How weak is the AI?

[Lynxx]

Hi there, interest in civ rising again after a break since civ 4.

I have looked though a couple of lets plays and got the general impression that the game looks "pretty good". However, the one thing I find a bit alarming is how weak the AI seems to be! It builds next to no units early game and seems to struggle a lot even later in the game. I guess all the streamers have been playing on prince? But even so, is anyone else worried about the AI in this game? What information is there available on the AI except for watching lets plays?

 

[Cyberian]

The game itself is very good but yes the AI is very weak even for prince.

There is still a chance that the preview build is months old and they have a better build internally already where they massively improved the AI.
Although that is not a given. The game is more complex due to unstacking the cities, changed movement and districts and such. It is possible that the AI will be rubbish on release.

 

[King Jason]

It's probably not tuned fully yet. There are a lot of very obvious examples that are suggestive of that fact. Two things to consider - the press build is accessed by everyone from Deity players who can take on a team of 5 A.I. by themselves (FilthyRobot) to players who have practically no idea how to play civ, but their still 'press' level streamers (TotalBiscuit).

So on the one hand, Filthy will make the game look like an absolute cheesefest - even with the silly mistakes he made over the first week of play. On the other hand, the game can't actually be all that hard when you're just trying to show people how it works. Otherwise someone like TotalBiscuit might lose all of their games in a manner that prevents him from experimenting and experimenting with most of it.

It's likely the same sort of build that the devs used for streaming and in one stream they themselves said something to the affect of "Well, we can't have it so we could lose while showing off the game, that'd be no fun".

Even without those anecdotes - I'm convinced that the difficulty we're viewing is not an accurate reflection of the difficulty of the game. Whether it's not fully tuned yet, intentionally lobotomized, or both.

 

[Lynxx]

The game itself is very good but yes the AI is very weak even for prince.

There is still a chance that the preview build is months old and they have a better build internally already where they massively improved the AI.
Although that is not a given. The game is more complex due to unstacking the cities, changed movement and districts and such. It is possible that the AI will be rubbish on release.

This is the same impression I have gotten. To bad, while I do play multiplayer once in a while I would mainly be playing against the computer... I guess I will wait and read the reviews and comments here on the forum after the game has been released.

 

[Lord Yanaek]

It's really hard to tell from some pre-release press build with AI on Prince level, that being said and as much as i like the Civilization franchise ever since i discovered it with Civ3, the AI have always been pretty bad. High difficulty AI appeared strong because of silly economic bonuses (like double production,science,happiness and most of the ancient techs pre-researched) until you discover how to "cheese" it.

Now, from what i've read, AI will no longer get (as much) silly bonuses which i find great cause those bonuses forced you to over-focus every game in a similar way and cut-out plenty of options in the early game, making every game much more similar to the previous one. However, it also means they are probably experimenting with new ways to make the AI "interesting" if not "competitive" on higher difficulties. As a result, i don't expect the AI to be fully polished at release date.

Hopefully, they will manage to improve it with patches over the first few months after release but i don't have too high hopes it would be great at the time of release. Fortunately i'll be learning the game myself so a weaker AI won't be that big an issue as long as it's not totally dumb.

 

[Deggial]

The actual tactical AI doesn't seem to be so bad:
It moves and shoots, it doesn't water under enemy fire, it keeps fortified units in cities to increase city deffense & not commit suicide, it switches to unit-production under siege (even in close-by cities to help out). It seems to prefer larger numbers (if available) to attack enemy cities. Even more, some impressive "trick moves" were reported.

The above "if available" is the only real issue, accompanied with lack of unit upgrades. But increasing the numbers of built units is definitely just a matter of tweaking and *very* responsive to boni at higher difficulties. The same is true for unit upgrades. And don't forget, that more advanced units have higher upkeep. Maybe it's even a smart move to keep outdated units around and only modernize them if necessary.

 

[rreebb]

And don't forget, that more advanced units have higher upkeep. Maybe it's even a smart move to keep outdated units around and only modernize them if necessary

That's a good point. (BTW, I like how upkeep is now tied to the unit's era, not the player's era.) Although the Prince AI never did seem to actually do the upgrades when pressured.

For human players it may not be such a great strategy, except when intending to invite AIs to declare war, since the military strength is affected (and I'm sure that is still a big factor in deciding whether to DoW a player).

 

[darko82]

Very weak. The AI and the game needs to be challenging. Otherwise, it's not fun.

Basically, the gameplays clearly show that the AI is not working as supposed to. Werid behaviours are everywhere.

Its stupidity hurts my eyes.

I hope they can improve it. Or maybe they can't?

The sooner they release the DLL, the better, the sooner modders can improve the AI.

 

[ThisNameIsTooLo]

Now, from what i've read, AI will no longer get (as much) silly bonuses which i find great cause those bonuses forced you to over-focus every game in a similar way and cut-out plenty of options in the early game, making every game much more similar to the previous one. However, it also means they are probably experimenting with new ways to make the AI "interesting" if not "competitive" on higher difficulties.

The Agenda system that they introduced seems to be taking inspiration from four very old video game foes: the ghosts of Pac-Man. Scroll down to Chapter 4 of this article: http://www.gamasutra.com/view/feature/3938/the_pacman_dossier.php?print=1

Rather than designing each rival civ to individually be a threat, the developers have built up each AI player to be single-mindedly focused on one front and less cautious in all other areas; but each one will have a very different point of focus that your actions can trigger. Thus, they work together without meaning to work together; when you exploit the weaknesses of one civ, it opens you up to be exploited by another civ in turn.

 

[treadwin]

I think we are going to be surprised by the AI.

The very first release where they only got the first 150 turns, the reviewers were getting absolutely hammered by the barbarians. Then when the early play version came out, everything was much easier. I suspect that the devs know exactly how to make the game harder, they just didn't, because they were only releasing one difficulty level and they wanted to make it so that even the least-skilled reviewer could get through a game first time.

I think anyone who thinks that they are going to load up the game at emperor-level or above on release and win by 1100AD is deceiving themselves mightily.

 

[Ryika]

Combat AI doesn't look as bad as I would have expected.
Doesn't look 'good', but not horrible.

 

[darthtoon]

AI will probably be on the same level as Civ 5

 

[MkLh]

I think we are going to be surprised by the AI.

The very first release where they only got the first 150 turns, the reviewers were getting absolutely hammered by the barbarians. Then when the early play version came out, everything was much easier. I suspect that the devs know exactly how to make the game harder, they just didn't, because they were only releasing one difficulty level and they wanted to make it so that even the least-skilled reviewer could get through a game first time.

I think anyone who thinks that they are going to load up the game at emperor-level or above on release and win by 1100AD is deceiving themselves mightily.

The difficulty level has nothing to do with the AI, it's all about bonuses.

 

[treadwin]

The difficulty level has nothing to do with the AI, it's all about bonuses.

I disagree completely. When the AI is not a challenge those bonuses are easy to achieve. When the AI gets more challenging, getting the bonuses before you need to research them is much harder.

 

[gunnergoz]

I find it hard to believe the AI seen in the press build is the release version, but I may be wrong. The barbarians can be problematic and the AI is pretty savage and relentless in that department, but much of that has to do with the barbarian camps pumping out a unit per turn apparently with no restrictions. If the player survives that, the regular civs' AI behavior seems fairly tame and that makes me think it was deliberately nerfed in the press build so preview players could survive & learn without undue frustration to sour their experience.

 

[agonistes]

I disagree completely. When the AI is not a challenge those bonuses are easy to achieve. When the AI gets more challenging, getting the bonuses before you need to research them is much harder.

Not the bonuses he's talking about.

 

[Eagle Pursuit]

The actual tactical AI doesn't seem to be so bad:
It moves and shoots, it doesn't water under enemy fire, it keeps fortified units in cities to increase city deffense & not commit suicide, it switches to unit-production under siege (even in close-by cities to help out). It seems to prefer larger numbers (if available) to attack enemy cities. Even more, some impressive "trick moves" were reported.

The above "if available" is the only real issue, accompanied with lack of unit upgrades. But increasing the numbers of built units is definitely just a matter of tweaking and *very* responsive to boni at higher difficulties. The same is true for unit upgrades. And don't forget, that more advanced units have higher upkeep. Maybe it's even a smart move to keep outdated units around and only modernize them if necessary.

I agree with this. The AI seems fine in combat. I've been watching Filthy and Arumba chase a barbarian spearman around, and the thing is damned evasive. The barbarians are a good example of good tactical AI because their attacks come pre-loaded with units.

The problem that I see with the AI is two-fold.

It doesn't know how to compensate for weak starts. In Filthy and Arumba's game, Norway got a bad start and didn't seem to be doing the right thing to fix that. It was on a coast without fresh water. The AI should have made a Builder to make improvements with Housing. It overproduced Scouts and not Builders.

Secondly, the AI doesn't appear to be producing enough units or maybe not upgrading the ones it has. I see the AI with Swordsmen and Slingers way too late in streams from several different players. Or maybe the AI isn't playing to its strengths with regards to strategic resources. It wants lots of melee units, but doesn't have iron or something.

 

[skyclad]

I know in previous civ games the AI had free or discounted unit upgrades, this small thing alone might help them a lot since a big part of the problem to me seems to be their out dated units.

 

[rastak]

I'll agree with most comments. It looks quite weak but I think most are correctable for the final build or shortly afterward.

Unit upgrades and not having enough units seems very clear. Will be interesting to see what King and above look like the 21st.

 

[Wolle68]

It seemed to be kicking Ed Beach's butt on emperor level today (with bonuses, sure). It certainly knew how to build enough units.