The AI is too bad

df1_2

Chieftain
Joined
Feb 23, 2006
Messages
44
I play the world 1000 AD scenario, and its like impossible to loose.

As Saladin and China I play on the toughest setting without any problem of winning.

Yesterday I tried India on the second hardest setting... I was behind in score and especially in research the whole game to most of the other leaders, especially the european ones.

I played agressive and focused much more on military than research (as always).

And because the AI makes so stupid military decision, there was no problem beating down the leaders that had much better military.

What do you think, do we need a harder setting to make up for the poor AI?
 
MrCynical said:
Can you beat the toughest setting (or anywhere near it) on a random map?

Probably not, havent tried to much, but I think its the same. They start out with a big lead, but once you start fighting they play stupid.
 
The AI does make very poor military decisions. The AI usually sends its stack of doom to attack my well defended cities instead of my captial that's near a coast with only a warrior defending it in 1900.
 
Ronin228 said:
The AI does make very poor military decisions. The AI usually sends its stack of doom to attack my well defended cities instead of my captial that's near a coast with only a warrior defending it in 1900.

Yes, I see crazy stuff, like attacking my stack with a infantery before artillery.

And suddenly attacking my fortified unit on a hill (protecting no resource) with 75 % defence bonus, where he must have a combat odds of like 0.2 %.

And also, they can declare war and then sit back with 90 % of their army in their cities, and fight with a harmless army at the front.
 
df1_2 said:
What do you think, do we need a harder setting to make up for the poor AI?

No, we need better AI. Merely ratchetting up the difficulty level means I'm not playing a wily opponent that intrigues me, but pre-set conditions, and I find that boring.
 
It's easy to demand a better AI, but, especially with a game like Civ, programming one is an entirely different story.
 
df1_2 said:
I play the world 1000 AD scenario, and its like impossible to loose.

As Saladin and China I play on the toughest setting without any problem of winning.

Yesterday I tried India on the second hardest setting... I was behind in score and especially in research the whole game to most of the other leaders, especially the european ones.

I played agressive and focused much more on military than research (as always).

And because the AI makes so stupid military decision, there was no problem beating down the leaders that had much better military.

What do you think, do we need a harder setting to make up for the poor AI?

Saladin and China Are Designed to be easy (the Default difficulty for Saladin is Deity) Try playing as one of the European powers on Deity Level... or the Incas
 
Krikkitone said:
Saladin and China Are Designed to be easy (the Default difficulty for Saladin is Deity) Try playing as one of the European powers on Deity Level... or the Incas

I'm playing this scenario as the Holy Roman Empire right now. It's my first full game, and it hasn't been too hard yet. I took the Ottoman Empire, the Arabs and the Indians. I have removed the Russians from Eurasia (though they have 1 city in South America), and I'm in the process of eliminating the French. It's 1955 and I have a pile of Panzers zooming through the remains of the French settlements in Northern Central Asia. They should be finished in a few turns. Then they can rot in exile on Madagascar. MUHAHAHAH!!! :lol:
 
Ronin228 said:
The AI does make very poor military decisions. The AI usually sends its stack of doom to attack my well defended cities instead of my captial that's near a coast with only a warrior defending it in 1900.

If I recall, the Civ 3 AI originally knew which of your cities was the weakest defended, and launched its attack there. But so many fans complained, they patched Civ3 so it didn't have that info. I don't think the AI ever recovered. It sucked ever since. Personally, I didn't mind the AI knowing my weakest city and attacking it. It prevents people from using "the one warrior defense" as you put it.

My guess is that so many fans complained, that they deliberately make the AI very easy to beat in battle. Let's face it, people like to win this game. Wars take a long time. Most people can't handle investing a hour or two in a Civ War and not coming out ahead.

So, instead, the AI gets research bonuses and starts out with extra units.
I liked the old way, where they tried to attack your weakest city.
 
If I recall, the Civ 3 AI originally knew which of your cities was the weakest defended, and launched its attack there. But so many fans complained, they patched Civ3 so it didn't have that info. I don't think the AI ever recovered. It sucked ever since. Personally, I didn't mind the AI knowing my weakest city and attacking it. It prevents people from using "the one warrior defense" as you put it.

The snag was that this too could be exploited. You could deliberately leave a city far behind the frontline undefended, and the AI would march its stack past border cities that, while more defended that your weakest city, it could easily capture. This would result in its troops marching through enemy territory for 10 turns getting picked off, while achieving nothing.

The AI in Civ 4, even in combat, is better than in Civ 3. That is to say it now has the intelligence of a slightly less ******** ant than it used to. It could certainly still use some more improvement, and while it is very difficult to program an AI for a game like Civ, I think there are a few improvements that ought to be fairly simple. For example, getting the AI to actually use its units rather than having them all sitting in cities while a comparatively small enemy force pillages it back into the stone age. Or to actually build up a decent attacking force if its first one gets wiped out, rather than sending in its units a few at a time as they get built.

The Earth 1000AD scenario really isn't the best one to judge the AI's competance on though, given how unequal the starting situations of the various civs are. I'd try a few random maps before claiming Deity level presents no problem.
 
bfordyce said:
If I recall, the Civ 3 AI originally knew which of your cities was the weakest defended, and launched its attack there. But so many fans complained, they patched Civ3 so it didn't have that info. I don't think the AI ever recovered. It sucked ever since.

I hated the AI knowing where your units were. It was a prime example of trying to improve the AI on the cheap. (Improve the AI war tactics? - nah, we'll just let it know where all the human player's units are.) I agree that it needs this sort of advantage, given how stupid it is.

I have a theory about Civ4: The biggest problem the AI has is cultural defense. Siege weapons are critical to reduce cultural defense and weaken city defenders, but the AI doesn't grasp that. That's why it's rare for the AI to war effectively in Civ4. Logically, they either have to fix the problem or do away with cultural defense.

Fixing the problem would be the hardest. Doing away with cultural defense would be easier, but even easier is... give the AI more advantages at higher levels! Yeah, that's the ticket. ;)
 
I hate how when I play Civ III the computer knows what I'm doing. A lot of warfare is manouvre and surprise. It really annoyed me the first time enemy forces swooped in out of nowhere to attack my force.

In Civ IV, the AI is better. Far from perfect, but it's better. I can live with that. And I've seen plenty of bombardment.
 
The A.Is bombard plenty, my suggestion is to download one of the many ods here on CFC, those tend to make the game much harder(more complexity) on the higher difficulty levels.
 
The AI usually only bombards with ships. They never park a stack on ahill across a river and bombard the crap out of you until there is nothing left and then send their massive army to an adjacent space across the river and send there former bombradiers on kamikaze attacks and then finish you off with a hail of units. Would that be considered a "run on" sentence?
 
anyone ever seen an AI bomber? I havent. But at least the AI units cross rivers instead of attacking across them, or is that because I increased river defense to 50%? I do think that the civ4 AI knows where your workers are. Too often they walk by my defensive hills and aim straight at my workers who are deep in my territory and thus should be unknown to the AI.
 
The AI is getting better, but it isn't perfect. My reasonable ideal would be to have a different AI for each civ, and let the players tweak those AI's and post their new files for player review.

The advantage of this kind of system is that as people play the game and learn the strengths/weakness' of the units the AI can keep up with the different strategies.

Different AI's could also use different types of algorithms and/or advantages. One AI might value different techs and make a judgement call, while another might follow a "tech path" programmed by a player, and a third might cheat civ III style.

If the AI for team X is randomly chosen from a "Team X" AI list available to all players each game . . . then computer could get a lot smarter and become very unpredictable.
 
I think the AI in CIV is much improved over the earlier games in the series. It's far from perfect, but CIV is a complex game to code AI for.

In one of my first games of CIV, an AI on a separate continent declared war on me. I figured it would end up being a faux war, since I had no intention of attacking that tribe - I hadn't built up enough of a fleet yet (was still middle ages). I was pleasantly surprised a couple of turns later to see a large invasion force had been dropped off, and the AI was starting to bombard my city defense down while pillaging around the city. I had been so overconfident from Civ III that I never expected an invasion!

In a recent Earth game with 18 AI's, I am playing as the Greeks and struggling to keep up with tech. In the late classical/early mideival period, I was attacking Saladin and my main army was on the Arabian peninsula. Catherine took that moment to attack. She had 3 land options of invading my land...she tried the first one, and I stopped her. She withdrew, and attacked over the second path...where again I stopped her. Again, Catherine withdrew, and launched an attack over the third path. Eventually peace was made without me losing any territory, but I was impressed at Catherine's tenacity and willingness to withdraw and try somewhere else when I stopped her forces.

Of course, that didn't stop Saladin and Caeser in that game from acting like complete idiots and getting obsorbed into my Hellenistic Empire. When I get back to that game...Catherine is next!
 
MrCynical said:
The snag was that this too could be exploited. You could deliberately leave a city far behind the frontline undefended, and the AI would march its stack past border cities that, while more defended that your weakest city, it could easily capture. This would result in its troops marching through enemy territory for 10 turns getting picked off, while achieving nothing.

Interesting. I didn't know about that exploit. I just remember lots of people complaining on the net that they only left one warrior defending a city on the coast, and they were crying because the AI landed all its ships there when they "shouldn't have been able to know" that city was weakly defended. :lol:
So, I thought it was a fix to appease the whiners. I guess they need a hybrid approach. I wouldn't mind the AI knowing the defense of every city and then choosing it's battles taking into account defense strength and travel time.



MrCynical said:
The AI in Civ 4, even in combat, is better than in Civ 3. That is to say it now has the intelligence of a slightly less ******** ant than it used to. It could certainly still use some more improvement, and while it is very difficult to program an AI for a game like Civ, I think there are a few improvements that ought to be fairly simple. For example, getting the AI to actually use its units rather than having them all sitting in cities while a comparatively small enemy force pillages it back into the stone age. Or to actually build up a decent attacking force if its first one gets wiped out, rather than sending in its units a few at a time as they get built.

You know what I'd like to see the AI do? When the human player starts amassing a stack of units on the border, it should respond by shifting a lot of units to that front. It's too easy to make a stack of 6-8 axmen right on the border of an AI player, and he does nothing about it. Then you have an easy invasion. I think the AI also needs to adjust to the specialization of the new units.. For example, in one war, the AI had 6 archers fortified in a city. It send out 2 of them in a weak attempt to pick off one of my stray axmen (as a stack of axmen was being created outside the city). Obviously, the best move is to leave those 2 archers there and let them defend against the invasion that is about to happen.


MrCynical said:
The Earth 1000AD scenario really isn't the best one to judge the AI's competance on though, given how unequal the starting situations of the various civs are. I'd try a few random maps before claiming Deity level presents no problem.

True, I'm still on emperor. I haven't tried levels harder than that yet.
 
bkwrm79 said:
I hate how when I play Civ III the computer knows what I'm doing. A lot of warfare is manouvre and surprise. It really annoyed me the first time enemy forces swooped in out of nowhere to attack my force.

What's totally ridiculous is that when you are trying to sneak around with a supposedly stealth unit like a submarine, the Civ III AI still knows exactly where it is and beelines a ship right over to it to sink it.

The biggest improvements I've seen in the Civ IV AI are:
  • Looking at combat odds before attacking and no longer attacking cities when the odds are ridiculous.
  • Keeping the SOD together. The Civ III would send out combined arms but units still traveled individually towards the target at whatever their movement rate happened to be. Fast attackers always ended up out in front while the slower defenders were a couple tiles behind.
  • Using artillery, although still not enough.
  • Pillaging
 
Back
Top Bottom