What aspect of Civ should Firaxis most improve?

**Master Zen** · June 30, 2004, 23:56

I have to say it: AI.

The AI is terrible, and has always been terrible in every Civ game, the only reason the harder difficulty levels are hard is because of the awsome handicaps and cheats they are allowed. I, for once, would like to see an AI capable at military strategy with good combined arms coordinations, launching of large naval and ground invasions etc. I would be quite willing to keep a simplified combat model if and only if the AI is drastically improved.

I have included other aspects of Civ to vote upon, but in case I forgot one, just post it.

-MZ

**Jonny** · July 1, 2004, 00:07

I agree entirely. The AI is the most important part of strategy games like these (unless you play only MP), and Civ's AI needs some improvement. I'll take a better, more strategical AI over added complexity any day, and I'm sure a lot of other people would too.

I will add, though, that scenario editing capability was a close second choice for me in this poll. That's another area that needs some improvement over Civ3 as well.

**Skanky Burns** · July 1, 2004, 00:21

What a pointless poll.
You do realise who is the lead developer for cIV, don't you?

**Vince278** · July 1, 2004, 00:31

In order of preference:
1. AI
2. Economics
3. Techs
4. Editing
5. Cultural

The rest doesn't matter to me much (except for the banana

). If you have to choose between cute graphics, awesome sound, or a more robust AI please choose the AI always.

**Robert Plomp** · July 1, 2004, 02:57

The cIIIv AI is one of the best AI's you'll find in any game available.
Do you have any idea how hard it is to program an AI for complex games like civilization?

I hate it that it needs to cheat, but there is no other option with the current computers.
For a simple game of chess IBM needed to create the huge big-blue computer that was able to beat the best chess-player. And that was most of all because big blue was filled with millions of former chess games so it could refer to those games for what to do.

That's impossible with civ, since civ is different every game.

The AI is good, and I surely hope that cIV will have an even better AI, but claiming that the cIIIv AI is worthless does no justice to Soren.

Go Soren, you're the man

Culture/Religion! That's the thing to focus on.

**Master Zen** · July 1, 2004, 03:07

The Civ3 AI is a definite improvement over the Civ2 AI that I agree wholeheardtedly.

But it pales in comparison with other more complex games, particularly wargames.

**Robert Plomp** · July 1, 2004, 03:26

There aren't many complex wargames that can compare to civ. Of course, simply war is simple enough. Take simple RTS games.

But civ is much more than war. The AI must chose between war, development, diplomacy, culture and economy. War is just a small part of the game.

**Vince278** · July 1, 2004, 05:20

I hope I never implied the AI is worthless but I'm sure we all can agree there is room for improvement and it is one of the most important parts of the game.

**Robert Plomp** · July 1, 2004, 05:48

Not you, but Master Zen did so:
"The AI is terrible"

**Urban Ranger** · July 1, 2004, 06:02

Originally posted by Master Zen
The Civ3 AI is a definite improvement over the Civ2 AI that I agree wholeheardtedly.

But it pales in comparison with other more complex games, particularly wargames.

Wargames have limited options, like chess. Computer programs are good at chess.

Strategic games such as Civ are like Go. Computers suck at Go.

**Brent** · July 1, 2004, 13:07

I vote Scenario Editing, not so much because of any specific way to improve it (I am very satisfied with the C3 editor), but because it's what I care about the most. One specific way to improve it would be AI to analyze the scenario you've made.

Things you didn't list: Quantity of premade scenarios. Quantity of premade civs.

**Dauphin** · July 1, 2004, 13:27

An AI will never be good enough. If there are simple ways of improving it, by all means the additions will be welcome, but I wouldn't focus attention on it beyond implementing tried and tested models and using ideas from other games.

**Master Zen** · July 1, 2004, 13:41

Originally posted by CyberShy
Not you, but Master Zen did so:
"The AI is terrible"

And I stand by my words.

- The AI has no clue on how to use artillery
- The AI has no concept of concentration of force
- The AI cannot use effectively the slider
- The AI cannot stage invasions involving more than one transport
- The AI is easily tricked via it's own cheats (ex. leave a city undefended as a lure and it'll go after it even if it can't see it)
- The AI sucks at workforce optimization
- The AI sucks at tile improvement

If this isn't a terrible AI, I don't know what is.

**Drachasor** · July 1, 2004, 13:57

I have to agree with Master Zen, the AI completely stinks. It can go up against even a fairly green player on even terms so long as the green player can defend themselves.

While the Civ III AI is a bit better at defending itself, the AI still isn't that great.

Of course the biggest change the AI could use is in the concentration of force and proper use of transports for an invasion (as Master Zen touched on). If it could just manage to attack correctly and with numbers, then it would be a more fearsome opponent. As it is now, the difficulty system completely relies on cheats, and everyone knows the no-cheat difficulty level is a joke. I'd like to see this changed.

As for comments about it being too difficult to program the AI, bah, I say. The AI can already find targets to go after, so how much harder would it be to teach it to wait until it has a sufficient force gathered before it attacks? The same goes with the transport issue. That alone is a massive issue.

Worker useage really stinks too, and part of that is because it is a bit unclear what the optimum strategy is to use workers, but the problem isn't insurmountable, the AI could still be improved here. (If I may toot my own horn, then the thread I started a few days ago--my Crazy Idea--would probably aid worker usage. Specialized cities means it is easier to know how to improve them).

I know, and I am sure Master Zen knows, that a perfect AI isn't feasible. Indeed, I am sure that at the higher difficulties there will still be some level of cheating. However, that doesn't mean that you couldn't get rid of the blatantly stupid and harmful things the AI does to itself. Even better, design an AI that will make some stupid and self-hurting decisions at lower difficulties, but these gradually disappear as it gets tougher.

All this would merely require time debugging the AI, finding out what is causing the stupid behavior, and making an algorithm to fix it. Getting rid of really stupid AI behavior that is caused merely by 1 or 2 factors, is not rocket science. I admit the AI would still probably have some stupidities not easily resolved, but I see no reason why most of Zen's list couldn't be fixed.

-Drachasor

**Vince278** · July 1, 2004, 23:35

If most of what Master Zen lists is implemented we would see a big improvement. I'd also like to see the AI have more personality and to be more human-like in actions.

**Tassadar500** · July 1, 2004, 23:46

I'd like to be able to think, where is Rome (or whatever civ) going to go next? Instead of, oh, they're just going to go to this city like always. Which means, I want the AI to know different possible strategies.

**asleepathewheel** · July 1, 2004, 23:51

better ai with fewer bonuses and cheats. Is it to be in xml as well? I would love to be able to download ai mods that would mimic other players. And I want ai civs to function more differently ala smac. Civ3, they play out similarly, just a bit more or less agressive. I want the civs to feel different. tighter spacing, looser spacing, heavy ag,heavy mine, etc etc.

**Robert Plomp** · July 2, 2004, 00:29

Master Zen, you have valid points.
There is for sure room for improvement.
But the AI is very good, for sure compared to other AI's in other games.

If you want to see an AI that plays like a human player.... forget it, you will not see it before your grandchildren have beards.

But I surely hope it'll improve from cIIIv.
But I can't say that the cIIIv AI sucked since Soren did a very good job on it.

**Urban Ranger** · July 2, 2004, 04:15

I keep forgetting.

Make the next game play more like an empire builder instead of a-collection-of-cities builder.

That means shift certain things up to national level.

**Urban Ranger** · July 2, 2004, 04:20

Originally posted by Drachasor
As for comments about it being too difficult to program the AI, bah, I say. The AI can already find targets to go after, so how much harder would it be to teach it to wait until it has a sufficient force gathered before it attacks? The same goes with the transport issue. That alone is a massive issue.

It's not that simple.

Suppose you find a target. The next question is, how much is a sufficient force? Then, you have to decide whether you have that available. If not, decide whether you want to switch the production of some cities to army units.

It's not like just moving N-K4.

**Silpy** · July 2, 2004, 07:09

I have three main compliants with the AI

1) They don't understand when they have old units. By 2000 AD they usally still have masses of Warriors and Chariots.

2) They don't seem capable of serious naval/air threats. I've never faced an AI with a airforce of more than 3-5 units.

3) They don't seem capable of using artillery.

The AI in Civ II seem better in use of air units than in Civ III!

Do other players come to the same concusions?

**Dominae** · July 2, 2004, 07:23

I vote AI, but the Tech model for me is a very close second.

Research in Civ3 is silly because it's so easy to trade for techs. This means that it's not really important what you research, so long as you have something to trade to the AI at every few turns or so. The decision to research a tech for the abilities/units/etc. it unlocks is typically irrelevant, since you can just trade for that tech when the opportunity arises (and it always does). Ever get the the feeling that the Ancient era just blew by in a couple of turns due to the AIs swapping scientific secrets? And please do not get me started on how you can just go to war if you find yourself lagging behind in tech.

I would like to see a game mechanism that restricts the power of tech trading (including extortion). This would make the whole research part of the game a lot more fun. Any ideas?

**Drachasor** · July 2, 2004, 09:32

Originally posted by Urban Ranger

It's not that simple.

Suppose you find a target. The next question is, how much is a sufficient force? Then, you have to decide whether you have that available. If not, decide whether you want to switch the production of some cities to army units.

It's not like just moving N-K4.

If the AI cheats and knows what you have in and around the City (as it does currently), then it should be pretty darn easy, now shouldn't it?

Otherwise it should do what any normal player would do, assume a somewhat tough fight against 3-4 units, and build an army large enough to supply 2-4 defenders after you take it, and give at least an 80% chance of victory (against the presumed opponent). Of course, there's no reason why they couldn't use the spy system to find out what's in a city.

As for switching production to military units, that shouldn't be too difficult, though production is another area that needs improvement. It does need some sort of priority queue, but if it doesn't have the troops to take a city, then it shouldn't be attacking (even if it can't figure out how to change its queues). It would be better off holding on to its current force for defensive purposes.

-Drachasor

**punkbass2000** · July 2, 2004, 11:33

Originally posted by Drachasor

If the AI cheats and knows what you have in and around the City (as it does currently), then it should be pretty darn easy, now shouldn't it?

This is a ridiculous question. I suspect you don't know much about programming. IF that is not the case, then maybe you should take on such an 'easy' task.

**Drachasor** · July 2, 2004, 12:15

Originally posted by punkbass2000

This is a ridiculous question. I suspect you don't know much about programming. IF that is not the case, then maybe you should take on such an 'easy' task.

So, you are saying that if the AI knows all the units in a city, then it will not be able to do a simple calculation to find out the sufficient force to achieve an 80% chance of victory (or higher)? That's is what is ridiculous. There mathematics involved are straightforward, and if you somehow think computers can't handle that, then you're just plain wrong.

For instance if 2 of Unit B is sufficent to defeat a fortified unit A in a city 80% of the time, and there are 4 unit A's in that city, then 8 unit Bs will be sufficient to defeat the force in that city 80% of the time (we'll assume the combat is one round, but multiple rounds only change the potential force involved--which can also easily be calculated if the AI assumes the opponent will switch everything to the defense of that city, that is the AI plays it safe).

Are you somehow telling me the computer can't do the combat calculations a few million times in a few seconds? Because you are quite wrong if you think that. I may not be an expert at programming, but I know my bit.

As I said, the switching of production in the AI cities would be a bit tricky (to balance), but it naturally isn't impossible to get a so-so to decent implementation. Heck, even keep the current one, but have it take into account numbers before it attacks. How the heck can you think the AI can't figure out its chances?

-Drachasor

**Japher** · July 2, 2004, 12:20

"What aspect of Civ should Firaxis most improve?"

The manual and box art

As I know nothing about AI, being only a post bot, I will go with City/Tile improvement

**punkbass2000** · July 2, 2004, 12:34

Originally posted by Drachasor

So, you are saying that if the AI knows all the units in a city, then it will not be able to do a simple calculation to find out the sufficient force to achieve an 80% chance of victory (or higher)? That's is what is ridiculous. There mathematics involved are straightforward, and if you somehow think computers can't handle that, then you're just plain wrong.

For instance if 2 of Unit B is sufficent to defeat a fortified unit A in a city 80% of the time, and there are 4 unit A's in that city, then 8 unit Bs will be sufficient to defeat the force in that city 80% of the time (we'll assume the combat is one round, but multiple rounds only change the potential force involved--which can also easily be calculated if the AI assumes the opponent will switch everything to the defense of that city, that is the AI plays it safe).

Are you somehow telling me the computer can't do the combat calculations a few million times in a few seconds? Because you are quite wrong if you think that. I may not be an expert at programming, but I know my bit.

As I said, the switching of production in the AI cities would be a bit tricky (to balance), but it naturally isn't impossible to get a so-so to decent implementation. Heck, even keep the current one, but have it take into account numbers before it attacks. How the heck can you think the AI can't figure out its chances?

-Drachasor

Look, as has already been stated by several people, there are many, many more factors than what you are suggesting. Just for starters, what if amassing this 80% victory chance requires the deployment of all the military units the AI currently has?

There is clearly more to this than a simple combat calculator. IF that were not so, do you honestly believe SOren would not have implemented this ages ago?

**Drachasor** · July 2, 2004, 12:56

I don't know why it hasn't been implemented, but when over the course of 10 or so turns the AI attacks the same city 5 or so times with one unit each time, then something is going wrong.

And yes, there are other factors, and certainly the AI already considers some units as relatively free to move about and others as not very free to move about (the later being city defenders). It seems to me that if the AI finds that it can't attack the city with sufficient force to take it, then it should probably not attack (especially if the attack requires declaring war). If it has decided it should really take that city or that it should go to war with you at some near point, then it should start building units for war. Certainly the AI shouldn't be attacking an city that can heal all of its troops in 1 turn if it doesn't think it can destroy most of the defenders....yet the AI DOES do this. That is something that isn't hard to stop, but could quite possibly dramatically improve the game. Suddenly the AI wouldn't be losing troops for stupid reasons, which means it would build up a sizeable force (for it is not tossing them away like garbage), and then eventually it would attack with that force.

Clearly something is currently a bit off, and it wouldn't be too hard to fix. Now, perhaps the reason why that isn't there is because there ARE a lot of other things to work on in the AI, and so sometimes less complicated details are missed. Afterall, if things can be "balanced" by making the AI cheat to a fantastic degree, then proper unit usage and other things can be ignored. I am merely proposing they aren't ignored to the same extent.

Clearly already the AI has some definition of the minimum number of troops it should have, all it needs now is to consider its war footin a bit more, and devote more resources to a war if it deems that war will be difficult. More complicated than current implementions? Yes, but it isn't impossible. It does require the AI to have some means to consider what city improvements and what cities are more important relative to when it can get an army out. I don't expect a perfect implementation of this, but I think it could be improved quite a bit.

-Drachasor

**Brent** · July 2, 2004, 14:46

AI with personality would be good.

I want the manual to have more documentation on the editor and premade scenarios.

**Jon Shafer** · July 2, 2004, 14:48

I voted AI, but Economics is close behind.

The AI's great flaws only appear when you become very good at the game. The vast majority of players "out there" who have never heard of Apolyton play at Chieftan and Warlord and get a good run at their money.

But I'm greedy, and the fact that I am very good at the game and I do enjoy playing it means that I would like more of a challenge to keep me busy.

The AI will be exposed to scripting, which means there ought to be a chance for some pretty heavy editing stuff, allowing for custom AIs that can be tweaked. What really matters is how much is exposed to that scripting - the more the merrier. And the better the community can make it.