonuses and boosts have always been one of the most annoying aspects of CIV. Yes it makes the game more difficult, so it achieves the end goal, but it seems cheap and grows old.

I was really hoping with CIV 6 and all the complex systems and changes that they made, that they would finally really scale AI. Hopefully this is a big system upgrade in the first expansion, because it would be really fun to see how well a deity level AI would use all the systems at its disposal.

Only a tiny fraction of the player pool will play on higher levels. It doesn't make sense to develop a superb AI and then dumb it down for a vast majority of players. This is why an AI scaling won't happen soon.

 

[ToothedBomb]

Only a tiny fraction of the player pool will play on higher levels.

I'd play on higher difficulties if the AI didn't receive ridiculous bonuses, effectively turning the game into 'how can I abuse the AI as much as possible."

 

[Lyoncet]

I'd play on higher difficulties if the AI didn't receive ridiculous bonuses, effectively turning the game into 'how can I abuse the AI as much as possible."

Agreed. For me in Civ IV and V, the trick was to find the balance between "too easy" and "too tedious." I expect VI to be no different in that regard.

 

[treadwin]

I'd play on higher difficulties if the AI didn't receive ridiculous bonuses, effectively turning the game into 'how can I abuse the AI as much as possible."

I agree that beating some of the higher AIs turn the game into more of a puzzle game than a strategy game, because you are forced to play a certain way, to exploit any advantage you have available, in order to win.

However I, like other people have said, play the game at the point where I have a good challenge. I'm not good, but I feel that my level is still above a game where the AI is as good as it can be, without bonuses, so I don't really care what they do below that level.

I appreciate that writing AI for Civ is a massive task though. I am not surprised that bonuses are required in order to create a challenge at higher difficulties.

 

[Ovarwa]

Hi,

The difference between playing a good AI versus playing a bad AI that has bonuses and cheats is like the difference between playing a good chess player versus playing a 6yo child who gets to have all his pawns work like queens.

Both games can be fun, especially if you are still trying to figure out the rules to chess yourself, but they are not at all the same thing. Once you know how to play the game, the difference is obvious. You might play versus the child if you're babysitting but less likely if you want a good game of chess.

Anyway,

Ken

 

[Acken]

Everybody would like a good AI. Im not sure why some people feel the need to repeat it would be better than bonuses.
But then expecting a challenge with 0 bonuses is completely delusional for these games unless you re terrible yourself.

It has to go both ways. The base AI has to be good enough especially when the improvements are easy otherwise it looks too foolish. Rhere is a lot of improvements here that can be made. And then it gets challenging through bonuses. Having a good base AI also allows for easier tuning of the bonuses.

Its also the job of the designer to give a good set of bonuses. Its a problem in civ5 when the AI starts so far ahead but cant keep up in the late game. It makes the start frustrating and the end game a foregone conclusion.

As for being stuck in some strategies on higher difficulties I seriously question what people think will change to that problem with a more efficient AI. Its a balancing issue mostly. Its true that some things will become more available like the great library for example but if you think youll be able to pass on rationalism then think again.

 

[Ovarwa]

Everyone recognizes that game AI requires cheats and/or bonuses as game difficulty increases. I'm not sure why some people feel the need to repeat that everything is just fine the way it is.

 

[Selereth]

The fundamental abilities of the AI should be decent enough, with bonuses supplementing that on the higher difficulty settings. For example, no matter what level the AI is, it should be able to move and shoot on the same turn, escort its settlers, make rational trade deals, know how to aim for a particular victory condition, etc.

 

[Ornen]

Anyone who says they are not disappointed in finding out AI doesn't scale past Chiefdom is lying to themselves.

Don’t make generalizations before you check your own naivety. Has a single Civ game ever scaled AI performance with difficulty? The number of games that adjust AI behavior AT ALL based on difficulty is very, very low. For any one you name, I could probably provide a list of 100 games that do not vary AI based on difficulty. Because that's over 99% of games.

Why would a developer purposefully withhold an AIs behavior solely for the sake of reduced challenge? Far, far easier to give out bonuses/penalties than to rewrite the AI for 8 levels of difficulty. I'm ready for the myth that that's on the table to die off, because it distorts any discussion of the AI.

Like nearly all video games, Civ 6 will scale one baseline AI up and down to contend with differently skilled players. We can only hope that AI is competent enough it does not rely wholly on those bonuses at higher difficulties. Anyone expecting otherwise is readying their hopes to be dashed, and anyone who generalizes their overblown expectations on EVERYONE is just being ridiculous.

 

[rpgalon]

Also, the more brain cells the AI has, the more you will have to wait for the game to process the next turn.

Sometimes it's better to make the AI a little bit less complex, and compensate with bonus, just to make the process faster. Don't know if that is the case for Civ VI AI.

 

[Lucius_]

I'm more curious what specific bonuses the AI will have in Civ VI. I know they will play at a lower difficulty level, but what bonuses do they have on top of that? I think it was mentioned they get extra production from tiles. Do they also get a base production bonus on top of the tile bonus and the difficulty level bonus? They did not mention this, but it seems like the AI might start with a bonus warrior and settler. Is that the case at higher levels? Do they start with free techs? Do they have amenity bonuses on top of their difficulty level bonus? In CiV, I always felt like the AI had bonuses on top of bonuses, and it was just too much. I don't mind the AI getting bonuses, but I hate that means there are things I simply cannot do, for example, the player simply can NOT build the Great Library, the Great Lighthouse, etc. No matter what, it can't be done against the AI at high levels (without the most extreme of favorable conditions) because of it's cheats. Makes me feel like those wonders shouldn't even be in the game then.

 

[Acken]

Everyone recognizes that game AI requires cheats and/or bonuses as game difficulty increases.

Lol.

Also, the more brain cells the AI has, the more you will have to wait for the game to process the next turn.

Sometimes it's better to make the AI a little bit less complex, and compensate with bonus, just to make the process faster. Don't know if that is the case for Civ VI AI.

Yes although that's mostly true of the tactical AI (and builder logic). Most of the city production, diplomacy, policy selection etc are not that intensive in computation. It's harder to make a good algorithm here. Mostly because it's not easy for the coder to know himself what exactly the AI should do in most situations (besides some obvious situations). But running it is pretty fast unless you overcomplicate it.

On the other hand moving units and combat is completely the reverse. It's pretty easy to know what the AI should do but it's very intensive in computation the more you ask for tactical checks. Running the pathfinder again and again in loops is what takes the most time during AI turns.

I don't mind the AI getting bonuses, but I hate that means there are things I simply cannot do, for example, the player simply can NOT build the Great Library, the Great Lighthouse, etc. No matter what, it can't be done against the AI at high levels (without the most extreme of favorable conditions) because of it's cheats. Makes me feel like those wonders shouldn't even be in the game then.

Yes. It's mostly a design problem where instead of having bonuses making you feel in danger and at risk for the whole game they just give the AI a headstart until you catch up and blow it up again.

 

[KrikkitTwo]

Lol.
Yes. It's mostly a design problem where instead of having bonuses making you feel in danger and at risk for the whole game they just give the AI a headstart until you catch up and blow it up again.

Rubber band bonuses would probably help here.
As the (average and/or best) human player does better compared to (the average and /or best) AI in some feature...All AIs get bigger bonuses to that (not things the player sees, so combat bonus would be constant...but production, growth, tech, civic bonuses could fluctuate as the human player does better or worse)

So you would never "pullahead" of the AI on Deit (as I get more techs, their techs get cheaper...As my empire gets bigger, their troops/districts/population gets cheaper)

 

[Acken]

Kinda. Not really sold on bonuses depending on human actions.
There are 2 ways to do this imo. Either you give varrying bonuses to the AI to tune it to a specific strength on a per era basis (which is what CBP does) or you make sure there is always at least a strong AI in the late game due to it eating its neighbour (this is more the approach I took for my own mod). The later feels more natural but it also frustrates some player to have only like 3 civs left at the end.

 

[treadwin]

Rubber band bonuses would probably help here.
As the (average and/or best) human player does better compared to (the average and /or best) AI in some feature...All AIs get bigger bonuses to that (not things the player sees, so combat bonus would be constant...but production, growth, tech, civic bonuses could fluctuate as the human player does better or worse)

So you would never "pullahead" of the AI on Deit (as I get more techs, their techs get cheaper...As my empire gets bigger, their troops/districts/population gets cheaper)

I would absolutely hate this. From a game-design perspective, for a game like this, I would imagine this would be an absolute no-no. There would be no reward for good play in the game.

 

[Magil]

I would absolutely hate this. From a game-design perspective, for a game like this, I would imagine this would be an absolute no-no. There would be no reward for good play in the game.

I tend to agree. While it might "work", it would feel all wrong for a game that's competitive like Civ. And make no mistake, Civ is a competitive game, you start with X Civs but there can only be one winner, there are no joint victories and everyone who doesn't win loses. "Rubber-banding" would just irritate me to no end.

 

[stealth_nsk]

You're making this a lot more complicated than it needs to be. Yes, the AI has a lot of different algorithms to run and factors to consider when it's making a given decision. But quite often it can be boiled down to a few easily visible factors. Again, I will point to things like diplomatic modifiers and agendas. You know why Victoria sent her settler to another continent to settle--that's her agenda. You know why she denounced you, because of a warmonger penalty.

There's a big difference between effective ai and visible ai. You actually picked a good examples, because both agenda and reacting to warmonger penalties are actually things which make AI less effective, they are added for visibility. Parameters which are taken into account by the "effective" part of the AI are way more complex.

 

[Grotius]

Also, remember that the AIs are programmed to role-play, which requires them to play sub-optimally in some circumstances. England is required to like civs on their own continent, even if that's not the best strategy, and to dislike civs abroad, even if that's suboptimal. Sometimes it might be in their interests to backstab the player, as we humans do without a second thought, but their personalities may not permit this. It's rather like the AI in World of Warcraft: if it played optimally, it would target the healer and not allow itself to be taunted by the tank. But that would annoy us humans because we'd view it as unfair or unrealistic. Soren Johnson gave a talk about this once, called "Playing to Lose":

There's also the massive challenge of coding an AI to play a game on a map that is many times larger than a chessboard and whose terrain configuration changes with every game. Someone said earlier it wants the AI to pursue a victory type. Heck, I'm not always capable of doing that as a human player. For that matter, even tactical decisions sometimes confound me: Which tech next? Which tile to develop? Which tile for the next city?

I'm not saying I don't want a challenging AI. Of course I do. I just keep my expectations in check -- and I role-play myself. E.g., I often refrain from ever declaring war. I know I'm not the strongest player, but that one house rule makes the game plenty challenging for me on harder levels in Civ V.

 

[dturtle1]

In regards to Rubber-banding it does have a fatal side effect of disincentivizing good play. Why play well when you can just play poorly and achieve the same result? You could of course make it affect the A.I only but that still gives you a situation when playing not quite as well as you can is better than playing to the best of your ability. There will be cracks and creases in the rubber banding mechanic were playing at 90% efficiency is better than playing at 95 % efficiency etc.
Secondly, because rubber-banding is reactive their will always be an issue of response time. It encourages you to "play down" to your competition and then pull away right at the end to win. Rubber-banding has been implemented in car games for years without giving a satisfactory result.

In saying that A.I only boni has its own issues as well as it almost always results in parts of the game being locked out from the player due to balance vagaries. Not all game mechanics fit the same scaling mould (if they do they are invariably shallow and boring) and these rough edges work directly against simple, scalable boni. You cant have complex mechanics and try it scale the difficulty using simple boni without losing the available complexity of the mechanics. it Is like taking a highly detailed picture and then trying to reduce the color depth, you will lose graphics quality.

Maybe a workable solution involved a hybrid approach. Give the A.I boni that make up for their shortcomings but also give the A.I boni soft catch up mechanics to help the A.I respond to situations outside of its programming. maybe something like a Civ who is playing peacefully gets a free Archer Unit every x turns (up to maximum of say 3) if attacked, or If the Starting capital doesn't have access to fresh water it gets a free Granary or at a reduced cost. Little limit breakers that help alleviate the flaws inherent in "Build Queue" A.I design. The hardest thing to programme is an A.I that can adapt to the unexpected, the 5% chance outcomes. By looking at trends over a large sample areas however you can implement "stop loss" techniques that mitigate against these effects, which dont actually require the A.I to adapt to be successful.

 

[skyclad]

It's mostly a design problem where instead of having bonuses making you feel in danger and at risk for the whole game they just give the AI a headstart until you catch up and blow it up again.

I agree that this is what is really the problem here! Its no fun when the AI gets a ton of free units and tech right at the start but still cant keep up.

Which is why I think AI bonuses should probably scale with era, or something like that. Previous civ games had more rubber band mechanics, such as techs being discounted when others have discovered them, or trade routes giving science based on how many techs behind you were. All that's left of that seems to be tech stealing, which has also been "nerfed" to not be the whole tech but simply the eureka moment.

Well, I have to try the game first. But I think its a mistake to front load the AI with bonuses right from the start and not help them in the late game. It means if you survive the early game and come out ahead, you will coast to victory. But to get there, you have to play very specific/abusive in order not to get crushed by the AIs early game boosts.