Why does all Civ AI suck?

[Ondskan]

Honestly, why!?

There's other games with better AI and more complex gameplay like Hearts of Iron 2 for example.

Seriously. CIV AI can't stack units, doesn't know how to use armies, can't use artillery properly, naval invasions are often bad I hear, they accept peace to often and declare war at bad times.

etc etc.

And ontop of that its hardcoded so you can't go in and change it without breaking a bunch of laws and hacking the hell out of the game.

I guess Civ 4 is better, especially in the diplomacy department but still!

How could they settle with such a bad AI? I'm soooo sick of the difficulties just giving more and more bonuses to the useless AI until it could be replaced by a 3 year old and still beat you.

Moderator Action: Changed title. (inappropriate language)
Please read the forum rules: http://forums.civfanatics.com/showthread.php?t=422889

 

[warpstorm]

In Civ IV all of the AI code and data it uses is there for an enterprising person to change (using the SDK). Firaxis made this fairly easy (and perfectly legal) to mod if you want to (without hacking up your original game - your mod can coexist with the original game). Go for it.

 

[potatokiosk]

There's other games with better AI and more complex gameplay like Hearts of Iron 2 for example.

The HoI2 AI isn't that great. If it had to deal with random maps it would probably be just as bad as the civ AI.

 

[Till]

1. The AI doesn't need to be good in order to satisfy the average gamer. It is supposed to be a challenge, but it doesn't have to play to win.
2. You can't finalize the AI until the gameplay rules have been set, which tends to happen rather late in development leaving little time.
3. Most games are too complex for machine learning to be used on a large scale, making the AI very predictable.

 

[Maniacal]

4. Programming complex AI is not easy.
5. Deadlines and limited time.
6. People can also be lazy and/or incompetent.

 

[Dachs]

The HoI2 AI isn't that great. If it had to deal with random maps it would probably be just as bad as the civ AI.

This is a good point. Random maps can be a bit difficult to program stuff for.

 

[Sub]

7. They sold millions upon millions of copies of Civ IV so clearly amazing AI isn't needed to make money, hence little incentive to improve it

 

[azzaman333]

It's much easier to program a semi-competant AI and give it bonuses to make it harder than a genuinely good AI.

 

[Ondskan]

7. They sold millions upon millions of copies of Civ IV so clearly amazing AI isn't needed to make money, hence little incentive to improve it

Yep. People haven't learned to truly vote with their money and influence their new rulers (the corporations). People should learn this soon or perish. Or revolt! Vive la revolucion!



Really though, the games (CIV) themselves are awesome so people buy them.
But yeah the AI is horrible. I think what gamingcompanies aren't realising is that more and more people pirate not only to skip the whole teadious thing of "paying" for something but also because few games nowdays are worked really hard on. They are actually rushed, it's still fun to play them but they have nowhere near the value of some older games.

Still CIV games aren't part of that category and ithey are definetly worth buying. It might be easier to create a semi-competent AI but its definetly possible to create a competent to. Just takes more testing.
CIV 5 better have a good AI or I'm not buying it.
Granted CIV 1-3 weere still unique concepts but today there is no need to further develope the concept as before. The need is for real improvement.

The AI doesn't need to be good in order to satisfy the average gamer. It is supposed to be a challenge, but it doesn't have to play to win.

Oh god I hate this.
I think they took this a step further in SPORE and a few other games.
They made a sort of selfaware AI. But not in the good sense, that it learns and developes.
But in a bad sense:

Depending on if you play bad or good the AI will become worse or better...in the end always addapting so that you can beat it.
Instead of YOU rising up to the challenge and trying to beat the AI.

What the hell are they thinking?

3. Most games are too complex for machine learning to be used on a large scale, making the AI very predictable.

Definetly not. Todays machines can handle it and todays understanding of shortcuts in AI programing is high. Sure, shortcuts aren't as good as: React to Y if enemy does X but they are still better than a nearly unresponsive AI like the CIV 3 AI is for example in battle. In diplomacy it is responsive.
But in battle it's not. It doesn't actually respond to anything more than if you have a big or small stack near its cities otherwise it has a set goal and pursues it without any strategic depth.

 

[warpstorm]

...or you can open the Civ 4 SDK and roll your own AI.

 

[Darth_Pugwash]

As others have said it's not easy to program a decent AI for a complex strategy game, among other reasons.

The AI in Civ 4 was actually pretty decent though IMHO, better than the AI in Hearts of Iron 2... if you want to see a truly woeful computer opponent check out any of the Total War games, especially Rome and Medieval.

 

[Ondskan]

...or you can open the Civ 4 SDK and roll your own AI.

Hehe I play mostly Civ 3 and would like to play some AC if the AI was better.

Civ 4 definetly has an even better diplomacy system but since its even more complex I wouldn't ever dream of fixing the AI myself.

That's something that should be done BEFORE the game is released.



With a little luck Civ 5 will be really good =) I hope for the best.

 

[Till]

Yep. People haven't learned to truly vote with their money and influence their new rulers (the corporations). People should learn this soon or perish. Or revolt! Vive la revolucion!



Really though, the games (CIV) themselves are awesome so people buy them.
But yeah the AI is horrible. I think what gamingcompanies aren't realising is that more and more people pirate not only to skip the whole teadious thing of "paying" for something but also because few games nowdays are worked really hard on. They are actually rushed, it's still fun to play them but they have nowhere near the value of some older games.

Still CIV games aren't part of that category and ithey are definetly worth buying. It might be easier to create a semi-competent AI but its definetly possible to create a competent to. Just takes more testing.
CIV 5 better have a good AI or I'm not buying it.
Granted CIV 1-3 weere still unique concepts but today there is no need to further develope the concept as before. The need is for real improvement.



Oh god I hate this.
I think they took this a step further in SPORE and a few other games.
They made a sort of selfaware AI. But not in the good sense, that it learns and developes.
But in a bad sense:

Depending on if you play bad or good the AI will become worse or better...in the end always addapting so that you can beat it.
Instead of YOU rising up to the challenge and trying to beat the AI.

What the hell are they thinking?




Definetly not. Todays machines can handle it and todays understanding of shortcuts in AI programing is high. Sure, shortcuts aren't as good as: React to Y if enemy does X but they are still better than a nearly unresponsive AI like the CIV 3 AI is for example in battle. In diplomacy it is responsive.
But in battle it's not. It doesn't actually respond to anything more than if you have a big or small stack near its cities otherwise it has a set goal and pursues it without any strategic depth.

Actually, rules like you describe are the best and most widely used "shortcuts". The problem is that you can have only so many rules before you get inconsistencies and contradictions, and that the AI is incapable of learning, making it predictable. Once you've figured out the rules, the AI stops a challenge.
That is why i was referring to machine learning. Unfortunately, Civ's state space is far too large for effective learning. You would either need to crop it, or divide it into several learning problems. The former has the disadvantage that the AI would be partially blind, making it easy to exploit. The latter has the problem that it is difficult to formulate distinct sub problems.

 

[warpstorm]

Soren Johnson (the designer and AI programmer for Civ 4) wrote an article about Civ's AI for last Game Developer Conference. He decided that the AI shouldn't be so good that it routinely crushes the player, it should be just good enough to make the player think they're facing some kind of challenge before letting them win. Would you really want to lose more games of Civ than you win? Answer honestly.

 

[Ondskan]

Soren Johnson (the designer and AI programmer for Civ 4) wrote an article about Civ's AI for last Game Developer Conference. He decided that the AI shouldn't be so good that it routinely crushes the player, it should be just good enough to make the player think they're facing some kind of challenge before letting them win. Would you really want to lose more games of Civ than you win? Answer honestly.

Hell yeah. Actually every time I am winning I quit the game because it's to easy.
If anything I'd want the AI to become more determined as I am winning.
But I'd just want it to be determined as hell all the time instead.
I hate gave developers who think that people don't want to loose.
At least give us the option for an actually difficult AI.

His reasons sound bogus and the talks of someone who's interested in rushing the game. If they have absurd difficulties such as SID and Diety then they might aswell have improved AI difficulties.

I must say though that for a communtiy that deals in one of the most strategic games there is, there is little demand for strategic depth at least when I play multiplayer. And little demand for challenge.
For example in Civ 3 multiplayer you 9 out of 10 times get whining people who if the teams are even the slightest unfair won't play or if you find a substitute thats slightly better/worse will regard the game as over. (this despite the fact that beating a harder player will get you a bigger skill advance on the ladder and a lesser skill drop if you loose against him/her).
Or if they are loosing, instead of fighting on and seeing how long they can survive they prefer to start a new game.

I am definetly not that sort of player and I hope at least that most people out there are like me.
Games were made for a reason, beyond the fun it's meant to be a challenge to rise to. The early arcade games would be games you'd try TIME and TIME over until you beat them, often with friends, competing.

 

[potatokiosk]

Soren Johnson (the designer and AI programmer for Civ 4) wrote an article about Civ's AI for last Game Developer Conference. He decided that the AI shouldn't be so good that it routinely crushes the player, it should be just good enough to make the player think they're facing some kind of challenge before letting them win. Would you really want to lose more games of Civ than you win? Answer honestly.

If the AI is improved they can just lower the AI bonuses to maintain the same level of difficulty for the player.